Skip to main content
Autonomy

AI and Aerodynamics at Insect Scale: MIT’s Bumblebee‑Sized Robot

MIT soft robotics lab did an amazing job right at the intersection of physical systems, cyber and software. Small robots and insect-related autonomy have fascinated me for awhile. The autonomy of an insect brain is simple enough to just start to touch the boundaries of AI. The aero and mechanics of insects is still mind-bendingly ahead of our best planes (power, maneuver, agility, range).

MIT’s latest insect-scale robot fuses new soft “muscle” actuators, a redesigned four-wing airframe, and an AI-trained controller to achieve truly bug-like flight. Multi-layer dielectric elastomer actuators provide high power at much lower voltage, while long, hair-thin wing hinges and improved transmissions cut mechanical stress so the robot can hover for about 1,000 seconds and execute flips and sharp turns faster than previous microrobots. On top of that hardware, the team uses model-predictive control as an expert “teacher” and distills it into a lightweight neural policy, enabling bumblebee-class speed and agility in a package that weighs less than a paperclip and opens a realistic path to autonomous swarms for pollination and search-and-rescue.

Tiny flying robots have long promised to help with tasks that are too dangerous or delicate for larger drones. Building such machines is extraordinarily difficult: the physics of flapping‑wing flight at insect scale and the constraints of small motors and batteries limit endurance and manoeuvrability. Over the past four years, engineers at the Massachusetts Institute of Technology (MIT) have made three major breakthroughs that together push insect‑scale flight from lab curiosity toward practical autonomy.

I built an animation using three.js to show just how cool this flight path can be.

The insect’s flight path is modeled using a discrete-time Langevin equation, which simulates Brownian motion with aerodynamic damping. At each time step \( t \), the velocity vector \( \mathbf{v} \) is updated by applying a stochastic acceleration \( \mathbf{a}_{\text{rand}} \) and a damping factor \( \gamma \) (representing air resistance):
$$ v_{t+1} = \gamma \, v_t + \mathbf{a}_{\text{rand}} $$
where \( \gamma = 0.95 \) and the components of \( \mathbf{a}_{\text{rand}} \) are drawn from a uniform distribution \( \mathcal{U}(-0.05, 0.05) \). The position \( \mathbf{p} \) is then integrated numerically:
$$ \mathbf{p}_{t+1} = \mathbf{p}_t + \mathbf{v}_{t+1} $$
This results in a “random walk” trajectory that is smoothed by the inertial momentum of the simulated robot, mimicking the erratic yet continuous flight dynamics of a small insect.

Use mouse to rotate/zoom.

Flying robots smaller than a few grams operate in a very different aerodynamic regime from traditional drones. Their wings experience low Reynolds numbers where lift comes from unsteady vortex shedding rather than smooth airflow; getting useful thrust requires wings to flap hundreds of times per second with large stroke angles. At such high frequencies, actuators often buckle and joints fatigue. The robots have extremely tight power and mass budgets, making it hard to carry batteries or onboard processors. Early insect‑like microrobots could just barely hover for a few seconds and needed bulky external power supplies.

A microrobot flips 10 times in 11 seconds.
Credit: Courtesy of the Soft and Micro Robotics Laboratory

MIT’s Soft and Micro Robotics Laboratory, led by Kevin Chen, set out to tackle these problems with a combination of novel soft actuators, mechanically resilient airframes and advanced control algorithms. The resulting platform evolved in three stages: improved artificial muscles (2021), a four‑wing cross‑shaped airframe with durable hinges and transmissions (January 2025) and a learning‑based controller that matches insect agility (December 2025).

The first breakthrough addressed the “muscle” problem. Conventional rigid actuators were too heavy and inefficient for gram‑scale robots. Chen’s group developed multilayer dielectric elastomer actuators—soft “muscles” made from ultrathin elastomer films sandwiched between carbon‑nanotube electrodes and rolled into cylinders. In 2021 they unveiled a fabrication method that eliminates microscopic air bubbles in the elastomer by vacuuming each layer after spin‑coating and baking it immediately. This allowed them to stack 20 alternating layers, each about 10 µm thick , without defects.

Key results from this work included:

  • Lower voltage and more payload: The new actuators operate at 75% lower voltage and carry 80% more payload than earlier soft actuators. By increasing surface area with more layers, they require less than 500V to actuate yet can lift nearly three times their own weight.
  • Higher power density and durability: Removing defects increases power output by more than 300% and extends lifespan. The 20‑layer actuators survived more than 2 million cycles while still flying smoothly.
  • Record hovering: Robots powered by these actuators achieved a 20‑second hovering flight, the longest yet recorded for a sub‑gram robot. With a lift‑to‑weight ratio of about 3.7:1, they could carry additional electronics.

The low‑voltage soft muscles solved a critical bottleneck: the robots could now carry lightweight power electronics and eventually microprocessors instead of being tethered to an external power supply. This breakthrough laid the hardware foundation for future autonomy.

Aerodynamic and structural redesign: four wings, durable hinges and long endurance

The second breakthrough came in early 2025 with a redesign of the robot’s wings and transmission. Previous generations assembled four two‑wing modules into a rectangle, creating eight wings whose wakes interfered with each other and limiting control authority. The new design arranges four single‑wing units in a cross. Each wing flaps outward, reducing aerodynamic interference and freeing up central volume for batteries and sensors.

To exploit the power of the improved actuators, Chen’s team built more complex transmissions that connect each actuator to its wing. These transmissions prevent the artificial muscles from buckling at high flapping frequencies, reducing mechanical strain and allowing higher torque. They also developed a 2 cm‑long wing hinge with a diameter of just 200 µm, fabricated via a multistep laser‑cutting process. The long hinge reduces torsional stress during flapping and increases durability; even slight misalignment during fabrication could affect the wing’s motion.

Performance improvements from this structural redesign were dramatic:

  • Extended flight time: The robot can hover for more than 1,000 seconds (~17 minutes) without degradation of precision —100 times longer than earlier insect‑scale robots.
  • Increased speed and agility: It reaches an average speed of 35 cm/s and performs body rolls and double flips. It can precisely follow complex paths, including spelling “MIT” in mid‑air.
  • Greater control torque: The new transmissions generate about three times more torque than prior designs , enabling sophisticated and accurate path‑following flights.

These mechanical innovations show how aerodynamic design, materials engineering and manufacturing precision are intertwined. By reducing strain on the wings and actuators, the platform gains not only endurance but also the stability needed for advanced control.

Autonomy, AI and Systems Engineering

The final breakthrough is in control theory as it intersects autonomy. Insect‑scale flight demands rapid decision‑making: the wings beat hundreds of times per second, and disturbances like gusts can quickly destabilize the robot. Traditional hand‑tuned controllers cannot handle aggressive maneuvers or unexpected perturbations. MIT’s team collaborated with Jonathan How’s Aerospace Controls Laboratory to develop a two‑step AI‑based control scheme.

Step 1: Model‑predictive control (MPC). The researchers built a high‑fidelity dynamic model of the robot’s mechanics and aerodynamics. An MPC uses this model to plan an optimal sequence of control inputs that follow a desired trajectory while respecting force and torque limits. The planner can design difficult maneuvers such as repeated flips and aggressive turns but is too computationally intensive to run on the robot in real time.

Step 2: Imitation‑learned policy. To bridge the gap between high‑end planning and onboard execution, the team generated training data by having the MPC perform many trajectories and perturbations. They then trained a deep neural network policy via imitation learning to map the robot’s state directly to control commands. This policy effectively compresses the MPC’s intelligence into a lightweight model that can run fast enough for real‑time control.

The results show how AI enables insect‑like agility:

  • Speed and acceleration: The learning‑based controller allows the robot to fly 447% faster and achieve a 255% increase in acceleration compared with their previous best hand‑tuned controller.
  • Complex maneuvers: The robot executed 10 somersaults in 11 seconds, staying within about 4–5 cm of the planned trajectory and maintaining performance despite wind gusts of more than 1 m/s
  • Bio‑inspired saccades: The policy enabled saccadic flight behavior, where the robot rapidly pitches forward to accelerate and then back to decelerate, mimicking insect eye stabilization strategies.

An external motion‑capture system currently provides state estimation, and the controller runs offboard. However, because the neural policy is much less computationally demanding than full MPC, the authors argue that similar policies may be feasible on tiny onboard processors. The AI control thus paves the way for autonomous flight without external assistance.

A key message of MIT’s microrobot program is that agility at insect scale does not come from a single innovation. The soft actuators enable large strokes at lower voltage; the cross‑shaped airframe and long hinges reduce mechanical strain and aerodynamic interference; and the AI‑driven controller exploits the robot’s physical capabilities while respecting constraints. Chen’s group emphasizes that hardware advances pushed them to develop better controllers, and improved controllers made it worthwhile to refine the hardware.

This co‑design philosophy—optimizing materials, mechanisms and algorithms together—will be essential as researchers push toward untethered, autonomous swarms. The team plans to integrate miniature batteries and sensors into the central cavity freed by the four‑wing design, allowing the robots to navigate without a motion‑capture system. Future work includes landing and take‑off from flowers for mechanical pollination and coordinated flight to avoid collisions and operate in groups. There are also open questions about enabling onboard perception—using tiny cameras or event sensors to close the loop—and about energy management to extend endurance beyond 10,000s.

MIT’s bumblebee‑sized flapping robot illustrates how progress in materials science, precision fabrication, aerodynamics and AI can converge to solve a hard problem. The low‑voltage, power‑dense actuators prove that soft materials can outperform rigid designs, the four‑wing airframe with durable hinges unlocks long endurance and high torque, and the hybrid MPC/learning controller shows that sophisticated planning can be compressed into hardware‑friendly neural policies. Together, these advances give the microrobot insect‑like speed, agility and endurance.

While still reliant on external power and motion capture, the robot’s modular design and AI controller suggest a roadmap to fully autonomous operation. As the team integrates onboard sensors and batteries, insect‑scale robots could move from labs to fields and disaster zones, pollinating crops or searching collapsed buildings. In that future, the intersection of autonomy and aerodynamics will be defined not by a single breakthrough but by a careful co‑design of muscles, wings and brains.

By 0 Comments
Autonomy

AI Hackers

AI Systems Are Now Hunting Software Vulnerabilities—And Winning.

Three months ago, Google Security SVP Heather Adkins and cryptographer Bruce Schneier warned that artificial intelligence would unleash an “AI vulnerability cataclysm.” They said autonomous code-analysis systems would find and weaponize flaws so quickly that human defenders would be overwhelmed. The claim seemed hyperbolic. Yet October and November of 2025 have proven them prescient: AI systems are now dominating bug bounty leaderboards, generating zero-day exploits in minutes, and even rewriting malware in real time to evade detection.

On an October morning, a commercial security agent called XBOW—a fully autonomous penetration tester—shot to the top of HackerOne’s U.S. leaderboard, outcompeting thousands of human hackers. Over the previous 90 days it had filed roughly 1,060 vulnerability reports, including remote-code-execution, SQL injection and server-side request forgery bugs. More than 50 were deemed critical. “If you’re competing for bug bounties, you’re not just competing against other humans anymore,” one veteran researcher told me. “You’re competing against machines that work 24/7, don’t get tired and are getting better every week.”

XBOW is just a harbinger. Between August and November 2025, at least nine developments upended how software is secured. OpenAI, Google DeepMind and DARPA all released sophisticated agents that can scan vast codebases, find vulnerabilities and propose or even automatically apply patches. State-backed hackers have begun using large-language models to design malware that modifies itself in mid-execution, while researchers at SpecterOps published a blueprint for an AI-mediated “gated loader” that decides whether a payload should run based on a covert risk assessment. And an AI-enabled exploit generator published in August showed that published CVEs can be weaponized in 10 to 15 minutes—collapsing the time defenders once had to patch systems.

The pattern is clear: AI systems aren’t just assisting security professionals; they’re replacing them in many tasks, creating new capabilities that didn’t exist before and forcing a fundamental rethinking of how software security works. As bug-bounty programs embrace automation and threat actors deploy AI at every stage of the kill chain, the offense-defense balance is being recalibrated at machine speed.

A New Offensive Arsenal

The offensive side of cybersecurity has seen the most dramatic AI advances, but it’s important to understand the distinction between what’s happening: foundational model companies like OpenAI and Anthropic are building the brains, while agent platforms like XBOW are building the bodies. This distinction matters when evaluating the different approaches emerging in AI-powered security.

OpenAI’s Aardvark, released on Oct. 30, is described by the company as a “security researcher” in software form. Rather than using static analysis or fuzzers, Aardvark uses large-language-model reasoning to build an internal representation of each codebase it analyzes. It continuously monitors commits, traces call graphs and identifies risky patterns. When it finds a potential vulnerability, Aardvark creates a test harness, executes the code in a sandbox and uses Codex to propose a patch. In internal benchmarks across open-source repositories, Aardvark reportedly detected 92 percent of known and synthetic vulnerabilities and discovered 10 new CVE-class flaws. It has already been offered pro bono to select open-source projects. But beneath the impressive numbers, Aardvark functions more like AI-enhanced static application security testing (SAST) than a true autonomous researcher—powerful, but incremental.

Google DeepMind’s CodeMender, unveiled on Oct. 6, takes the concept further by combining discovery and automated repair. It applies advanced program analysis, fuzzing and formal methods to find bugs and uses multi-agent LLMs to generate and validate patches. Over the past six months, CodeMender upstreamed 72 security fixes to open-source projects, some as large as 4.5 million lines of code. In one notable case, it inserted -fbounds-safety annotations into the WebP image library, proactively eliminating a buffer overflow that had been exploited in a zero-click iOS attack. All patches are still reviewed by human experts, but the cadence is accelerating.

