mihas1o

From zero to AI Agent Engineer. Crypto × AI

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

Every procedure you follow was written by someone who did not know about today.

That works until it doesn't. And the moment it stops working is the only moment that ever mattered.

Dekker spent a career on that gap: what actually holds a complex system together when the rulebook runs out. It is never the rulebook.

Information, execution, credentials are all rulebook. Three of your four skill types expire the day the situation stops matching the manual.

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

A senior title tells you someone survived. It does not tell you they learned anything.

Kahneman's test for expertise was never seniority. It was whether the environment was regular enough to teach, and whether the person got real feedback from it.

Most of what sits on your resume was earned in environments that no longer exist. Docs rewritten. Tools deprecated. Fields moved.

Three of the four skill types expire that way. The fourth one has no expiry date and no certificate, so almost nobody tracks it.чс

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

You have been collecting skills that were designed to rot.

Docs change. Tools die. Fields move. Three of the four types you own are on a clock, and the clock started the day you learned them.

The fourth one never expires. It also never appears on a resume.

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

An hour saved anywhere except the bottleneck is not saved. It's a mirage.

Eli Goldratt built a whole theory on that one sentence, and it explains why your productivity efforts keep failing.

Every system has one constraint. One thing that actually sets how much gets through. Improve anything else and the system's output does not move, because it was never what was holding you.

You feel productive. Nothing arrives faster.

He was blunt about where this leads in practice. Stop starting. Start finishing.

Which sounds like a slogan until you notice most of us do the exact opposite, all day, and call it being responsive.

Goldratt was a physicist. He wrote The Goal in 1984 as a novel because nobody would read the theory. It sold millions and is still required reading in operations programmes forty years later.

He gave his last talk in 2011 and died shortly after returning to Israel.

[mihas1o @mihas1o]

https://t.co/PtC4IOgb1E

[mihas1o @mihas1o]

32,000 views in ten years.

That's the entire audience for the talk that explains why your week keeps falling apart.

I went looking for whoever first proved this, and found the exact chart I had drawn myself.

Don Reinertsen. In 1991 he published the first paper showing that queueing theory applies to product development. Not factories. Desk work.

Thirty eight minutes in, he puts up a slide called The Effect of Capacity Utilization. Queue size on one axis, how booked you are on the other. Flat, flat, flat, then vertical after 80%.

I built that same curve for an article this week without knowing his slide existed. It's not a coincidence. It's a formula, and there is only one shape it can make.

His finding underneath it is the part people miss.

In manufacturing the queue is visible. Boxes between machines, anyone can count them. In knowledge work the identical queue exists and nothing sums it. One item in a thread, one in a doc, one in a promise you made in a kitchen.

So it grows for years, and the only symptom is that everything takes longer than it used to.

And the line I keep returning to: most people think the sooner they start something, the sooner it gets done. It's the opposite. Starting is what slows down everything already moving.

He lists this talk on his own site. Free, since 2015.

[mihas1o @mihas1o]

https://t.co/PtC4IOgb1E

[mihas1o @mihas1o]

@AxiomExchange already set up my bnb wallet

[mihas1o @mihas1o]

Deming used to run an experiment where volunteers scooped beads out of a box. White beads were good. Red beads were defects.

Then he managed them.

He praised the ones who got fewer red beads. He warned the ones who got more. He put up quality posters. He ranked them. He fired the worst performer and promoted the best.

None of it changed anything, because none of them could. The proportion of red beads in the box was fixed. Every person was scooping from the same box with the same paddle.

The differences between them were noise, and he spent an hour treating noise as if it were performance.

That is the whole demonstration, and people who watched it in the eighties said it rearranged how they saw their own companies.

Here is why I keep thinking about it.

When your week goes badly you look for the person to correct, and the person available is you. Work harder. Get up earlier. Be more disciplined.

Sometimes that's right. Often you are scooping from a box you never looked inside.

[mihas1o @mihas1o]

https://t.co/PtC4IOgb1E

[mihas1o @mihas1o]

https://t.co/PtC4IOgb1E

[mihas1o @mihas1o]

For forty years an MIT professor gave the same one hour lecture every January. Every year the hall filled past capacity, people on the stairs, a crowd in the corridor straining to hear.

