Essays

Aerospace and engineering thinking, applied to meaning, work, and the inner life. For technical founders and senior leaders rebuilding the thing they built when it stopped fitting.

For founders and technical leaders

You built a company, or a senior technical function, and it stopped fitting you. Start here.

Why RFA Should Celebrate Isar Aerospace Reaching Orbit

Isar Aerospace reached orbit on September 5, 2026, on only its second flight, the first privately funded European rocket to do it. Its closest rival, Rocket Factory Augsburg, has flown 0 times, so it had no market share to lose, and everything around it moved: investors, satellite customers, spaceports, engineers, and the ESA contracts that require orbit before 2028. The pattern is the 4-minute mile: John Landy had run 4:02 six times and never under, then ran 3:57.9 just 46 days after Roger Bannister, and 2,349 men have followed. The record is zero-sum, the club is not. For technical founders: why a rival's win is free growth while customers still doubt the product works, how to tell when the market flips to price and delivery, and what the last rival win actually cost you.

September 7, 20265 min read

SpaceX Is Boeing in 1929

Boeing sells you the 787 but not the engines: an airline picks GE or Rolls-Royce separately, and the engine makers earn a very large share of the money in aviation. That split was not natural. In 1929 Boeing, Pratt & Whitney and United Air Lines were one company, until the Air Mail Act of 1934 broke them into three, and the punishment became the shape of the industry for 90 years. SpaceX, Rocket Lab and Blue Origin build everything under one roof today, the way Boeing did in 1929, and the split is starting (Blue Origin sells the BE-4 to ULA). For technical founders: why the split is the way in that doesn't require beating the leader, and why the engine makers' flying testbeds teach the rule that keeps an experiment readable: change one variable at a time.

September 7, 20266 min read

Europe Will Have a SpaceX Competitor by 2049

In the mid 1970s Airbus had 16 finished airliners parked at Toulouse that nobody wanted to buy. It went on to pass Boeing, and the climb never came from building a better 747 (its one direct try, the A380, never earned back its development money): it bet on a requirement nobody had written down (2 engines when serious meant 3 or 4), bought its engine ready-made from GE instead of designing one, and cut airlines' pilot costs for decades with cockpit commonality. For technical founders: why copying the leader is the expensive option, what the Airbus play would look like against Starship, and why the gap between being right and being seen to be right can be 20 years wide.

August 28, 20267 min read

They Flew It 1,700 Times. It Barely Changed.

The Soyuz family has flown roughly 8,500 rocket engines since 1957, more than every Merlin SpaceX has ever launched, and the engine is close to what it was in 1957. The Space Shuttle got all 405 of its engine flights back and inspected every one, and produced the most expensive rocket engine ever operated. So neither volume nor teardowns explain SpaceX, and the real answer is narrower: lots of data, lots of volume, and the freedom to keep changing the design. For technical founders: why your loop is probably too long, and how to start from the thing that needs to improve rather than from where the data happens to be.

August 26, 20267 min read

The Physics of Rock Bottom

Starcloud just raised $250M at a $2.3B valuation to put data centers in orbit. Orbital mechanics has a rule for low points: a burn at the bottom of the orbit raises the far side of the path, and the same fuel buys the most change where you are moving fastest. A 25-year University of Michigan study of 2,420 lives, published in Nature Mental Health the same week, found the human version: the people who started lowest had the highest odds of marked improvement. For founders in a hard season: the low point is a place on the path, not the end of the story, and each point on the loop has its own burn.

August 21, 20266 min read

The Recipe Was Public. The Cooking Took 11 Years.

China's LandSpace just landed an orbital-class booster on solid ground, the 2nd company ever to do it after SpaceX. Every step was public: SpaceX livestreamed the maneuver in 2015, the physics is in textbooks, and for years an AI would walk you through it in 10 seconds. It still took 11 years and 1 crash. For technical founders: why the knowledge was never the moat, what AI actually compresses and what it cannot, and why simple and easy are different things.