Anthropic, meanwhile, is taking a fundamentally different path—one that involves building specialized training environments for red teaming. The company has devoted an entire team to training a foundational red team model. This approach represents a bet that the future of security AI lies not in bolting agents onto existing models, but in training models from the ground up to think like attackers.

The DARPA AI Cyber Challenge (AIxCC), concluded in October, showcased how far autonomous systems have come. Competing teams’ tools scanned 54 million lines of code, discovered 77 percent of synthetic vulnerabilities and generated working patches for 61 percent—with an average time to patch of 45 minutes. During the final four-hour round, participants found 54 new vulnerabilities and patched 43, plus 18 real bugs with 11 patches. DARPA announced that the winning systems will be open-sourced, democratizing these capabilities.

A flurry of attack-centric innovations soon followed. In August, researchers demonstrated an AI pipeline that can weaponize newly disclosed CVEs in under 15 minutes using automated patch diffing and exploit generation; the system costs about $1 per exploit and can scale to hundreds of vulnerabilities per day. A September opinion piece by Gadi Evron, Heather Adkins and Bruce Schneier noted that over the summer, autonomous AI hacking graduated from proof of concept to operational capability. XBOW vaulted to the top of HackerOne, DARPA’s challenge teams found dozens of new bugs, and Ukraine’s CERT uncovered malware using LLMs for reconnaissance and data theft, while another threat actor was caught using Anthropic’s Claude to automate cyberattacks. “AI agents now rival elite hackers,” the authors wrote, warning that the tools drastically reduce the cost and skill needed to exploit systems and could tip the balance towards the attackers.

Yet XBOW’s success reveals an important nuance about agent-based security tools. Unlike OpenAI’s Aardvark or Anthropic’s foundational approach, XBOW is an agent platform that uses these foundational models as backends. The vulnerabilities it finds tend to be surface-level—relatively easy targets like SQL injection, XSS and SSRF—not the deep architectural flaws that require sophisticated reasoning. XBOW’s real innovation wasn’t its vulnerability discovery capability; it was using LLMs to automatically write professional vulnerability reports and leveraging HackerOne’s leaderboard as a go-to-market strategy. By showing up on public rankings, XBOW demonstrated that AI could compete with human hackers at scale, even if the underlying vulnerabilities weren’t particularly complex.

Defense Gets More Automated—But Threats Evolve Faster

Even as defenders deploy AI, adversaries are innovating. The Google Threat Intelligence Group (GTIG) AI Threat Tracker, published on Nov. 5, is the most comprehensive look to date at AI in the wild. For the first time, GTIG identified “just-in-time AI” malware that calls large-language models at runtime to dynamically rewrite and obfuscate itself. One family, PROMPTFLUX, is a VBScript dropper that interacts with Gemini to generate new code segments on demand, making each infection unique. PROMPTSTEAL is a Python data miner that uses Qwen2.5-Coder to build Windows commands for data theft, while PROMPTLOCK demonstrates how ransomware can employ an LLM to craft cross-platform Lua scripts. Another tool, QUIETVAULT, uses an AI prompt to search JavaScript for authentication tokens and secrets. All of these examples show that attackers are moving beyond the 2024 paradigm of AI as a planning aide; in 2025, malware is beginning to self-modify mid-execution.

GTIG’s report also highlights the misuse of AI by state-sponsored actors. Chinese hackers posed as capture-the-flag participants to bypass guardrails and obtain exploitation guidance; Iranian group MUDDYCOAST masqueraded as university students to build custom malware and command-and-control servers, inadvertently exposing their infrastructure. These actors used Gemini to generate reconnaissance scripts, ransomware routines and exfiltration code, demonstrating that widely available models are enabling less-sophisticated hackers to perform advanced operations.

Meanwhile, SpecterOps researcher John Wotton introduced the concept of an AI-gated loader, a covert program that collects host telemetry—process lists, network activity, user presence—and sends it to an LLM, which decides whether the environment is a honeypot or a real victim. Only if the model approves does the loader decrypt and execute its payload; otherwise it quietly exits. The design, dubbed HALO, uses a fail-closed mechanism to avoid exposing a payload in a monitored environment. As LLM API costs fall, such evasive techniques become more practical.

Consolidation and Friction

These technological leaps are reshaping the business of cybersecurity. On Nov. 4, Bugcrowd announced that it will acquire Mayhem Security from my good friend David Brumley. His team was previously known as ForAllSecure and won the 2016 DARPA Cyber Grand Challenge. Mayhem’s technology automatically discovers and exploits bugs and uses reinforcement learning to prioritize high-impact vulnerabilities; it also builds dynamic software bills of materials and “chaos maps” of live systems. Bugcrowd plans to integrate Mayhem’s AI automation with its human hacker community, offering continuous penetration testing and merging AI with crowd-sourced expertise. “We’ve built a system that thinks like an attacker,” Mayhem founder David Brumley said, adding that combining with Bugcrowd brings AI to a global hacker network. The acquisition signals that bug bounty platforms will not remain purely human endeavours; automation is becoming a product feature.

The Mayhem acquisition also underscores the diverging strategies in the AI security space. While agent platforms like XBOW focus on automation at scale, foundational model teams are making massive capital investments in training infrastructure. Anthropic’s multi-billion-dollar commitment to building specialized red teaming environments dwarfs the iterative approach seen elsewhere. This created substantial competitive pressure: when word spread that Anthropic was spending at this scale, it generated significant fear of missing out among both startups and established players, accelerating consolidation moves like the Bugcrowd-Mayhem deal.

Yet adoption is uneven. Some folks I spoke with are testing Aardvark and CodeMender for internal red-teaming and patch generation but won’t deploy them in production without extensive governance. They worry about false positives, destabilizing critical systems and questions of liability if an AI-generated patch breaks something. The friction isn’t technological; it’s organizational—legal, compliance and risk management must all sign off.

The contrast between OpenAI’s and Anthropic’s approaches is striking. OpenAI’s Aardvark, while impressive in benchmarks, functions primarily as enhanced SAST—using AI to improve traditional static analysis rather than fundamentally rethinking how security research is done. Anthropic, by contrast, is betting that true autonomous security research requires training foundational models specifically for offensive security, complete with vast training environments that simulate real-world attack scenarios. This isn’t just a difference in tactics; it’s a philosophical divide about whether security AI should augment existing tools or replace them entirely.

Attackers, by contrast, face no such constraints. They can run self-modifying malware and LLM-powered exploit generators without worrying about compliance. GTIG’s report notes that the underground marketplace for illicit AI tooling is maturing, and the existence of PROMPTFLUX and PROMPTSTEAL suggests some criminal groups are already paying to call LLM APIs in operational malware. This asymmetry raises an unsettling question: Will AI adoption accelerate faster on the offensive side?

What Comes Next

Experts outline three scenarios. The Slow Burn assumes high friction on both sides leads to gradual, manageable adoption, giving regulators and organizations time to adapt. An Asymmetric Surge envisions attackers hurdling friction faster than defenders, driving a spike in breaches and forcing a reactive policy response. And the Cascade scenario posits simultaneous large-scale deployment by both offense and defense, producing the “vulnerability cataclysm” Adkins and Schneier warned about—just delayed by organizational inertia.

What we know: the technology exists. Autonomous agents can find and patch vulnerabilities faster than most humans and can generate exploits in minutes. Malware is starting to adapt itself mid-execution. Bug bounty platforms are integrating AI at their core. And nation-state actors are experimenting with open-source models to augment operations. The question isn’t whether AI will transform cybersecurity—that’s already happening—but whether defenders or attackers will adopt the technology faster, and whether policy makers will help shape the outcome.

Time is short. Patch windows have shrunk from weeks to minutes. Signature-based detection is increasingly unreliable against self-modifying malware. And AI systems like XBOW, Aardvark and CodeMender are running 24 hours a day on infrastructure that scales infinitely.

By 0 Comments
Personal Projects & Interests

To Act or To Trust: A false Choice?

Credit: WSJ

There’s a pastor in Moscow, Idaho who wants you to fight. Doug Wilson and the Christian Nationalists tell us the culture is at stake—we must engage, strategize, win. Anything less is retreat, surrender, unfaithfulness. This resonates with me. I’m a military man, trained and willing to fight. I’m also a conservative, convicted to preserve the good in society. This requires action and courage. My friends who attend his church are wise, committed and productive citizens.

Doug Wilson’s Icon

However, a recent visit with a friend re-introduced me to Rowan Williams, who offers not “theology that bites back” but the mystic’s voice: “Consider the lilies. They neither toil nor spin.” Trust God. Let go. This sounds like wisdom—until it sounds like fatalism. Or worse, Islam’s inshallah: whatever will be, will be.

Which is it? Fight for culture or let God do the fighting?

The question feels urgent, binary. Kierkegaard would recognize it immediately—the tyranny of “Either/Or” thinking. Pick a side. You’re either with us or against us.

Rowan Williams—Welsh poet, patristics scholar, former Archbishop of Canterbury—spent a decade frustrating everyone by refusing to choose.

During his 2002-2012 tenure, activists wanted decisive stands. Conservatives wanted tradition defended. Progressives wanted him leading the charge. Everyone wanted a general willing to marshal troops.

Why Ya’ll Fighting?

Instead, they got a contemplative who wrote poetry, translated fourth-century mystics, and gave maddeningly nuanced answers. When pressed on sexuality debates threatening to split global Anglicanism, he chose patient dialogue over ideological purity. The right called him weak. The left called him complicit.

But Williams wasn’t confused. He was inhabiting the tension rather than resolving it. He learned this from Christianity’s most unlikely revolutionaries—the Desert Fathers.

In his book Silence and Honey Cakes, Williams introduces these strange rebels. When fourth-century Christianity became the empire’s official religion and bishops wore imperial robes, thousands of Christians fled to the Egyptian desert.

These weren’t quietists. They saw that “winning” the culture war meant losing everything that mattered. The empire embraced the church; the church became imperial. To bishops negotiating with emperors, accumulating wealth, defining orthodoxy through political power, the Desert Fathers said: No.

They didn’t storm halls of power or write manifestos. They walked into the wilderness with nothing and sat down.

And from that wilderness, they changed everything.

Here’s what Williams sees: their withdrawal was the most political act imaginable. They weren’t escaping the culture war—they were fighting it on completely different terms.

The empire said power comes from influence, wealth, control. The Desert Fathers built nothing, owned nothing, controlled nothing.

The culture said identity comes from role, status, accomplishments. The Desert Fathers literally forgot their names in silence.

The economy said value equals productivity. The Desert Fathers sat absolutely still for hours, producing nothing, calling it the world’s most important work.

This drove visitors crazy. Pilgrims traveled weeks seeking wisdom, and the hermit would say: “Go, sit in your cell, and your cell will teach you everything.”

It sounds like non-engagement. But Williams argues the opposite: the desert wasn’t retreat from reality—it was the only place to see reality clearly.

The activist impulse runs deep in American Christianity. Save the culture. Win the argument. Build the institution. Doug Wilson’s approach resonates because it feels faithful, muscular. “Faith without works is dead,” right?

But Williams asks: What if all your frantic activity is baptized anxiety? What if your need to “fight for the culture” is actually ego—your need to matter, to win? What if activism is driven not by love but fear—fear of irrelevance, of God not showing up unless you make it happen?

The Desert Fathers called this logismoi—the swirling thoughts driving us, compulsions masquerading as virtues. Until you sit still enough to see these illusions clearly, every “good work” is contaminated by them.

In the desert, they learned apatheia—not apathy, but freedom from reactive passions. Acting from clarity rather than compulsion, love rather than fear, trust rather than control.

This drove Williams’ critics mad. They wanted reaction. He offered contemplation. They wanted tribal certainty. He offered nuanced discernment. They wanted a general. He gave them a monk.

But doesn’t this collapse into Islamic fatalism? If we’re supposed to trust God, why act at all?

Here’s Williams’ crucial synthesis:

The fatalist says: “My actions don’t matter, so why try?” Despair disguised as piety.

The activist says: “Everything depends on me, so I must never rest.” Pride disguised as faithfulness.

The contemplative says: “God is acting, and I get to participate—but the outcome isn’t mine to control.” This is freedom.

Look at Jesus. Forty days in the desert before public ministry. Regular withdrawal sustaining public work. He engaged culture—healing, teaching, confronting power. But he acted from deep rootedness in the Father, from abundance not anxiety.

The night before his arrest, Peter draws a sword. Jesus says: put it away. Not because he’s passive, but because he’s operating from completely different understanding of power. That’s not fatalism. That’s the most radical trust in human history.

Kierkegaard’s Either/Or presents seemingly incompatible life choices: aesthetic or ethical, pleasure or duty. But Kierkegaard knew the title itself is the trap. The mature person doesn’t choose between them; they integrate them at a higher level in the “religious” stage.

Williams does the same with contemplation and action. We must trust and act. Withdraw and engage. Cultivate silence and speak prophetically.

Think of it like breathing. Inhale: contemplation, withdrawal, receiving. Exhale: action, engagement, giving. You can’t live by only inhaling—you’ll burst. You can’t live by only exhaling—you’ll collapse.

The question isn’t calculating the perfect ratio. It’s: Are you rooted deeply enough to engage wisely?

But here’s where Williams’ approach faces its most devastating critique: What about when evil demands immediate action?

Consider the lilies while babies are being aborted? Cultivate silence while DEI policies systematically discriminate? Would Williams have been Churchill or Chamberlain?

Edmund Burke haunts us: “The only thing necessary for the triumph of evil is for good men to do nothing.” Hitler didn’t need more Rowan Williams types—he needed the church to fight.

When Bonhoeffer returned to Germany in 1939, he chose conspiracy and resistance over safe contemplation. When Wilberforce fought the slave trade for decades, he wasn’t told to “consider the lilies.”