It was not about code. Not about maths. Not about AI, which is what he actually ran the lab for.

It was about how to speak.

Patrick Winston opened it by telling the room that their success would be determined mostly by how well they spoke, then by how well they wrote, and only then by the quality of their ideas. In that order.

He taught at MIT for almost fifty years. He died in 2019 and the tradition died with it.

The recording is the single most watched thing MIT has ever put online. Ahead of every engineering course they publish.

Which tells you something about what people actually needed and never got taught.

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

The world will try to define you by a skill.

That's why you have a major. That's why you have a job title. You're an engineer, then a front-end engineer, and then people hand you front ends and ask you to engineer them.

Bret Victor called that the path of the craftsman, and said plainly it's the most common one there is.

He spent the first half of this talk demonstrating tools he built that made a room full of programmers applaud repeatedly, which almost never happens at a technical conference.

Then he stopped and explained that none of it came from solving problems.

It came from a principle he decided on years earlier and refused to drop. Every tool in the demo was that one belief, applied.

The skills he used to build them are fourteen years old now and half of them are gone.

The principle isn't.

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

The man who spent forty years studying how people get good at things gave his last interview in 2020, then died a few weeks later.

Anders Ericsson. Florida State. The researcher whose work Gladwell turned into the ten thousand hour rule, which Ericsson then spent years explaining was not what he found.

What he actually found is less comfortable.

Hours don't do it. Experience doesn't do it. In several medical specialties, doctors twenty and thirty years in score worse on objective measures than people two years out of training.

They didn't get worse. Nothing in a normal working day pushed them past what they could already do, so nothing improved. The years accumulated. The skill sat still.

His distinction is narrow and it decides everything: repetition is not practice. Practice is working at the edge of what you can do, with feedback fast enough to correct you before the mistake sets.

Without that, you are not building a skill. You are automating one, including the parts of it that are wrong.

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

You can read something three times, feel like you know it, and know nothing.

Barbara Oakley has a name for that: the illusion of competence.

What happens is your brain confuses recognition with retrieval. The page looks familiar, so it registers as learned. Nothing has actually been stored in a form you can pull out when you need it.

The test is small and unpleasant. Close the material. Try to produce it from nothing. Most of what felt solid will not be there.

She points out you can memorise a large number of words in a language and still not know what a single one of them means. Recognition all the way down.

The reason this matters more now than it used to: recognition is free. Everything looks familiar because you have scrolled past a version of it. Feeling informed and being informed have never been further apart.

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

A professor gave his students a test on global health.

Two countries, which has higher child mortality. Five questions like that. The students were Swedish, top of their year, best grades in the system.

They scored worse than random.

Hans Rosling ran the same test on his faculty colleagues. Better than the students. Still worse than chance.

His line for it: they knew statistically significantly less about the world than chimpanzees, because a chimpanzee picking bananas at least gets fifty percent.

The students were not stupid and they were not uninformed. They had learned the world once, and the world had moved, and nobody told them.

That is what an expiring asset looks like from the inside. It doesn't feel like ignorance. It feels like knowing.

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

Seven people died 73 seconds after liftoff.

The engineers who built the booster had spent the night before the launch begging NASA not to fly. Too cold. The seals won't hold. They said it for hours.

Management overruled them.

Then came the commission, and a room full of people with the right credentials explaining that nobody could have known.

Richard Feynman was on that commission. He did what none of the others did: he went to the people who actually built the thing and asked them.

What he found was not a technical mystery. Management put the odds of losing a shuttle at one in a hundred thousand. The engineers doing the work put it at about one in a hundred.

Same agency. Same data. A thousandfold difference in what they believed.

Here is the part that should bother you.

Nobody in that story lacked information. The engineers had it. Management had the engineers. Feynman had the same reports everyone else had.

The difference was knowing which question would actually settle it, and being willing to ask people below you.

Credit: Richard Feynman, CNN interview, 1986.