August 20, 20266 min read

Stop Collecting Moves

SoftBank just put $200M into machines nobody taught how to dig, because no motion can be recorded when the ground is different with every bucket. The lab behind it never gave its robot a gait either. It taught it when to trust its own reading of the ground and move briskly, and when to shorten the step. For technical founders: why the moves you earned yourself carry unnoticed prerequisites and quietly stop paying, and how the layer above them comes apart into 3 things you can practice.

August 19, 20266 min read

The Answer Was 42

Figure is on its 7th robot hand and threw out the one design 1X still ships. Its CEO calls tendons a complete local maximum; 1X runs the same architecture at 25 degrees of freedom and rates it past 2 million cycles. Two serious teams, the same mechanism, opposite verdicts. For technical founders: why a quick win is a small hill, why no two of us are standing on the same terrain even when the problem looks identical, and the difference between being told what another person can see and being helped to go and look from several other places yourself.

August 15, 20266 min read

It Works. It Costs You Both Hands.

For 12 years every walking exoskeleton cleared for home use borrowed both of the user's arms. On August 11, 2026 the FDA cleared Wandercraft's Eve, the first that balances itself. For technical founders and senior leaders: why a fix that works can still be expensive, how to spot the places your own hands are holding something upright, and how to move the balancing into the system instead.

August 13, 20266 min read

Most Things Aren't Broken. They're Out of Fuel.

On July 21, 2026, a robot with two arms launched to fly a year across space and bolt jetpacks onto aging satellites, extending them at least 6 years instead of replacing them. For technical founders and senior leaders: why we mistake being low on fuel for being broken and reach for a costly rebuild, why maintenance stays invisible while launches get the applause, and how to tell what only needs refueling from what the world has truly moved past.

August 13, 20266 min read

What Did You Put on the Test Stand This Week?

Rocket Lab just posted the best quarter in its history, and the only number anyone asked about was a launch date its CEO says not to watch. For technical founders and senior leaders: why an outcome date is a poor instrument for your own progress even though a close, public deadline really does make you finish, why a tank that never ruptures in testing has not been qualified, and how to swap the deadline you keep re-promising yourself for a weekly count you control.

August 12, 20268 min read

Effort Sets the Floor, Not the Ceiling

A former Rocket Lab engineer raised 7.35 million dollars for Outlier Space, which makes materials in orbit where felt gravity falls about a millionfold, because some crystals and drugs only form right in free fall. For technical founders and senior leaders: why effort raises your floor but never your ceiling, why venture returns follow a power law in which about 4 in 100 bets carry almost everything, and how to back your real potential and let it change instead of trying to force a 10x with more hours.

August 11, 20266 min read

Change the Plan, Not the Vision

SpaceX built the world's largest supercomputer to train an AI to beat Anthropic, then rented most of it to Anthropic for about 1.25 billion dollars a month, and it still pays for the rockets. For technical founders and senior leaders: why a bet that looks like a loss still feeds the goal when the vision above it does not move, why you change your plans freely but your vision least often and most carefully, and how to build so every try, win or lose, becomes a data point instead of a dead end.

August 6, 20267 min read

Your Moat Just Became a Download

Google's Gemini Robotics 2 can teach a robot a body it has never seen from fewer than 200 examples, turning a skill that once needed a PhD into something close to a download. For technical founders: why a hard skill is a head start and not a moat, why the same cheapening is also your lever to disrupt, and how the real moat splits in two, the company's (what you own that is scarce) and yours (how fast you learn the next hard thing).

August 5, 20266 min read

The Problem You Want to Solve Isn't the One in Your Way

K2 Space doubled its value in seven months. A company, or a career, grows at the rate it finds the single thing most in its way and clears it, then the next. For technical founders and senior leaders: solve the problem that actually limits you, not the one you want to solve, and start early on the constraints whose clock you do not own.

August 1, 20266 min read

Which Robot Has the Best Hip?

Between two models, Unitree reversed the order of the joints in its robot's hip. The world's best humanoid makers cannot agree on how to build one, and a 2025 running-robot deliberately chose the older order to keep its legs light. There is no best hip, only the best hip for what you are optimizing for. For founders and anyone building a life: why copying the leader's answer imports requirements that were never yours, and how to copy the reasoning instead.