So is Williams wrong?

No—but he’s incomplete if we don’t understand what contemplation actually produces.

The Desert Fathers weren’t pacifists—they were the fiercest warriors against evil. But they understood the deepest evils are spiritual, and you can’t fight spiritual battles with carnal weapons. Athanasius, who fought Arianism for decades, learned his theology from Anthony of the desert. The contemplative produced the fighter.

Those silent monks became the theological backbone resisting imperial heresy. When emperors demanded doctrinal compromise, the desert tradition held firm. Their withdrawal gave them clarity and courage to resist.

Contemplation isn’t an alternative to fighting evil—it’s what equips you to fight the right battles in the right way for the right reasons.

The activist who never contemplates will fight—but against what? With what weapons? Driven by what spirit? We’ve seen conservatives become indistinguishable from the left in tactics, spirit, tribalism. They’re fighting like the world fights, which means they’ve already lost.

DEI is evil when it judges people by immutable characteristics rather than character and competence. Fight it. But how? By becoming equally tribal, equally discriminatory—just reversed? That’s not winning; that’s becoming what you oppose.

The pro-life movement’s greatest victories haven’t come from politicians screaming on cable news—they’ve come from crisis pregnancy centers, changed hearts, decades of patient work.

So practically, what does this mean?

First, check your motivations. Is this flowing from prayer or anxiety? From love or ego? Anxious activism burns out or becomes monstrous.

Second, cultivate depth. If your activism isn’t rooted in regular silence, solitude, prayer—it will corrupt you. Bonhoeffer practiced the disciplines even while plotting against Hitler.

Third, embrace the long game. The Desert Fathers planted seeds that wouldn’t flower for generations. Wilberforce fought twenty years before victory.

Fourth, go smaller, deeper, truer. Don’t grasp for cultural dominance. Build communities of alternative practice. The early church didn’t defeat Rome by winning elections—they built a parallel society so compelling the empire converted.

Fifth, know your real enemies. The Desert Fathers’ battles were internal—pride, vainglory, anger. The same spiritual evil producing DEI—pride disguised as compassion—lives in your heart too.

After a decade trying to hold together a fracturing communion, Williams stepped down and returned to teaching, writing, poetry. Some saw defeat. But maybe it was the most desert move of all: releasing the need to control outcomes, trusting God with the church.

Since leaving Canterbury, Williams has written more explicitly about economic justice, climate change, empire. Freed from institutional power, he’s become more pointed. The withdrawal enabled deeper engagement.

The question Williams forces us to face isn’t whether to engage or trust, fight or pray. It’s: From what source are you living?

Are you acting from union with God, or from anxious need to make something happen? Is your engagement an overflow of contemplation, or a substitute for it?

The Desert Fathers fled not because they didn’t care about the world, but because they loved it too much to let empire define what caring looks like.

Williams spent a decade demonstrating you can be deeply engaged while refusing to let the world’s anxiety determine your response. You can lead without grasping. Care without controlling. Act decisively while holding outcomes loosely.

The culture won’t be saved by Christians who trust God so much they do nothing. Nor by Christians who fight so hard they become indistinguishable from every other power-seeking tribe.

It might be saved by communities who’ve been to the desert. Who’ve sat still enough to see their illusions clearly. Who act not from anxiety but abundance. Who fight not from fear but love. Who engage not to win but to witness.

Who trust God enough to act. And act in ways that demonstrate trust.

The Desert Fathers didn’t defeat empire by organizing coalitions. They defeated it by refusing to let it define reality. By living so differently that people traveled hundreds of miles to ask: “How do you have such peace?”

And when pilgrims arrived expecting great wisdom, the hermits would say: “Go sit in your cell, and your cell will teach you everything.”

Maybe that’s exactly what we need to hear.

Not because culture doesn’t need engagement, but because we can’t engage faithfully until we’ve learned to sit still. Until we’ve discovered that God’s kingdom comes not by might nor power, but by a Spirit we can only encounter in silence.

Then—only then—when we speak, we’ll have something worth saying. When we act, we’ll act from depth. When we fight, we’ll fight for the right things in the right ways.

We’ll fight like people who’ve learned to trust. And trust like people free to act.

Which is to say: we’ll breathe. In and out. Contemplation and action. Desert and city.

Both. And. Not either. Or.

Maybe the way forward isn’t choosing between the culture warrior and the contemplative. It’s becoming the kind of person who’s been silent enough to have something worth saying when they speak.

By 0 Comments
Personal Projects & Interests

Reformed Christian Perspective on Israel and Modern Judaism

Some American Christians are rethinking what they believe about Israel and Judaism. Shifting geopolitical realities, rising tensions both domestically and internationally, and increasingly polarized discourse have forced questions that have been long settled for me. But before we can answer “What should Christians think about Jews?”—we need to untangle what we’re actually talking about.

We can call something “Jewish”, when it functions as at least five different categories that we habitually collapse into one:

Ethnicity: Genetic descent from Abraham through Isaac and Jacob
Culture: Rich literary and intellectual traditions, distinctive languages (Yiddish, Ladino, Judeo-Arabic), cuisine and foodways, music (klezmer, liturgical, contemporary), art and cinema, philosophical contributions, scientific achievements, centuries of diaspora experience, humor and storytelling traditions, family and community practices, distinct historical memory and commemoration
Faith: Religious belief and practice ranging from secular to ultra-Orthodox
Geopolitics: The modern State of Israel and Middle Eastern policy
Theology: God’s covenant relationship with His chosen people

And within the category of faith alone, “Judaism” spans an enormous spectrum. Reform Jews may not believe in a personal God. Conservative Jews navigate tradition and modernity. Modern Orthodox professionals may study Talmud daily and some ultra-Orthodox communities reject Zionism and don’t recognize the State of Israel. Many friends are secular Jews who identify culturally but not religiously. Even among the religiously observant, there are vast differences in how they relate to foundational texts—Torah (the five books of Moses), Nevi’im (the Prophets), Ketuvim (the Writings), the Talmud (rabbinic discussions and legal interpretations), Midrash (interpretive commentaries), Kabbalah (mystical traditions), and countless generations of rabbinical responsa. Some communities prioritize halakhic (legal) study, others emphasize ethical teachings, still others focus on mystical experience or philosophical theology.

When we ask “What should the Christian posture be toward Jews?” which Judaism are we discussing? The Hasidic community in Brooklyn? The Reform temple down the street? The Israeli soldier? The Hollywood producer? The Talmud scholar? These aren’t interchangeable categories, and our theology must be precise enough to account for these distinctions—not just in what we believe, but in how we relate, engage, and bear witness.

The confusion isn’t just academic. When we lump together ethnicity, faith, culture, and geopolitics, we end up with theological frameworks that can’t distinguish between critique of Israeli policy, rejection of Talmudic tradition, and hatred of Jewish people. We need to disentangle these threads before we can think biblically about any of them.

In the recent words of Stephen Wise, “Jews get to define Judaism, others get to decide if they accept us as we see ourselves”. That’s fair. My answer then of Who and What is provided by the manifold conversations I’ve had with the people who practice their Jewish faith. For me that means two groups: secular Jews largely in military, tech and/or business and faithful religious Jews I know largely through the classroom or conservative political affiliations.

For the purpose of this post, I’m going to assume we are talking about the theological core and shared beliefs of the Jewish faith today as understood through my personal relationships and a bit of reading.

The Biblical Framework or What Christian Scripture Actually Says

Paul’s treatment of Israel in Romans 11 provides the definitive framework for Reformed thinking about Judaism. He describes a remnant chosen by grace (11:5)—not wholesale abandonment but divine preservation. He speaks of partial hardening, not total rejection (11:25)—Israel’s blindness is temporary and purposeful, not permanent apostasy. Most remarkably, Paul uses present tense—they are “beloved for the sake of their forefathers” (11:28), not were beloved. God’s love for Israel continues in the present, not just the past. The gifts and calling of God are irrevocable (11:29)—God doesn’t break covenants, even when His people stumble.

This language is incompatible with viewing Judaism as a pagan or demonic religion. Paul explicitly warns Gentile believers against arrogance toward the natural branches (11:18-22). If post-Temple Judaism were intrinsically demonic, Paul’s entire argument collapses.

Revelation’s Vision: Israel in God’s Future

Revelation’s vision of the end times offers a striking confirmation of Israel’s enduring place in God’s purposes. When the Apostle John sees the future unfold, he doesn’t witness Israel’s erasure or replacement—he sees their distinct preservation. The 144,000 sealed from the twelve tribes of Israel appear in Revelation 7 and 14, not as a metaphor emptied of ethnic meaning, but as testimony to God’s faithfulness to His covenant people. The New Jerusalem itself bears witness to this continuity and its twelve gates are named for the twelve tribes, inscribed into the architecture of God’s eternal city (Revelation 21:12). Even the apocalyptic measuring of the temple and those who worship there (Revelation 11:1-2) maintains Israel’s specific identity in God’s final purposes. Whether we read these passages literally or symbolically, the theological point remains unshakeable, namely that God has not abandoned His covenant relationship with Israel as a people.

This biblical vision helps us understand what “demonic” actually means in Scripture’s categories. The word isn’t a catch-all for “things Christians disagree with”—it has specific, defined boundaries. Scripture reserves “demonic” language for idol worship and service to false gods (Deuteronomy 32:17; 1 Corinthians 10:20), conscious opposition to Christ as the Antichrist spirit (1 John 2:22), and occult practices and divination explicitly condemned in the Law (Leviticus 19:31; Deuteronomy 18:10-12). These are clear categories with clear markers.

Rabbinic Judaism, even in its tragic rejection of Jesus as Messiah, doesn’t fit these categories. Jews continue to worship the Creator God of Abraham, Isaac, and Jacob—the same God Christians worship, though they cannot see His full revelation in Christ. They maintain the Hebrew Scriptures as authoritative and binding, the very Scriptures that testify to Jesus. They order their lives around prayer, repentance, and holiness, seeking to live before the God who gave Torah at Sinai. They faithfully preserve the Sabbath, the biblical festivals, and the covenant markers that God Himself instituted. This isn’t apostasy to demons—it’s spiritual blindness to the fulfillment that has come.

The center of gravity remains the God of Israel, not Molech or Baal. This is blindness to fulfillment, not apostasy to paganism.

The Talmud Question: Corruption vs. Apostasy

One argument made recently (by folks in my hometown of Fort Worth) is that a careful reading of the Talmud reveals a wholly corrupted Jewish faith that ancient Jews wouldn’t recognize. But this claim requires significant qualification. First, not all Jews hold the Talmud (the oral tradition) on par with the written tradition. The Jewish world is far more diverse in its relationship to rabbinic texts than many critics acknowledge.

The Talmud does present legitimate concerns for Christians. It can obscure grace under layers of legal reasoning, making the encounter with God’s mercy harder to see. Some passages reflect hostility toward Christianity, shaped by centuries of conflict and persecution. It solidifies a system that, without Christ, cannot ultimately save. Certain mystical traditions drift toward problematic territory that should give us pause. These are real issues that deserve honest theological engagement.

But we must be careful not to mistake corruption for apostasy, or development for departure. The Talmud is fundamentally an in-house Jewish attempt to order life before the God of Israel according to Torah. It’s not a manual for worshiping a different deity. It’s rabbinic Judaism wrestling with the question: How do we remain faithful to the covenant when the Temple is destroyed and the priesthood scattered? This distinction matters enormously when we’re trying to assess whether we’re dealing with blindness to fulfillment or apostasy to demons.

Consider Jesus’ own approach to Pharisaic tradition, which became the foundation for what evolved into rabbinic Judaism. He blasts their additions and burdens in Matthew 23, calling out their hypocrisy and legalistic distortions with searing clarity. Yet in the same breath, He acknowledges “they sit in Moses’ seat,” recognizing their legitimate authority to interpret Torah. He debates within the framework of Jewish tradition, not as an outsider confronting a foreign religion. This is the posture of a reformer calling His people back to covenant faithfulness, not a missionary encountering paganism.

What continues from ancient Judaism to its modern expression tells us something crucial. The same God remains central—the Shema, “Hear, O Israel: the LORD our God, the LORD is one,” is still prayed daily. The same Scriptures are read, studied, and revered. The same covenant markers—circumcision, Sabbath, dietary laws—structure Jewish life. The same fundamental hope persists, even if Messianic expectation is understood differently. Many of the same prayers are still recited, some dating back to Temple times. This isn’t wholesale replacement; it’s recognizable continuity.

What changed after the destruction of the Temple represents adaptation within tradition rather than abandonment of it. Temple sacrifice was replaced by prayer and study, following the prophetic principle that God desires mercy more than sacrifice. Priestly authority gave way to rabbinic interpretation as the community needed new leadership structures. Messianic hope was deferred rather than recognized in Jesus—a tragic blindness, but still hope directed toward the God of Abraham. The oral Torah was codified in Talmud and Midrash to preserve what had been transmitted verbally for generations. Theological emphasis shifted from sacrifice to ethics and law as the community sought ways to maintain covenant relationship without the Temple system.

This is development within the same tradition, not the creation of a new religion. A first-century Pharisee, transported to a modern Orthodox synagogue, would recognize far more than he’d find foreign. He would hear familiar prayers, see familiar rituals, recognize the Torah scroll and its reverent treatment. He might be shocked by some theological developments, puzzled by certain innovations, but he wouldn’t think he’d stumbled into a temple of Baal or Molech. The center of gravity—the God of Israel, the authority of Scripture, the covenant relationship—remains recognizably continuous.

Contemporary Jewish belief spans an enormous spectrum, and any analysis that treats it as monolithic fails before it begins.

Judaism is diverse and committed faith

Within contemporary Judaism, the diversity of belief and practice is staggering. Ultra-Orthodox communities structure entire lives around intensive Talmud study and strict halakhic observance, often living in insular neighborhoods where religious law governs every detail from sunrise to sunset. Modern Orthodox Jews navigate a different balance, engaging Torah deeply while participating fully in modern professional and cultural life—doctors and lawyers who spend their evenings studying ancient texts. Conservative Judaism attempts to honor historical tradition while applying critical scholarship to its sources, creating communities that look traditional but think historically. Reform Judaism emphasizes ethical monotheism over ritual observance, seeing Judaism primarily as a moral framework rather than a legal system. And millions of secular Jews maintain strong cultural identity—celebrating Passover, mourning the Holocaust, supporting Israel—while holding no particular religious beliefs at all.

The claim that “all rabbinic Jews reject God dwelling with man” reveals a fundamental unfamiliarity with Jewish sources and practice. The concept of the Shekhinah—God’s divine presence dwelling among His people—remains absolutely central to traditional Jewish thought across denominations. Every synagogue service explicitly invokes God’s presence. Hasidic traditions speak constantly of encountering the divine in everyday life. Jewish mysticism, from medieval Kabbalah to modern Hasidism, emphasizes divine immanence with an intensity that would surprise critics. Contemporary Jewish philosophers like Abraham Joshua Heschel and Martin Buber have written profoundly about divine-human encounter. To claim Judaism categorically denies God’s presence is simply false.

When Christians carelessly label Judaism “demonic,” a cascade of consequences follows that extends far beyond theological error. We forget that our own Scriptures emerged from Jewish scribes, that our Savior lived as a Torah-observant Jew, that our first apostles were all Jewish believers who saw Jesus as Messiah, not as founder of a new religion. We undermine Paul’s carefully constructed argument in Romans about Israel’s ongoing significance in God’s purposes. We create space for actual antisemites to claim Christian validation for their hatred. We destroy any credible witness to Jewish communities—who would listen to someone who’s already declared them demonic? And most dangerously, we align ourselves with ideological movements that have historically led to persecution and atrocity.

Luther’s Warning: When Evangelical Frustration Becomes Genocidal Blueprint

The Reformed tradition must reckon honestly with how this trajectory has played out in our own history. Martin Luther provides the most sobering example—and for those of us who love Luther, who have been shaped by his theological courage, his biblical insight, his unwavering commitment to justification by faith alone, this reckoning is painful. I count myself among Luther’s admirers. His stand at Worms, his translation of Scripture, his hymns, his exposition of Galatians—these have nourished my faith and the faith of millions. Which makes his writings on the Jews not just historically troubling but personally grievous. We cannot love Luther rightly without lamenting this aspect of his legacy deeply.

In his earlier years, Luther criticized the Catholic Church’s treatment of Jews and held hope for mass Jewish conversion once the Gospel was freed from papal corruption. His 1523 work “That Jesus Christ Was Born a Jew” showed genuine concern for Jewish evangelism and criticized Christian mistreatment.

On the Jews and Their Lies (1543)

But when Jews failed to convert in the numbers Luther anticipated, his frustration curdled into something far darker. By 1543, Luther published “On the Jews and Their Lies,” a document so venomous that the Nazis would later display it at Nuremberg rallies. Luther called for burning synagogues, destroying Jewish homes, confiscating prayer books and Talmudic writings, forbidding rabbis from teaching, abolishing safe conduct for Jews on highways, banning usury, and forcing Jews into manual labor. His theological justification? That the Jews’ rejection of Christ proved them to be children of the devil, their synagogues “a den of devils,” their worship demonic.

The progression is instructive and terrifying. Luther began with orthodox Christian conviction—faith in Christ is necessary for salvation. He added urgent evangelistic hope—surely Jews will recognize their Messiah when the Gospel is clearly preached. When reality disappointed—Jews remained unconvinced—theological frustration transmuted into demonization. And demonization inevitably produced calls for persecution. If Jews are demonic, if their worship is satanic, if their very presence pollutes Christian society, then violence becomes not just permissible but pious.

Four centuries later, Nazi propagandists didn’t have to invent Christian antisemitism—they simply dusted off Luther and gave his recommendations modern implementation. When Kristallnacht came on this day (9 Nov) in 1938, synagogues burned across Germany on November 9th—Martin Luther’s birthday. The switch can happen: righteous evangelical urgency can become dark ethnic hatred; theological conviction can become demonization.

The lesson isn’t that we should soft-pedal the Gospel or pretend Jewish rejection of Christ doesn’t matter. Luther was right that Jesus is the only way to salvation. He was right that post-Temple Judaism cannot save. He was catastrophically, damnably wrong in moving from “Jews need Christ” to “Jews are demonic” to “Jews should be persecuted.” The slide from the first to the second happens when we lose Paul’s nuance in Romans 11. The slide from the second to the third is inevitable—it’s simply a matter of time and political opportunity.

A faithful Reformed perspective maintains what might seem like contradictory truths, holding them in tension as Scripture does. Salvation is in Christ alone—any religious system that rejects Jesus cannot ultimately save. Judaism without Christ remains incomplete and broken, with the veil over Moses still unlifted, as Paul describes in 2 Corinthians. Yet simultaneously, Israel remains beloved for the sake of the patriarchs, their gifts and calling irrevocable. The root of the olive tree is holy, and we Gentile believers are grafted into Israel’s story, not the reverse. And crucially for our historical moment, antisemitism—the hatred of Jewish people—stands in direct opposition to God’s purposes and the Gospel itself. These truths don’t contradict; they complete each other.

This framework allows us to evangelize Jewish people with urgency and love, understanding that faith in Christ is essential for salvation while recognizing we’re speaking to those who already know the God of Abraham. It enables us to oppose antisemitism wherever it emerges—whether in progressive spaces that cloak hatred in anti-Zionism or in conservative circles that traffic in conspiracy theories. We can appreciate Judaism’s remarkable preservation of Scripture and its ongoing witness to monotheism, even while maintaining that this witness remains incomplete without Christ. Most importantly, we can hold theological clarity without resorting to demonization, recognizing the profound mystery of Israel’s future restoration that Paul describes in Romans 11:25-26.

These theological commitments have concrete implications for how Christians engage the world today. In theological discussion, we must reject the simplistic “Judaism is demonic” rhetoric that has gained traction in some corners of the internet and most alarmingly in the Church. The distinction between spiritual blindness and pagan apostasy isn’t semantic hairsplitting—it’s the difference between biblical fidelity and dangerous error. Before making sweeping pronouncements about Judaism, Christians should immerse themselves in Romans 9-11, where Paul wrestles with these very questions with far more nuance than most contemporary commentators manage.

In political engagement, the stakes are even higher. Christians must refuse to platform antisemites, regardless of how much we might agree with their positions on other issues. Supporting Israel’s right to exist doesn’t require blind endorsement of every policy decision, just as loving Jewish neighbors doesn’t mean abandoning theological convictions. But we must be clear—theological disagreement never justifies persecution, marginalization, or hatred. When antisemitism appears in progressive spaces or conservative ones, Christians must call it out with equal vigor.

The most transformative engagement, though, happens in personal relationships. Building genuine friendships with Jewish neighbors, learning about Judaism from practicing Jews rather than plucking random passages from the Talmud to construct caricatures, sharing the Gospel with love rather than contempt—these ordinary interactions matter more than grand theological pronouncements. Christians can celebrate what Judaism has faithfully preserved through centuries of persecution while pointing to its fulfillment in Christ. We can study together, disagree deeply, and still recognize our shared heritage in the God of Abraham, Isaac, and Jacob.

Church teaching bears special responsibility in this moment. Pastors must preach the whole counsel of Scripture on Israel, including the difficult passages that don’t fit neatly into either supersessionist or dispensationalist categories. Replacement theology that erases Israel’s ongoing significance contradicts Paul’s explicit teaching, but so does any framework that ignores Judaism’s tragic blindness to its own Messiah. Churches must teach Christian history honestly, including the shameful legacy of Christian antisemitism that provided theological cover for persecution and genocide. Only by facing this history squarely can we prepare our congregations to engage Jewish friends, neighbors, and colleagues with both theological clarity and genuine love.

To declare Judaism “demonic” is to saw off the branch we’re sitting on. Christianity emerges from Judaism, fulfills Judaism’s hopes, and shares Judaism’s Scriptures. Our Savior was a Torah-observant Jew who prayed the Shema, kept the Sabbath, and celebrated Passover. The apostles were Jews who saw Jesus as Israel’s Messiah, not a foreign deity. Yes, modern Judaism’s rejection of Jesus is spiritually fatal. Yes, the Talmudic tradition includes problematic elements. Yes, we must evangelize Jewish people with urgency. But we must do so recognizing what Paul knew—this is family business. We’re dealing with elder brothers who can’t see the family resemblance in the One they reject, not strangers worshiping foreign gods. They are “enemies for your sake” but “beloved for the sake of the fathers” (Romans 11:28).

The Reformed tradition at its best provides the clarity our moment demands. Unwavering commitment to salvation in Christ alone, coupled with deep respect for God’s irrevocable calling of Israel. This isn’t theological compromise—it’s biblical fidelity. The way forward isn’t through demonization but through faithful witness, proclaiming Christ as the fulfillment of Israel’s hope while standing firmly against those who would harm the people through whom salvation came to the world. This is the Reformed position, the biblical position, and the only position that takes seriously both the Gospel’s exclusivity and God’s covenant faithfulness. As Paul concludes his meditation on Israel’s mystery—”Oh, the depth of the riches both of the wisdom and knowledge of God! How unsearchable are His judgments and His ways past finding out!” (Romans 11:33). That humility—not internet boldness or reactionary provocations—should mark our engagement with the mystery of Israel and the Jewish people.

By 0 Comments
Personal Projects & Interests

The Data Doesn’t Lie: How 266 Runs Revealed The Truth About My Marathon Goal

For the BMW Marathon in Dallas, I’m targeting 7:30/mile pace—a 3:16:45 finish, what training plan should I follow from today? Work has been insane stressful, and has destroyed my training. Can AI help me here?

My long runs felt good at that pace for 8-10 miles, so I thought I was on track. I’m always testing to see how much I can do myself and I’m curious to use AI to coach me. AI isn’t useful without the right data, so I analyzed 266 runs from the past two years to build a good training plan.

The most recent data is the most important, so it was important that I took note of my run this morning. I ran a half marathon at 8:10 average pace with a heart rate of 137-148 bpm—very easy aerobic effort. I finished the last two miles at 7:42 and 7:49 pace.

Here’s what 266 runs over two years revealed:

DISTANCE RANGERUNSAVERAGE PACEBEST PACEPATTERN
3-4 miles928:146:44Speed is there!
6-7 miles318:257:23Solid training pace
10-11 miles88:117:47Sub-8:00 capability proven
13-14 miles37:487:53THE SWEET SPOT
14-15 miles27:547:41Strong mid-distance
16-17 miles28:268:20Starting to fade
18-19 miles28:117:44Inconsistent
20+ miles58:548:00THE PROBLEM

The pattern was clear: 13-14 mile average of 7:48 versus 20+ mile average of 8:32. A 44-second-per-mile dropoff. My best 20+ mile run: 8:00 pace. Still 30 seconds slower than goal. My average 20+ mile run: 8:32 pace. A 1:02/mile gap to close. But this morning’s half marathon told a different story. I ran 13.1 miles at low heart rate (137-148 bpm) and finished strong at 7:42-7:49 pace. “Very easy,” I noted afterward. “Could have done a lot more.” This suggests my aerobic base is much better than the historical 8:32 suggests. That average likely came from poor pacing on long runs, not lack of fitness.

My 3-4 mile best pace: 6:26. That’s 1:04 faster than my marathon goal. The problem isn’t speed—it’s extending that speed over distance. The gap: extending my 7:48 pace from 13-14 miles to 20+ miles, then racing smart for 26.2 miles. When you define it that specifically, you can build a plan to address it.

The dropoff from 7:48 pace (13-14 miles) to 8:32 pace (20+ miles) isn’t random—it’s physiological. Research on elite marathoners shows that even well-trained runners deplete muscle glycogen stores significantly after 90-120 minutes at marathon pace. For me, 13-14 miles at 7:48 takes about 1:47—right in the window where glycogen runs low. When I push to 20 miles without proper fueling and without training my body to use fat efficiently, I hit the metabolic wall. But fuel isn’t the only issue. My legs simply aren’t conditioned for the distance.

By “legs aren’t ready,” I mean muscular endurance breaks down, neuromuscular efficiency degrades, and form deteriorates. The quads that fire smoothly at mile 10 are misfiring by mile 18. The signal from brain to muscle becomes less crisp. Motor unit recruitment—the coordinated firing of muscle fibers—gets sloppy. I’m sending the same “run at 7:45 pace” command, but my legs execute it as 8:30 pace. Meanwhile, small biomechanical breakdowns compound: hip drops slightly, stride shortens, each foot strike becomes less efficient. Running 20 miles means roughly 30,000 foot strikes. If I haven’t progressively trained my legs to absorb that cumulative pounding, my body literally slows me down to prevent injury.

Studies on elite marathon training show successful marathoners spend 74% of their training volume at easy intensity (Zone 1-2) because it builds aerobic capacity without accumulating neuromuscular fatigue. My data suggests I was probably running too many miles too hard, accumulating fatigue faster than I could recover—especially at 48. Research on masters athletes shows recovery takes 10-20% longer after age 40. Some coaches recommend 10-12 day training cycles for older athletes instead of traditional 7-day cycles, allowing more space between hard efforts. If I’m not recovering fully before the next quality session, my 20+ mile pace suffers even more than it would for a younger runner.