July 30, 20266 min read

Don't Outsource Your Vision

This week a single decision made an entire class of robots illegal overnight, aimed at China's Unitree. A year earlier Tesla's Optimus got hit from the other side, stalled by a rare-earth magnet China would not export. Neither company changed its vision, only its supply chain. For founders and anyone building a life: why an outside shock, a ban, a shortage, a layoff, should change your plans but never your vision, and how to stop outsourcing the thing your whole future depends on.

July 29, 20266 min read

Integration Multiplies Opportunities and Problems

Rocket Lab put a satellite in orbit 16 hours after the order, then agreed to spend 8 billion dollars on Iridium to integrate even further. Vertical integration is a multiplier: it made them fast because their parts were already tight, and it would spread a loose system's problems just as fast. For technical founders and senior leaders deciding to take on more, why adding multiplies whatever state your system is already in.

June 29, 20266 min read

The Work Nobody Films

A viral robot-hand demo hides millions of invisible test cycles. The real work, in a company and in you, is the unseen reliability, not the breakthrough.

July 13, 20264 min read

The Thing You Built Stopped Fitting You

A design choice made for the Boeing 737 in the 1960s contributed to the deaths of 346 people half a century later, across two crashes. Not because the design was wrong, but because the requirements changed and a chain of patches grew unchecked. The same thing happens to founders and senior leaders: the original misalignment is rarely what breaks you, the unexamined patches stacked on top of it are.

June 8, 20268 min read

The Founder Who Outgrew Their Own Company

The airlines you fly today started by flying the mail, and survived by letting the mail go. For technical founders and senior engineers whose company, role, or product has quietly stopped fitting: how to find the real requirement under the thing you built, and why outgrowing it is a signal, not a failure.

June 16, 20267 min read

The Decision a Founder Can't Make: Pivot, Sell, Step Back, or Stay

Pilots compute a decision speed, V1, before every takeoff: the line past which not choosing is itself a choice. For technical founders stuck on the big decision, pivot, sell, step back, or stay: how to tell real waiting from avoidance, compute your own V1, and find a place to think when everyone close to you has a stake.

June 16, 20267 min read

More founder essays are landing here over the coming weeks.

The engineering-a-life series

Elon Musk's five-step algorithm, the overview plus one essay per step, in reading order.

The Five-Step Algorithm for Redesigning Anything

SpaceX deleted Super Heavy's landing legs and catches the booster with the launch tower instead. The decision came from a five-step algorithm whose steps are ordered by cost: question the requirements, delete, simplify, accelerate, automate. Most of us run it backwards, on our companies and our lives, and pay the most for what we should have deleted for free.

June 12, 20267 min read

Stop Optimizing Your Life. Start Specifying It.

I spent ten years getting really good at the wrong things. Then I learned something from Elon Musk's five-step algorithm for building rockets. It also works for building a life.

March 15, 202410 min read

Don't Build an SLS

NASA's newest moon rocket flies with engines from 1982 because Congress mandated heritage hardware. Most of us build our lives the same way, around parts we never picked.

April 28, 20267 min read

The Best Part Is No Part

SpaceX's newest rocket engine is missing a heat shield, a fire suppression system, and most of its external plumbing. The engineers there have a phrase for this approach. It might be one of the most important things I have ever read about how to live.

May 6, 20269 min read

Delete Where It Hurts

Surgical robots got better by deleting parts: the da Vinci went from four arms and four holes to one arm and one 2.5 cm port. In my University of Tokyo PhD we deleted ten of twelve electromagnetic coils on paper and gained a bigger workspace. The lesson for technical founders and senior leaders: deletion is not a diet. Delete at the boundary where failure is expensive, and let necessary complexity live where slack is cheap.

July 2, 20267 min read

Why Your Life Won't Run Smoothly Until You Stop Trying

We are always trying to make life more efficient, and that is the problem. Simplification is the third step of an engineering framework for living. Not minimalism, not optimization. It is finding the interfaces in your week and folding two things into one.

May 19, 20268 min read

The Weekly Retrospective: How to Run Sprints on Your Life

A rocket engine line that builds four hundred engines a year has fewer defects per engine than one that builds twenty. High cadence forces quality. The fourth step of an engineering framework for living: speed up the cycle, and run a five-minute weekly retrospective so you find problems while they are still small.