There’s also cardiovascular drift to consider. During prolonged running, cardiac output gradually decreases while heart rate increases to compensate. This is more pronounced at higher intensities. If I’m running long runs at or near race pace (7:30-7:50), I’m experiencing significant cardiovascular drift by mile 15-18. The effort to maintain pace increases exponentially. My 20+ mile pace of 8:32 might simply reflect the point where my cardiovascular system says “enough.”

My training strategy alternates run days with bike days—typically 18-28 miles at 15-18 mph. Research shows that cycling can maintain aerobic fitness while reducing impact stress, with one mile of running equaling approximately three miles of cycling for cardiovascular equivalence. This means my weekly aerobic load is higher than running mileage suggests: 45 miles running plus 85 miles cycling equals roughly 73 “running equivalent” miles per week. The cycling protects my legs while maintaining cardiovascular fitness—smart training at 48. But it also means my running-specific adaptations (neuromuscular patterns, impact tolerance, glycogen depletion management) might be underdeveloped relative to my aerobic capacity.

The data reveals a specific problem: I have speed (6:26 for 3-4 miles), good mid-distance endurance (7:48 for 13-14 miles), and strong aerobic fitness (cycling adds volume). But I haven’t trained the specific adaptation of holding sub-8:00 pace beyond 14 miles. This isn’t a fitness problem—it’s a specificity problem. The solution isn’t to run more miles. It’s to progressively extend the distance at which I can hold my proven 7:48 pace, while managing fatigue and recovery as a 48-year-old athlete.

Traditional marathon training plans prescribe long runs at “easy pace” or 30-60 seconds slower than race pace. That builds aerobic base, but it doesn’t address my specific limitation: I need to teach my body to hold quality pace for progressively longer distances.

This morning’s half marathon changes the starting point. Instead of beginning conservatively at 16 miles, I can start more aggressively at 18 miles next Saturday. The plan builds from there: 18 miles at 7:50 average pace (two miles easy warmup, 14 miles at 7:45-7:50, two miles easy cooldown). The goal is simple—extend this morning’s easy 13.1-mile effort to 14 quality miles. Week two pushes to 20 miles at 7:45 average, tackling my historical problem distance at a pace 47 seconds per mile faster than my 8:32 average. Week three peaks at 22 miles averaging 7:40 pace—the breakthrough workout that proves I can hold sub-7:40 pace for 18 continuous miles.

After peaking, the volume drops but the intensity holds. Week four: 16 miles at 7:35 average with 12 miles at race pace (7:30-7:35). Week five adjusts for Thanksgiving with 14 miles at 7:35. Week six is the dress rehearsal: 10 miles with six at 7:25-7:30 pace, confirming goal pace is ready. The progression is deliberate—each week either extends distance or drops pace by five seconds per mile, allowing physiological adaptation without overwhelming the system. Elite marathon training research supports this approach: progressive overload with strategic recovery.

Tuesday speedwork leverages my natural speed. My best 3-4 mile pace is 6:26—more than a minute per mile faster than marathon goal pace. Research consistently shows that running intervals 30-60 seconds faster than marathon pace improves race performance by increasing VO2 max, improving running economy, and creating a “speed reserve” that makes race pace feel controlled.

The plan starts with 8×800 meters at 6:30 pace (3:15 per repeat) with 90-second recovery jogs—establishing that I can run fast repeatedly. Week two builds to 10×800 at the same pace. Week three shifts to marathon-specific longer intervals: 6×1200 meters at 6:40 pace. Week four is a six-mile tempo run at 7:10-7:15 pace—faster than race pace, sustained effort. The final speedwork comes in week five: 6×800 at 6:25 pace for sharpness, not volume. Running 6:30 pace in workouts creates a one-minute-per-mile speed reserve over my 7:30 goal. Analysis of 92 sub-elite marathon training plans found successful programs include 5-15% high-intensity training. My Tuesday sessions provide exactly this stimulus.

At 48, recovery determines whether I arrive at race day peaked or exhausted. Complete rest days come every five to six days—no running, no cross-training, just rest. Easy runs stay at 6-7 miles at 8:20-8:30 pace, conversational effort that builds aerobic capacity without adding fatigue. Bike days alternate with run days: recovery rides of 18-22 miles at 15 mph with high cadence and low resistance, or moderate rides of 22-28 miles at 16 mph for steady aerobic work. The cycling maintains cardiovascular fitness and increases blood flow to running muscles while reducing impact stress. Research on masters runners consistently emphasizes that recovery adaptations—not just training adaptations—determine race day performance for athletes over 40.

Fueling practice matters as much as the miles themselves. My data shows pace dropping significantly after 90-120 minutes, suggesting glycogen depletion. Case studies on elite marathoners found optimal race-day fueling to be 60 grams of carbohydrate per hour, delivered as 15 grams every 15 minutes in a 10% carbohydrate solution. I need to practice this in training, not just on race day. Every long run over 90 minutes becomes a fueling rehearsal at goal pace.

This morning’s half marathon rewrites the race plan. I ran 13.1 miles at low heart rate (137-148 bpm), averaging 8:10 for the first 11 miles before closing the last two at 7:42 and 7:49 pace. My note afterward: “very easy, could have done a lot more.” That performance suggests my aerobic base is significantly better than my historical 8:32 average for 20+ miles indicates. That average likely came from poor pacing on long runs, not lack of fitness.

The conservative approach mirrors this morning’s pattern: controlled start, gradual build, strong finish. Miles 1-10 at 7:30-7:35 pace—just like this morning’s easy start—predicted split of 1:15:00 to 1:16:40. It will feel easy, almost too easy. The temptation will be to push harder. Don’t. Miles 11-20 settle into goal pace range at 7:28-7:32, predicted split of 1:15:00 to 1:16:00. This is half-marathon distance, and my data proves I can hold 7:48 pace for 13-14 miles. Running 7:30-7:35 is only 13-23 seconds per mile faster—controlled and sustainable. Miles 21-26.2 finish strong at 7:25-7:30 pace, just like this morning’s close. Predicted split: 38:45 to 39:00. Total projected finish: 3:15:00 to 3:16:45.

The aggressive approach—even 7:30 splits from mile one for a 3:16:45 finish—only makes sense if week three’s 22-mile peak run at 7:40 average feels as easy as this morning’s half marathon felt, if weather is perfect (50-55°F, low humidity, no wind), if taper goes flawlessly, and if I wake up on race day feeling exceptional. Otherwise, the conservative negative split strategy is smarter.

Three scenarios, all representing massive improvement over historical data. Best case: 3:15:00-3:16:00, requiring perfect execution of the conservative strategy and this morning’s negative split pattern. Probability: 45 percent, up significantly after this morning’s performance. Realistic: 3:16:00-3:17:00, solid execution with maybe imperfect conditions. Probability: 40 percent. Solid: 3:17:00-3:18:30, good race with slight fade or challenging conditions. Probability: 15 percent.

All three outcomes crush the historical 8:32 pace for 20+ miles. All three are victories. The goal isn’t to cling to 7:30 pace at all costs—it’s to run the smartest race possible given training data and this morning’s proof that the aerobic base is there.

I thought I was running long runs at 7:30 pace. The data showed 7:48 for 13-14 miles and 8:32 for 20+ miles. Memory is selective. Data isn’t.

But this morning’s half marathon revealed something the historical data missed: I ran 13.1 miles at a heart rate of 137-148 bpm—easy aerobic effort—and finished strong at 7:42 and 7:49 pace. Afterward, I noted “very easy, could have done a lot more.” That 8:32 average for 20+ miles wasn’t about fitness—it was about pacing. I’d been going out too hard and fading. The aerobic base is better than the numbers suggested.

The limitation isn’t speed—my best 3-4 mile pace is 6:26. It’s not aerobic fitness—the cycling adds significant volume and this morning proved the engine is strong. The gap is specificity: I haven’t trained to hold quality pace beyond 14 miles. At 48, I need more recovery than I did at 35. Research shows masters athletes need 10-20% more recovery time. The alternating run-bike schedule isn’t a compromise—it’s smart training that keeps me healthy enough to execute the progressive long runs that will close the gap.

Seven weeks to race day. Progressive long runs build from 18 to 22 miles at progressively faster paces. Tuesday speedwork at 6:30 pace creates a one-minute-per-mile reserve over goal pace. Complete rest every five to six days. Race strategy mirrors this morning’s pattern: controlled start, build into goal pace, finish strong.

Will I hit 3:16:45? Good chance—this morning proved the base is there. Will I run 3:16:00-3:17:00? More likely. Either way, it’s significantly faster than 8:32. The data showed the problem. This morning showed the solution. Now execute.

RESEARCH REFERENCES

  1. Stellingwerff, T. (2012). “Case Study: Nutrition and Training Periodization in
    Three Elite Marathon Runners.” International Journal of Sport Nutrition and
    Exercise Metabolism, 22(5), 392-400.
  2. Sports Medicine – Open (2024). “Quantitative Analysis of 92 12-Week Sub-elite
    Marathon Training Plans.”
  3. Tanaka, H. (1994). “Effects of cross-training. Transfer of training effects on
    VO2max between cycling, running and swimming.” Sports Medicine, 18(5), 330-339.
  4. Runner’s World / Marathon Training Academy (2024-2025). “Marathon Training After 50”
  • Research on masters athlete adaptations and recovery needs.
  1. Haugen, T., et al. (2019). “The Training and Development of Elite Sprint Performance: an Integration of Scientific and Best Practice Literature.” Sports Medicine – Open.
By One Comment
Personal Projects & Interests

Fixing the inlet on my Ryobi Pressure Washer

Some posts I write to inform others of cool stuff. Others posts I write to save someone else time when they google. This post is just a story of frustration, written to look like the second example. You have been warned.

Several years ago, I bought this pressure washer and then let it sit unused for far too long—a bad move. The Ryobi RY80940B, powered by a Honda GCV190 engine, delivers 3,100 PSI at approximately 2.5 GPM, more than enough force for serious cleaning but also enough to punish neglected hardware. For anyone working on one of these, the operating manual is available here: Ryobi RY80940B Service Manual (PDF).

When I finally tried to remove the hose, the metal fitting was stuck. The smart thing would have been to leave the hose stuck in there forever. I didn’t do that. I wanted to get the old hose out and, well, I’m an arrogant engineer and this was a battle of wills not of wits. I’ve spent the better part of an afternoon fighting with a broken brass fitting fused inside the outlet coupler of my Ryobi RY80940B pressure washer. What should have been a simple hose replacement turned into a problem. The brass insert from the old high-pressure hose sheared off cleanly, leaving a ring of corroded metal wedged inside the plastic spindle. I tried everything — prying, tapping, a bolt extractor — but the brass just spun or smeared instead of breaking free.

Eventually, I realized I was wasting time trying to rescue a part that costs less than lunch. The black collar that covers the quick-connect doesn’t unscrew; it slides over a spring and bearing set. Once you see the parts diagram for this model, it’s obvious the spindle (part #642058001) and collar (part #310993003) are meant to be replaced together.

Part 64 is wicked

Per the service manual, I first tried an 8 mm hex key to back out the fitting, but it immediately began to strip, producing nothing but a rain of metal shavings.

Prying out the Brass was a huge pain

After exhausting every polite method of extraction, I escalated. Since I knew the brass and aluminum had fused completely, I cut most of the seized coupler off with an angle grinder, then used diagonal cutters to peel the remaining sleeve away from the spindle. With the bulk removed, I applied heat from a torch, clamped a pair of vice grips onto what was left, and added a four-foot cheater bar for leverage. The grips slipped off more than once, but eventually I felt the faint give of motion—the threads finally broke free. Sometimes the “fix” isn’t surgical precision; it’s controlled brutality, executed carefully enough not to destroy the surrounding parts.

Because this took so much work. I’m going to show you two pictures.

Tech Details

So what happened here? The failure wasn’t merely mechanical — it was electrochemical and thermomechanical. The brass hose insert (nominally Cu–Zn alloy, ≈ \(Cu_{60}Zn_{40}\)) had been pressed into a steel spindle (Fe–C alloy with minor Mn, Si). Over years of intermittent water exposure, this dissimilar-metal couple set up a galvanic cell with the water acting as electrolyte. Brass is more noble than steel in the galvanic series:

$$E^\circ_{\text{Cu}^{2+}/\text{Cu}} = +0.337\text{ V}, \quad E^\circ_{\text{Fe}^{2+}/\text{Fe}} = -0.440\text{ V} $$

The potential difference,

$$\Delta E = E_{\text{cathode}} – E_{\text{anode}} \approx 0.777\text{ V},$$

was sufficient to drive anodic dissolution of iron and the formation of iron oxides/hydroxides.

The local electrochemical reactions were:

Anode (steel):

$$\text{Fe} \rightarrow \text{Fe}^{2+} + 2e^-$$

Cathode (brass surface):

$$\text{O}_2 + 2H_2O + 4e^- \rightarrow 4OH^-$$

The resulting \(Fe(OH)_2\) and \(Fe_2O_3\) corrosion products expanded volumetrically, wedging the interface tighter. The growth strain of oxide formation can be expressed as:

$$\varepsilon_v = \frac{V_{\text{oxide}} – V_{\text{metal}}}{V_{\text{metal}}}$$

For iron oxides, \(\varepsilon_v \approx 1.9\), meaning nearly a doubling in volume. This expansion induced radial compressive stress at the interface:

$$\sigma_r = \frac{E \, \varepsilon_v}{1 – \nu}$$

With \(E_{\text{Fe}} \approx 210\,\text{GPa}\) and \(\nu \approx 0.3\), localized stresses easily exceeded hundreds of MPa, enough to plastically deform the surrounding brass and produce microscopic cold-welding through diffusion and oxide bridging.

At the same time, dezincification occurred in the brass:

$$\text{CuZn} \rightarrow \text{Cu} + \text{Zn}^{2+}$$

leaving a porous copper-rich layer with high surface energy. Under compression and mild heat cycling, solid-state diffusion across the Cu–Fe interface promoted the formation of intermetallic compounds such as \(CuFe_2\) or \(Fe_3Zn_{10}\), effectively brazing the two metals together.

The end state was a compound joint bonded both by corrosion-product expansion and diffusion adhesion — a quasi-metallurgical weld. The torque required to shear it loose later had to exceed the yield strength of both alloys plus the adhesion energy of the oxide layer:

$$\tau_{\text{break}} \ge \frac{\sigma_y A + \gamma_{\text{adh}}}{r}$$

where \(\sigma_y\) is the yield stress, A the contact area, \(\gamma_{\text{adh}}\) the oxide interfacial energy, and r the effective radius. I estimate I needed breakaway torque of in the neighborhood of 80–160 N·m, danger close to the ~200-250 N·m that will break a Grade 8.8 M16 bolt — and this was a thin walled threaded tube. The geometry also doesn’t help. Since brass (CuZn) was threaded into the aluminum/steel pump body at M16 scale, the fine 1.5 mm pitch meant very tight thread flank contact. Over years of water exposure, galvanic corrosion products grew in the tiny 60° V thread roots, effectively cold-welding the joint.

So the “fused” fitting wasn’t rusted in the casual sense — it had become a galvanic, mechanically locked, diffusion-bonded composite system, obeying the same physics that make bimetallic corrosion joints and accidental cold-welds in space so hard to undo.

By One Comment
fly brain
Autonomy

The Connectome Scaling Wall: What Mapping Fly Brains Reveals About Autonomous Systems

In October 2024, the FlyWire consortium published the complete connectome of an adult Drosophila melanogaster brain: 139,255 neurons, 54.5 million synapses, 8,453 cell types, mapped at 8-nanometer resolution. This is only the third complete animal connectome ever produced, following C. elegans (302 neurons, 1986) and Drosophila larva (3,016 neurons, 2023). The 38-year gap between the first and third isn’t just a story about technology—it’s about fundamental scaling constraints that apply equally to biological neural networks and autonomous robotics.

A connectome is the complete wiring diagram of a nervous system. Think of it as the ultimate circuit board schematic, except instead of copper traces and resistors, you’re mapping biological neurons and the synapses that connect them. But unlike an electrical circuit you can trace with your finger, these connections exist in three-dimensional space at scales invisible to the naked eye. A single neuron might be a tenth of a millimeter long but only a few microns wide—about 1/20th the width of a human hair. The synapses where neurons connect are even tinier: just 20-40 nanometers across, roughly 2,000 times smaller than the width of a hair. And there are millions of them, tangled together in a three-dimensional mesh that makes the densest urban skyline look spacious by comparison. The connectome doesn’t just tell you which neurons exist—it tells you how they talk to each other. Neuron A connects to neurons B, C, and D. Neuron B connects back to A and forward to E. That recurrent loop between A and B? That might be how the fly remembers the smell of rotting fruit. The connection strength between neurons—whether a synapse is strong or weak—determines whether a signal gets amplified or filtered out. It’s the difference between a whisper and a shout, between a fleeting thought and a committed decision.

Creating a connectome is almost absurdly difficult. First, you preserve the brain perfectly at nanometer resolution, then slice it into thousands of impossibly thin sections—the fruit fly required 7,000 slices, each just 40 nanometers thick, about 1/1,000th the thickness of paper, cut with diamond-knife precision. Each slice goes under an electron microscope, generating a 100-teravoxel dataset where each pixel represents an 8-nanometer cube of brain tissue. Then comes the nightmare part: tracing each neuron through all 7,000 slices, like following a single thread through 7,000 sheets of paper where the thread appears as just a dot on each sheet—and there are 139,255 threads tangled together. When Sydney Brenner’s team mapped C. elegans in the 1970s and 80s, they did this entirely by hand, printing electron microscopy images on glossy paper and tracing neurons with colored pens through thousands of images. It took 15 years for 302 neurons. The fruit fly has 460 times more.

This is where the breakthrough happened. The FlyWire consortium used machine learning algorithms called flood-filling networks to automatically segment neurons, but the AI made mistakes constantly—merging neurons, splitting single cells into pieces, missing connections. So they crowdsourced the proofreading: hundreds of scientists and citizen scientists corrected errors neuron by neuron, synapse by synapse, and the AI learned from each correction. This hybrid approach—silicon intelligence for speed, human intelligence for accuracy—took approximately 50 person-years of work. The team then used additional machine learning to predict neurotransmitter types from images alone and identified 8,453 distinct cell types. The final result: 139,255 neurons, 54.5 million synapses, every single connection mapped—a complete wiring diagram of how a fly thinks.

The O(n²) Problem

Connectomes scale brutally. Going from 302 to 139,255 neurons isn’t 460× harder—it’s exponentially harder because connectivity scales roughly as n². The worm has ~7,000 synapses. The fly has 54.5 million—a 7,800× increase in edges for a 460× increase in nodes. The fly brain also went from 118 cell classes to 8,453 distinct types, meaning the segmentation problem became orders of magnitude more complex.

Sydney Brenner’s team in 1986 traced C. elegans by hand through 8,000 electron microscopy images using colored pens on glossy prints. They tried automating it with a Modular I computer (64 KB memory), but 1980s hardware couldn’t handle the reconstruction. The entire project took 15 years.

The FlyWire consortium solved the scaling problem with a three-stage pipeline:

Connectome Evolution

Mapping the Neural Universe

From 302 neurons in a tiny worm to 139,255 in a fruit fly—witness the exponential complexity that defines the frontier of neuroscience.

Exponential Growth
460× more neurons in 38 years
7,142× more synapses
Complexity: O(n²) connections

Stage 1: Automated segmentation via flood-filling networks
They sliced the fly brain into 7,000 sections at 40nm thickness (1/1000th the thickness of paper), generating a 100-teravoxel dataset. Flood-filling neural networks performed initial segmentation—essentially a 3D region-growing algorithm that propagates labels across voxels with similar appearance. This is computationally tractable because it’s local and parallelizable, but error-prone because neurons can look similar to surrounding tissue at boundaries.

Stage 2: Crowdsourced proofreading
The AI merged adjacent neurons, split single cells, and missed synapses constantly. Hundreds of neuroscientists and citizen scientists corrected these errors through a web interface. Each correction fed back into the training set, iteratively improving the model. This hybrid approach—automated first-pass, human verification, iterative refinement—consumed approximately 50 person-years.

Stage 3: Machine learning for neurotransmitter prediction
Rather than requiring chemical labeling for every neuron, they trained classifiers to predict neurotransmitter type (glutamate, GABA, acetylcholine, etc.) directly from EM morphology. This is non-trivial because neurotransmitter identity isn’t always apparent from structure alone, but statistical patterns in vesicle density, synapse morphology, and connectivity motifs provide signal.

The result is complete: every neuron traced from dendrite to axon terminal, every synapse identified and typed, every connection mapped. UC Berkeley researchers ran the entire connectome as a leaky integrate-and-fire network on a laptop, successfully predicting sensorimotor pathways for sugar sensing, water detection, and proboscis extension. Completeness matters because partial connectomes can’t capture whole-brain dynamics—recurrent loops, feedback pathways, and emergent computation require the full graph.

50 largest neurons of the fly brain connectome.
Credit: Tyler Sloan and Amy Sterling for FlyWire, Princeton University, (Dorkenwald et al., 2024)

Power Scaling: The Fundamental Constraint

Here’s the engineering problem: the fly brain operates on 5-10 microwatts. That’s 0.036-0.072 nanowatts per neuron. It enables:

  • Controlled flight at 5 m/s with 200 Hz wing beat frequency
  • Visual processing through 77,536 optic lobe neurons at ~200 fps
  • Real-time sensorimotor integration with ~10ms latencies
  • Onboard learning and memory formation
  • Navigation, courtship, and decision-making

Compare this to autonomous systems:

SystemPowerCapabilityEfficiency
Fruit fly brain10 μWFull autonomy0.072 nW/neuron
Intel Loihi 22.26 μW/neuronInference only31× worse than fly
NVIDIA Jetson (edge inference)~15 WVision + control10⁶× worse
Boston Dynamics Spot~400 W total90 min runtime–
Human brain20 W1 exaFLOP equivalent50 pFLOPS/W
Frontier supercomputer20 MW1 exaFLOP50 FLOPS/W

The brain achieves ~10¹² better energy efficiency than conventional computing at the same computational throughput. This isn’t a transistor physics problem—it’s architectural.

Why Biology Wins: Event-Driven Sparse Computation

The connectome reveals three principles that explain biological efficiency:

1. Asynchronous event-driven processing
Neurons fire sparsely (~1-10 Hz average) and asynchronously. There’s no global clock. Computation happens only when a spike arrives. In the fly brain, within four synaptic hops nearly every neuron can influence every other (high recurrence), yet the network remains stable because most neurons are silent most of the time. Contrast this with synchronous processors where every transistor is clocked every cycle, consuming 30-40% of chip power just on clock distribution—even when idle.

2. Strong/weak connection asymmetry
The fly connectome shows that 70-79% of all synaptic weight is concentrated in just 16% of connections. These strong connections (>10 synapses between a neuron pair) carry reliable signals with high SNR. Weak connections (1-2 synapses) may represent developmental noise, context-dependent modulation, or exploratory wiring that rarely fires. This creates a core network of reliable pathways overlaid with a sparse exploratory graph—essentially a biological ensemble method that balances exploitation and exploration.

3. Recurrent loops instead of deep feedforward hierarchies
The Drosophila larval connectome (3,016 neurons, 548,000 synapses) revealed that 41% of neurons receive long-range recurrent input. Rather than the deep feedforward architectures in CNNs (which require many layers to build useful representations), insect brains use nested recurrent loops. This compensates for shallow depth: instead of composing features through 50+ layers, they iterate and refine through recurrent processing with 3-4 layers. Multisensory integration starts with sense-specific second-order neurons but rapidly converges to shared third/fourth-order processing—biological transfer learning that amortizes computation across modalities.

Neuromorphic Implementations: Narrowing the Gap

Event-driven neuromorphic chips implement these principles in silicon:

Intel Loihi 2 (2024)

  • 1M neurons/chip, 4nm process
  • Fully programmable neuron models
  • Graded spikes (not just binary)
  • 2.26 μW/neuron (vs. 0.072 nW for flies—still 31,000× worse)
  • 16× better energy efficiency than GPUs on audio tasks

The Hala Point system (1,152 Loihi 2 chips) achieves 1.15 billion neurons at 2,600W maximum—demonstrating that neuromorphic scales, but still consumes orders of magnitude more than biology per neuron.

IBM NorthPole (2023)

  • Eliminates von Neumann bottleneck by co-locating memory and compute
  • 22 billion transistors, 800 mm² die
  • 25× better energy efficiency vs. 12nm GPUs on vision tasks
  • 72× better efficiency on LLM token generation

NorthPole is significant because it addresses the memory wall: in traditional architectures, moving data between DRAM and compute costs 100-1000× more energy than the actual computation. Co-locating memory eliminates this overhead.

BrainChip Akida (2021-present)

  • First commercially available neuromorphic chip
  • 100 μW to 300 mW depending on workload
  • Akida Pico (Oct 2024): <1 mW operation
  • On-chip learning without backprop

The critical insight: event-driven cameras + neuromorphic processors. Traditional cameras output full frames at 30-60 fps whether or not anything changed. Event-based cameras (DVS sensors) output asynchronous pixel-level changes only—mimicking retinal spike encoding. Paired with neuromorphic processors, this achieves dramatic efficiency: idle power drops from ~30W (GPU) to <1mW (neuromorphic).

RoboBees
Credit: Wyss Institute at Harvard University

The Micro-Robotics Power Wall

The scaling problem becomes acute at insect scale. Harvard’s RoboBee (80-100 mg, 3 cm wingspan) flies at 120 mW. Solar cells deliver 0.76 mW/mg at full sun—but the RoboBee needs 3× Earth’s solar flux to sustain flight. This forces operation under halogen lights in the lab, not outdoor autonomy.

The fundamental constraint is energy storage scaling. As robots shrink, surface area (power collection) scales as r², volume (power demand) as r³. At sub-gram scale, lithium-ion provides 1.8 MJ/kg. Metabolic fat provides 38 MJ/kg—a 21× advantage that no battery chemistry on the roadmap approaches.

This creates a catch-22: larger batteries enable longer flight, but add weight requiring bigger actuators drawing more power, requiring even larger batteries. The loop doesn’t converge.

Alternative approaches:

  1. Cyborg insects: Beijing Institute of Technology’s 74 mg brain controller interfaces directly with bee neural tissue via three electrodes, achieving 90% command compliance. Power: hundreds of microwatts for control electronics. Propulsion: biological, running on metabolic fuel. Result: 1,000× power advantage over fully robotic micro-flyers.
  2. Chemical fuels: The RoBeetle (88 mg crawling robot) uses catalytic methanol combustion. Methanol: 20 MJ/kg—10× lithium-ion density. But scaling combustion to aerial vehicles introduces complexity (fuel pumps, thermal management) at millimeter scales.
  3. Tethered operation: MIT’s micro-robot achieved 1,000+ seconds of flight with double aerial flips, but remains tethered to external power.

For untethered autonomous micro-aerial vehicles, current battery chemistry makes hour-long operation physically impossible.

The Dragonfly Algorithm: Proof of Concept

The dragonfly demonstrates what’s possible with ~1 million neurons. It intercepts prey with 95%+ success, responding to maneuvers in 50ms:

  • 10 ms: photoreceptor response + signal transmission
  • 5 ms: muscle force production
  • 35 ms: neural computation

That’s 3-4 neuron layers maximum at biological signaling speeds (~1 m/s propagation, ~1 ms synaptic delays). The algorithm is parallel navigation: maintain constant line-of-sight angle to prey while adjusting speed. It’s simple, fast, and works with 1/100th the spatial resolution of human vision at 200 fps.