May 31, 20268 min read

Don't Automate a Life That Shouldn't Exist

Automation is the fifth and last step of an engineering framework for living, and the one most people reach for first. The moment you automate something, you marry it: you commit to maintaining it forever. Automating a bad process does not fix it, it cements it. When to build, when to question the requirement instead, and why the last step points back to the first.

June 4, 20268 min read

On failure, risk, and decisions

What aerospace and cave diving teach about the failures that actually get you, and the decisions you keep deferring.

Fail Faster to Get There Sooner

One second before liftoff, Starship Flight 13 refused to fly. The surest way to reach something almost impossible is to fail at it as fast, as cheaply, and as often as you can, fixing every step on the way. On a cold email to the president of SpaceX, and what a rocket that stopped itself teaches about the failures worth having.

July 24, 20267 min read

Reusable Rockets, Disposable Joy

For most of the space age we threw every rocket away, until we learned to catch them and fly them again. Most of us still throw our own joy away the same way, one achievement at a time. The world got measurably better over the last century while our reported happiness stayed flat, and we barely notice either. What a caught booster reveals about restarting joy and peace: reusable over disposable, powered by meaning rather than money, and eventually built for everyone.

July 27, 20268 min read

Grounded Is Not Broken

Japan's H3 rocket lost a satellite because a payload adapter had delaminated months earlier, in a factory oven. The failure was old before it was visible. On damage that hides between the layers, the watermelon-green status report, and why grounded is not broken. It is tracing.

June 13, 20268 min read

Nothing Went Wrong for Ten Years

The failure that gets you is almost never the dramatic one. It is the small drift that went fine for years, until the day it did not. An aerospace engineer and cave diver on the normalization of deviance, for founders and senior leaders, and the skill of noticing the quiet erosion early.

June 7, 20269 min read

The Safest Rocket Never Launches

The only rocket with a zero percent chance of failure is the one that never leaves the pad, and that one has already failed. For founders and senior leaders sitting on a decision that will not move: how to tell the safe choice that protects something real from the one that just keeps you on the ground because flying is uncomfortable.

June 7, 20268 min read

It's Just a Way

There are two ways to bring a crew home from space: the Americans splash into the ocean, the Soviets land on dirt in Kazakhstan. Neither is the clever way and the other the workaround. Each side just used what it had. On unknown unknowns, the assumptions you cannot see until you stand somewhere else, and what raising kids across five countries taught an aerospace engineer about the laws that are only constraints he stopped noticing.

July 7, 20266 min read

You Don't Prevent the Failure. You Make It Safe.

Fail-safe engineering does not try to build something that never fails. It makes sure that when a part fails, it fails into a safe state. An aerospace engineer turned coach on why the most important thing he does has almost nothing to do with the questions he asks: building a space safe enough for the true thing to be said out loud.

June 4, 20267 min read

Burnout Is Not a Bug. It's a Spec Mismatch.

Burnout is not a personal weakness. It is what happens when a real human spec sheet is run inside a system whose requirements do not match. An engineer's view on what actually breaks, and why fixing the human is the wrong fix.

May 1, 202610 min read

Personal

The road here. NASA, Tokyo, a wall of rejections, and learning to read my own instruments.

I Got to NASA. The Search Didn't End.

I wrote my diploma thesis at NASA, did a PhD in Tokyo, applied to be an astronaut. Then an astronaut on a private tour showed me the search has no finish line. It only upgrades. What the search cost me, and what finally answered the question it was asking.

May 16, 20269 min read

My CV of Failures

A toddler falls down hundreds of times before walking. Adults trying something hard get a much emptier room. Here, for the record, is a partial list of the times I fell down, and what I notice now from this side of it.

May 1, 20269 min read

Tight Chest. Less Tight Chest. That Was the Whole Reading.

I sat on a call with ten coaches. They could read their bodies in 4K. I had two channels. Here's what happened when I started taking emotional data seriously.

April 30, 20266 min read

Get new essays by email

New essays, plus the occasional thing I only send by email.