The dragonfly’s visual system contains 441 input neurons feeding 194,481 processing neurons. Researchers have implemented this in III-V nanowire optoelectronics operating at sub-picowatt per neuron. The human brain averages 0.23 nW/neuron—still 100,000× more efficient than conventional processors per operation. Loihi 2 and NorthPole narrow this to ~1,000× gap, but the remaining distance requires architectural innovation, not just process shrinks.

Distributed Control vs. Centralized Bottlenecks

Insect nervous systems demonstrate distributed control:

  • Spinal reflexes: 10-30 ms responses without brain involvement
  • Central pattern generators: rhythmic movements (walking, flying) produced locally in ganglia
  • Parallel sensory streams: no serialization bottleneck

Modern autonomous vehicles route all sensor data to central compute clusters, creating communication bottlenecks and single points of failure. The biological approach is hierarchical:

  • Low-level reactive control: fast (1-10 ms), continuous, local
  • High-level deliberative planning: slow (100-1000 ms), occasional, centralized

This matches the insect architecture: local ganglia handle reflexes and pattern generation, brain handles navigation and decision-making. The division of labor minimizes communication overhead and reduces latency.

Scaling Constraints Across Morphologies

The power constraint manifests predictably across scales:

Micro-aerial (0.1-1 g): Battery energy density is the hard limit. Flight times measured in minutes. Solutions require chemical fuels, cyborg approaches, or accepting tethered operation.

Drones (0.1-10 kg): Power ∝ v³ for flight. Consumer drones: 20-30 min. Advanced commercial: 40-55 min. Headwinds cut range by half. Monarch butterflies migrate thousands of km on 140 mg of fat—200× better mass-adjusted efficiency.

Ground robots (10-100 kg): Legged locomotion is 2-5× less efficient than wheeled for the same distance (constantly fighting gravity during stance phase). Spot: 90 min runtime, 3-5 km range. Humanoids: 5-10× worse cost of transport than humans despite electric motors being more efficient than muscle (90% vs. 25%). The difference is energy storage and integrated control.

Computation overhead: At large scale (vehicles drawing 10-30 kW), AI processing (500-1000W) is 5-10% overhead. At micro-scale, computation dominates: a 3cm autonomous robot with CNN vision gets 15 minutes from a 40 mAh battery because video processing drains faster than actuation.

The Engineering Path Forward

The connectome provides a blueprint, but implementation requires system-level integration:

1. Neuromorphic processors for always-on sensing
Event-driven computation with DVS cameras enables <1 mW idle power. Critical for battery-limited mobile robots.

2. Hierarchical control architectures
Distribute reflexes and pattern generation to local controllers. Reserve central compute for high-level planning. Reduces communication overhead and latency.

3. Task-specific optimization over general intelligence
The dragonfly’s parallel navigation algorithm is simple, fast, and sufficient for 95%+ interception success. General-purpose autonomy is expensive. Narrow, well-defined missions allow exploiting biological efficiency principles.

4. Structural batteries and variable impedance actuators
Structural batteries serve as both energy storage and load-bearing elements, improving payload mass fraction. Variable impedance actuators mimic muscle compliance, reducing energy waste during impacts.

5. Chemical fuels for micro-robotics
At sub-gram scale, metabolic fuel’s 20-40× energy density advantage over batteries is insurmountable with current chemistry. Methanol combustion, hydrogen fuel cells, or cyborg approaches are the only paths to hour-long micro-aerial autonomy.

Minuscule RoBeetle Turns Liquid Methanol Into Muscle Power
Credit: IEEE Spectrum

Conclusion: Efficiency is the Constraint

The fruit fly connectome took 38 years after C. elegans because complexity scales exponentially. The same scaling laws govern autonomous robotics: every doubling of capability demands exponentially more energy, computation, and integration complexity.

The fly brain’s 5-10 μW budget for full autonomy isn’t just impressive—it’s the benchmark. Current neuromorphic chips are 1,000-30,000× worse per neuron. Closing this gap requires implementing biological principles: event-driven sparse computation, strong/weak connection asymmetry, recurrent processing over deep hierarchies, and distributed control.

Companies building autonomous systems without addressing energetics will hit a wall—the same wall that kept connectomics at 302 neurons for 38 years. Physics doesn’t care about better perception models if the robot runs out of power before completing useful work.

When robotics achieves even 1% of biological efficiency at scale, truly autonomous micro-robots become feasible. Until then, the scaling laws remain unforgiving. The connectome reveals both how far we are from biological efficiency—and the specific architectural principles required to close that gap.

The message for robotics engineers: efficiency isn’t a feature, it’s the product.

Explore the full FlyWire visualization at https://flywire.ai/

By 3 Comments
Engineering & Code

The GENIUS Act: From Crypto Experiment to National Security Infrastructure

When President Trump signed the GENIUS Act on July 18, 2025, he didn’t just regulate stablecoins—he legitimized an entire technology stack that will become core infrastructure for defense and intelligence agencies. By requiring full liquid reserves and exempting compliant stablecoins from securities regulations, the Act transforms what was once speculative technology into government-grade payment and settlement infrastructure.

This matters because modern military operations aren’t starved for bandwidth and processing power—they need immutable state, enforceable trust, and precise data distribution. The same technology that powers a dollar-pegged stablecoin can lock mission events in seconds, encode rules of engagement in self-executing code, and prove compliance without compromising operational security.

Command and control breaks when everyone maintains their own version of “truth.” Byzantine fault-tolerant consensus—the same mechanism that prevents double-spending in stablecoins—forces independent nodes to agree on each state change before it’s recorded. No more reconciliation queues. No more debating which ROE update is authoritative. Just a single, cryptographically locked sequence of events that can’t be quietly rewritten.

With that single, tamper‑proof log in place, autonomous systems can graduate from lone‑wolf behaviors to coordinated swarms: the ledger’s state becomes the authoritative “leader,” while each drone, satellite, or ground robot is a “follower” that subscribes to—and writes into—the same stream. AI planners can post composite tasks (sensor → shooter chains, refuel rendezvous, med‑evac corridors) as smart‑contract directives; individual platforms claim subtasks, stake cryptographic commitments, and update status the moment they act. Because every node sees identical state within seconds, hand‑offs and deconfliction happen in code, not over voice nets, and higher‑level orchestration engines can reason over the evolving mission graph in real time. The result is multi‑platform C2 where machines negotiate roles, verify execution, and re‑task themselves under a commander’s intent—while commanders and auditors retain an immutable play‑by‑play of who led, who followed, and whether the AI kept faith with policy.

But this tech can do more than just enable coordinated higher level autonomy. Supply chains fail when we can’t prove provenance. The tokenization standards legitimized by the GENIUS Act give every part, payment, and piece of data a cryptographic birth certificate. A counterfeit component breaks the chain instantly. Payment releases automatically when secure oracles confirm delivery and compliance. Cash cycles drop from 60-120 days to near real-time.

Zero-trust architectures stall without programmable policy. Smart contracts—now legally deployable for government use—convert access control from organizational charts into mathematical proofs. A fire mission only executes if the required cryptographic signatures align. Coalition partners can verify actions without seeing raw intelligence.

Before the GENIUS Act, stable‑coin rails and tokenised provenance rarely moved beyond DARPA demos because the scale‑up money wasn’t there: mission engineers could prove value in a sandbox, but CFOs and contracting officers lacked a legally accepted funding mechanism and a clear path to classify the R&D spend. The existence of large‑volume pilots—and the capital that follows—will harden the stack, integrate it with legacy systems, and drive down technical risk for the war‑fighter.

Now, with explicit statutory protection for properly-backed digital assets, defense contractors can build at scale. The same legal framework that protects a stablecoin’s dollar peg protects a tokenized part inventory, a programmable budget line, or a cryptographically-governed sensor feed.

The Act doesn’t just permit this infrastructure—it demands it. Full reserve requirements and regular attestations? That’s exactly the level of auditability the DoD needs for financial management. Bankruptcy-remote custody structures? Perfect for protecting government assets in contractor systems. Real-time settlement with cryptographic finality? Precisely what accelerated kill chains require.

The GENIUS Act creates a rare alignment between financial innovation and national security needs. The question isn’t whether this technology will reshape defense operations—it’s who will build it.

This will intersect with formal methods in a powerful way. Recent breakthroughs in large‑scale formal verification—driven by Amazon’s Byron Cook and Michael Hicks—slot perfectly into this Merkle‑anchored world: their push‑button model‑checking and property‑based compilation pipelines can exhaustively prove that a smart contract (or autonomous‑agent policy) preserves invariants like “never fire without dual‑key authorization” or “funds only move when supply‑chain proofs are valid” before the code ever deploys, while the Merkle tree records every compiled binary and configuration hash so field auditors can show the running system is exactly the one that passed the proofs.

Silicon Valley fintechs are already racing to capture the civilian market. They’ll build the protocols, control the standards, and rent access back to the government. Or defense primes can recognize this moment: the regulatory green light to build sovereign, military-grade distributed systems that we control from the ground up.

The experimental phase is over. The GENIUS Act provides the legal foundation. The technology works at massive scale. The operational need is real.

The only question left is whether the national security establishment will seize this opportunity to build its own cryptographic infrastructure—or wake up in five years dependent on systems designed for retail banking, hoping they can be retrofitted for combat.

By One Comment
Autonomy

Robot Bees and other Little Robots

Researchers in China are crafting cyborg bees by fitting them with incredibly thin neural implants (article: Design of flight control system for honeybee based on EEG stimulation). The story isn’t that biology replaces electronics; it’s that biology supplies free actuation, sensing, and ultra low power computation that we can piggyback on with just enough silicon to plug those abilities into our digital world. Through leveraging natural flight capabilities and endurance of biological hosts, you have an ideal covert reconnaissance or delivery platform ideal for military surveillance, counterterrorism operations, physically hacking hardware and disaster relief missions where traditional drones would be too conspicuous or lack the agility to navigate complex environments.

So, there are lots of news articles on this, but the concept of miniature flying machines is far from new. For over a decade before projects like Harvard’s pioneering RoboBee took flight in 2007, researchers were already delving into micro-aerial vehicles (MAVs), driven by the quest for stealthy surveillance and other specialized applications. Early breakthroughs, such as the untethered flight of the Delft University of Technology’s “DelFly,” laid crucial groundwork, proving that insect-scale flight was not just a sci-fi dream but an achievable engineering challenge. This long history underscores a persistent, fascinating pursuit: shrinking aerial capabilities to an almost invisible scale.

Unlike the “China has a robot Bee!” articles, I wanted to step back and survey this space and where it’s going, and, well, this is kinda a big deal. The plot below reveals the tug-of-war between a robot’s mass and its flight endurance. While a general trend shows heavier robots flying longer, the interesting data are in the extremes. Beijing’s “Cyborg Bee” is a stunning outlier, achieving incredible flight times (around 45 minutes) with a minuscule 0.074-gram control unit by cleverly outsourcing the heavy lifting to a living bee – a biological cheat code for endurance. Conversely, the “NUDT Mosquito-Sized Drone” pushes miniaturization to its absolute limit at 0.3 grams, but sacrifices practical flight time, lasting only a few minutes . This highlights the persistent “holy grail” challenge: building truly insect-scale artificial robots that can fly far–and obey orders. Even cutting-edge artificial designs like MIT’s latest robotic insect, while achieving an impressive 17 minutes of flight, still carry more mass than their biological counterparts. The plot vividly demonstrates that while human ingenuity can shrink technology to astonishing scales, nature’s efficiency in flight remains the ultimate benchmark.

Robo Bee Positioning

Researchers figured out how to deliver precise electronic pulses that create rough sensory illusions, compelling the bee to execute flight commands—turning left or right, advancing forward, or retreating on demand. And this works. During laboratory demonstrations, these bio-hybrid drones achieved a remarkable 90% command compliance rate, following directional instructions with near-perfect accuracy. By hijacking the bee’s neural pathways, China has effectively weaponized nature’s own engineering, creating a new class of biological surveillance assets that combine the stealth of living organisms with the precision of electronic control systems.

A Bee’s Brain

The Beijing team’s breakthrough lies in their precise neural hijacking methodology: three hair-width platinum-iridium micro-needles penetrate the bee’s protocerebrum, delivering charge-balanced biphasic square-wave pulses at a soothing ±20 microamperes—just above the 5-10 µA activation threshold for insect CNS axons yet safely below tissue damage limits. These engineered waveforms, typically running at 200-300 Hz to mimic natural flight central pattern generators, create directional illusions rather than forcing specific muscle commands. Each pulse lasts 100-300 microseconds, optimized to match the chronaxie of 2-4 micrometer diameter insect axons.

Figuring this out is more guessing than careful math. Researchers discovered these “magic numbers” through systematic parametric sweeps across amplitude (5-50 µA), frequency (50-400 Hz), and pulse width (50-500 µs) grids, scoring binary outcomes like “turn angle >30°” until convergence on optimal control surfaces. Modern implementations use closed-loop Bayesian optimization with onboard IMUs and nRF24L01 radios, reducing tuning time from hours to ~90 seconds while adding ±10% amplitude jitter to counteract the bee’s rapid habituation response.

You can’t figure this out on live bees. To get these measurements, the honeybee head capsule is isolated from the body while leaving the entire brain undamaged. The head is cut free and placed into a dye-loaded and cooled ringer solution in a staining chamber for 1 hour, then fixed in a recording chamber, covered with a cover-slip, and imaged under the microscope. After you do this, you can conduct experiments on hundreds of honeybees using low and high current intensity. After stimulation, the Isolated Pulse Stimulator was modulated to generate a dissociating pulse (20 µA DC, 15–20 s), which partially dissociated Fe³⁺ from the stimulating electrode and deposited it in surrounding brain tissue marking where they stimulated for post-mortem analysis.

Precise probe placement relies on decades of accumulated neuroanatomical knowledge. Researchers leverage detailed brain atlases created through electron microscopy and confocal imaging, which map neural structures down to individual synaptic connections. Before inserting stimulation electrodes, they consult these three-dimensional brain maps to target specific neural clusters with sub-millimeter accuracy. During experiments, they verify their targeting using complementary recording techniques: ultra-fine borosilicate glass microelectrodes with 70-120 MΩ tip resistance penetrate individual neurons, capturing their electrical chatter at 20,000 samples per second. These recordings reveal whether the stimulation reaches its intended targets—researchers can literally watch neurons fire in real-time as voltage spikes scroll across their screens, confirming that their three platinum-iridium needles are activating the correct protocerebral circuits. This dual approach—anatomical precision guided by brain atlases combined with live electrophysiological validation—ensures the cyborg control signals reach exactly where they need to go.

What was cool here is that they found a nearly linear relationship between the stimulus burst duration and generated torque. This stimulus-torque characteristic holds for burst durations of up to 500ms. Hierarchical Bayesian modeling revealed that linearity of the stimulus-torque characteristic was generally invariant, with individually varying slopes. This allowed them to get generality through statistical models accounting for individual differences.

Learning the signals is only half the battle. The power challenge defines the system’s ultimate constraints: while a flying bee dissipates ~200 milliwatts mechanically, the cyborg controller must operate within a ~10 mW mass-equivalent budget—about what a drop of nectar weighs. Current 3-milligram micro-LiPo cells provide only ~1 milliwatt-hour, forcing engineers toward energy harvesting solutions like piezoelectric thorax patches that scavenge 20-40 microwatts from wingbeats or thermoelectric films adding single-digit microwatts from body heat. Yes, the bee is the battery.

This power scarcity drives the next evolution: rather than imposing external commands, future systems will eavesdrop on the bee’s existing neural traffic—stretch receptors encoding wingbeat amplitude, Johnston’s organs signaling airspeed, and antennal lobe spikes classifying odors with <50ms latency at just 5 µW biological power. Event-driven neuromorphic processors like Intel’s Loihi 2 can decode these spike trains at <50 microjoules per inference, potentially enabling bidirectional brain-computer interfaces where artificial intelligence augments rather than overrides the insect’s 100-million-year-evolved sensorimotor algorithms.

Intel’s Lohi 2 Neuromorphic Chip

Challenges

Power remains the fundamental bottleneck preventing sustained cyborg control. A flying bee burns through ~200 milliwatts mechanically, but the electronic backpack must survive on a mere ~10 milliwatt mass budget—equivalent to carrying a single drop of nectar. Current micro-LiPo cells weighing 3 milligrams deliver only ~1 milliwatt-hour, barely enough for minutes of operation. Engineers have turned to the bee itself as a power source: piezoelectric patches glued to the thorax harvest 20-40 microwatts from wingbeat vibrations, while thermoelectric films capture single-digit microwatts from body heat. Combined, these provide just enough juice for duty-cycled neural recording and simple radio backscatter—but not continuous control. The thermal constraint is equally brutal: even 5 milliwatts of continuous power dissipation heats the bee’s dorsal thorax by 1°C, disrupting its olfactory navigation. This forces all onboard electronics to operate in the sub-milliwatt range, making every microjoule precious. The solution may lie in passive eavesdropping rather than active control—tapping into the bee’s existing 5-microwatt neural signals instead of imposing power-hungry external commands.

I summarize the rest of the challenges below:

ChallengeCandidate approachPractical limits
RF antenna sizeFolded‑loop 2.4 GHz BLE or sub‑GHz LoRa patterns printed on 20 µm polyimideNeeds >30 mm trace length—almost bee‑size; detunes with wing motion
Power for TxAmbient‑backscatter using 915 MHz carrier (reader on the ground)100 µW budget fits; 1 kbps uplink only (ScienceDaily)
Network dynamicsBio‑inspired swarm routing: every node rebroadcasts if SNR improves by Δ>3 dBScalability shown in simulation up to 5 000 nodes; real tests at NTU Singapore hit 120 roach‑bots (ScienceDaily)
LocalizationFusion of onboard IMU (20 µg) + optic‑flow + Doppler‑based ranging to readersIMU drift acceptable for ≤30 s missions; longer tasks need visual odometry

The cyborg bee’s computational supremacy becomes stark when viewed through the lens of task efficiency rather than raw processing power. While silicon-based micro-flyers operate on ARM Cortex processors churning through 20-170 MOPS (mega-operations per second), the bee’s million-neuron brain achieves full visual navigation on just 5 MOPS—neurons firing at a leisurely 5 Hz average. This thousand-fold reduction in arithmetic operations masks a deeper truth: the bee’s sparse, event-driven neural code extracts far more navigational intelligence per computation.

$$ \text{SEI} = \frac{\text{MAP vox}/\text{sec} \times \text{Pose bits}}{\text{MOPS}}$$

To show this, I made up a metric that combines each vehicle’s vision rate, positional accuracy and onboard compute into a single “task‑efficiency” score—the Simultaneous localization and mapping (SLAM)-Efficiency Index (SEI)—so we could compare a honey‑bee brain running on microwatts with silicon drones running on megahertz. Simply put, SEI measures how many bits of world-state knowledge each platform generates per unit of processing. With SEI, the bee achieves 290,000 bits/MOPS—over 100 times better than the best silicon autopilots. Even DJI’s Mini 4K, with its 9,600 MOPS quad-core processor, manages only 420 bits/MOPS despite throwing two thousand times more compute at the problem.

This efficiency gap stems from fundamental architectural differences: bee brains process visual scenes through parallel analog channels that compress optic flow into navigation commands without digitizing every pixel, while our drones waste cycles on frame buffers and matrix multiplications. The bee’s ring-attractor neurons solve path integration as a single rotation-symmetric loop, updating its cognitive map 200 times per second (synchronized with wingbeats) using mere microwatts. Silicon systems attempting equivalent SLAM functionality—feature extraction, bundle adjustment, loop closure—burn 5-15 watts on embedded GPUs. Until neuromorphic processors can match the bee’s event-driven, power-sipping architecture, cyborg insects will remain the only sub-gram platforms capable of autonomous navigation. The chart’s below tells the story on a logarithmic scale: that lonely dot at 290,000 SEI represents not just superior engineering, but a fundamentally different computational paradigm. Note that flapping‑wing robots cluster below 200 MOPS because that’s all a 168 MHz Cortex‑M4 can supply; adding bigger processors blows the weight budget at these sizes.

The race to miniaturize autonomous flight reveals a fundamental truth about computation: raw processing power means nothing without the right architecture. While teams worldwide chase incremental improvements in battery life and chip efficiency, Beijing’s cyborg bee has leverages biology to solve this problem. Just as xAI’s Grok 4 throws trillion-parameter models at language understanding while smaller, cleaner models achieve similar results with careful data curation, bees handle similar complexity with microwatts. The lesson isn’t to abandon silicon for biology, but to recognize that sparse, event-driven architectures trump dense matrix multiplications when every microjoule counts.

Looking forward, the convergence is inevitable: neuromorphic processors like Intel’s Loihi 2 are beginning to bridge the efficiency gap, while cyborg systems like Beijing’s bee point the way toward hybrid architectures that leverage both biological and silicon strengths. The real breakthrough won’t come from making smaller batteries or faster processors—it will come from fundamentally rethinking how we encode, embed, and process information. Whether it’s Scale AI proving that better algorithms beat bigger models, or a bee’s ring-attractor neurons solving navigation with analog elegance, the message is clear: in the contest between brute force and biological intelligence, nature still holds most of the cards.

By 4 Comments
Engineering & Code

Chinese Researchers Break 22-Bit RSA Encryption with Quantum Computer

A Chinese research team just pushed quantum computing one step closer to breaking the encryption that secures your bank account. The security that protects your credit card, banking, and private messages relies on a simple mathematical bet: it would take classical computers longer than the age of the universe to crack a 2048-bit RSA key. That bet is looking increasingly risky as quantum computers inch closer to breaking the math that secures the internet.

RSA encryption is a widely used asymmetric algorithm crucial for secure digital signatures, web browsing (SSL/TLS), online banking, messaging services, VPNs, and cloud infrastructure. The web depends on RSA and, given some recent news, it’s good to take a couple minutes and get a peek at where this is going.

This month (June 2025), researchers at Shanghai University (Wang, et al) published a study in the Chinese Journal of Computers that used a D-Wave Advantage quantum annealer to factor a 22-bit RSA integer—beating the previous 19-bit record by recasting factoring as a QUBO (Quadratic Unconstrained Binary Optimization) problem. While 22 bits falls far short of the 2048-bit keys securing the internet today, it demonstrates that quantum annealers can scale beyond past limits. The researchers demonstrated that quantum annealing can turn cryptographic attacks into combinatorial optimization problems.

Headlines can be deceptive—another Chinese group “hacked” lightweight block ciphers in 2024, and a D-Wave demo factored a contrived 2048-bit semiprime whose twin primes differed by just two bits—but genuine progress is visible in the research community.

IBM Starling

Quantum annealers differ fundamentally from classical computers by utilizing quantum bits (qubits), which exploit phenomena like superposition and entanglement. This enables quantum annealers to efficiently solve certain optimization problems. Although limited in application compared to general-purpose quantum computers, quantum annealers’ recent progress highlights their potential impact on cryptographic security.

Quantum Annealer
Shor’s Algorithm

To make sense of this, I made a visualization to compare quantum annealing to Shor’s Algorithm. The “Quantum Annealer” side shows particles behaving in a fluid, probabilistic manner – they move chaotically and gradually converge toward a central point, representing how quantum annealers explore solution spaces by leveraging quantum superposition and tunneling effects to find optimal solutions through a process of gradual energy minimization. On the right, “Shor’s Algorithm” displays particles organizing into a rigid, structured grid pattern, illustrating the deterministic, step-by-step nature of classical quantum algorithms that follow precise mathematical procedures to solve specific problems like integer factorization. The contrast between the organic, exploratory movement on the quantum annealer side and the methodical, ordered arrangement on the classical algorithm side captures the essential difference between these two quantum computing paradigms: one uses quantum mechanics to probabilistically search for solutions, while the other uses quantum mechanics to execute deterministic algorithms with exponential speedup over classical computers.

Factoring large integers remains the yardstick for quantum progress because RSA’s safety rests on how brutally hard that task is for classical machines. Jumping from a 19-bit to a 22-bit crack looks tiny beside a 2,048-bit key, yet the pace signals that quantum methods—whether Shor’s algorithm or annealing—are gaining real traction.

RSA is safe for the moment, but the danger is time-shifted. Attackers can copy ciphertext today and decrypt it when hardware catches up, so any data that must stay private into the 2030s already needs stronger wrapping. Agencies and corporations are mapping every legacy backup, TLS endpoint, and VPN tunnel that still leans on RSA, then swapping certificates and firmware signatures for post-quantum or hybrid ones as they come up for renewal. The White House, NIST, and NSA’s CNSA 2.0 suite have turned that housekeeping into policy, naming lattice schemes such as CRYSTALS-Kyber, Dilithium, and SPHINCS+ as the new default. Migration is messy only for systems without “crypto agility,” the design principle that lets you change algorithms as easily as you update software.

Elliptic-curve keys used in Bitcoin and Ethereum sit even lower on the future quantum cost curve, but they, too, will fall once hardware scales. The immediate leak risk lies in records with decade-long value—medical histories, merger drafts, cold wallets, long-lived SSL certs—because those files are already being siphoned for eventual decryption.

Quantum road maps show why urgency is justified. Google’s 105-qubit Willow chip has crossed the error threshold for full fault tolerance, and IBM projects two thousand logical qubits by 2033—enough to threaten roughly 1,000-bit RSA keys.

Specifically, IBM’s roadmap targets a fault-tolerant quantum computer with 200 logical qubits by 2029 (Starling) and 2,000 by 2033 (Blue Jay). Under Shor’s algorithm, factoring an \(n\)-bit RSA key requires roughly \(2n + 2\) logical qubits , so IBM’s Blue Jay could break RSA keys up to about 999 bits by 2033. Crucially, Shor also solves discrete-log problems, meaning 256-bit ECC keys—the basis for Bitcoin’s ECDSA and Ethereum’s EdDSA—would fall with far fewer qubits, making ECC instances more vulnerable to future quantum attacks than RSA at comparable classical security levels.

Start-ups betting on photonic qubits promise faster scaling, while skeptics such as Scott Aaronson note that engineering overhead grows faster than headline qubit counts suggest. Regulators aren’t waiting: Washington demands server inventories by 2027, Brussels ties NIS2 fines to quantum-safe readiness, and Beijing is pouring billions into domestic chip fabs.

Researchers track logical-qubit lifetimes, error-corrected cycles, and real factoring benchmarks rather than raw qubit totals. A lab that links thousands of physical qubits into hundreds of long-lived logical ones would mark the tipping point where costs start resembling supercomputers, not billion-dollar prototypes.

History rarely shouts as clearly as a Chinese factoring record and a government migration order in the same quarter. It’s wise to treat post-quantum upgrades as routine maintenance—rotate keys, adopt algorithm-neutral APIs, replace certificates on schedule—and a million-qubit announcement will be just another headline. Delay, and every RSA-protected archive on your network becomes a ticking clock, counting down to disclosure day.

Special thanks to Tom Berson, Brian LaMacchia and Pat Lincoln for insights into this area.

By 0 Comments

© 2025 TIM BOOHER