I wrote about the difference between knowing things and knowing which thing matters. Three of your four assets expire. That one doesn't:

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

In 1968 a man walked on stage and showed a room full of engineers the mouse, hypertext, video calls, and two people editing the same document at the same time.

None of it existed anywhere else. It would take the industry twenty to thirty years to ship most of it.

Douglas Engelbart. Stanford Research Institute. December 9th, 1968.

Here is the part that should bother you.

The room did not miss it. Xerox took the ideas within a few years, Apple took them from Xerox, everyone took them from Apple. Your screen right now is his demo, shipped.

They took the mouse. They took the windows.

They did not take the point.

Engelbart was not building an interface. He said so plainly: the work was about augmenting human intellect, changing how people solve hard problems together. That part had no product in it, so nobody picked it up, and he spent decades watching the world use his tools for everything except the thing he built them for.

The execution transferred. The judgment behind it did not.

I wrote about that split. Three of your four assets expire. The one that doesn't is the one nobody can copy off you:

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

The NSA said the recording didn't exist.

It existed. It had been sitting in their vault for 42 years.

When they finally admitted that, they gave a stranger reason for keeping it: they had no machine that could play it, and no law forced them to go find one.

On the tape was Grace Hopper. The woman who built the first compiler and helped create COBOL, speaking to the NSA workforce at Fort Meade in 1982 about what was coming.

It took the National Archives to get it out. Released August 2024. Forty two years late.

Here is the part that should bother you.

Nothing about the knowledge decayed. Her ideas were fine. The machine that could read them stopped existing, and that was enough.

That is exactly how a skill dies. Not because you got worse. Because the thing it lived inside stopped being played.

Credit: Rear Adm. Grace Hopper, NSA, 1982. Released by the NSA in 2024 with help from the National Archives.

I wrote about which of your four assets that happens to, and the one it can't touch:

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

An economist spent years trying to find out what a degree is actually paying you for.

His answer: mostly not the skills.

Bryan Caplan, George Mason University. His estimate is that around eighty percent of what education adds to your income comes from signalling, not from anything you learned.

The evidence that convinced him is small and hard to argue with.

The jump in income comes from finishing. Not from the years. Two people can sit through almost the same amount of school, and the one who walks out with the paper earns far more than the one who walks out without it. If you were being paid for what you know, the last sheet of paper would not be worth that much more than the year that preceded it.

And most of the material is gone within months anyway. You cannot be paid for knowledge you no longer have.

Here is the part that should bother you.

A signal only works in the room that issues it. Change country, change industry, walk into a room that has never heard of your school, and the thing that opened every door on one side of the line does nothing on the other.

You did not get worse. The currency just isn't accepted here.

Credit: Bryan Caplan, C-SPAN.

I wrote about which of your four assets survive a change of room and which do not:

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

Rome had concrete that set underwater. Aqueducts. Central heating.

Then it didn't.

Not because anyone destroyed the knowledge. Because the people who had it stopped practising, and nobody noticed until it was gone. It took a thousand years to get back.

Jonathan Blow gave a talk about this in 2019 and made an uncomfortable claim: it is happening again, right now, in software, and almost nobody working in the industry can see it.

His argument is that technological knowledge is not automatically preserved. It lives in people who use it. Stop using it for one generation and it is not archived somewhere, it is simply gone.

Here is the part that should bother you.

The same thing happens to a person, on a shorter clock. Whatever you can do faster and cleaner than everyone around you right now, you built by repetition. And it is worth exactly as long as the tool it lives inside.

Not because you got worse. Because the ground moved.

Credit: Jonathan Blow, DevGAMM Moscow, 2019.

I wrote about which of your skills survive that and which don't. Three of the four are already expiring:

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

An anesthesiologist and a radiologist go through the same training, same exams, same hospital.

One of them will develop real intuition. The other one won't. And it has nothing to do with how smart they are.

Daniel Kahneman explained why at Google in 2011.

The anesthesiologist has monitors. Every time something starts going wrong, the numbers tell him within seconds. Thousands of small corrections, each one answered immediately.

The radiologist reads a scan and finds out whether he was right months later, if he ever finds out at all.

Same intelligence. Same years on the job. One of them is building judgment and the other is building confidence, and from the outside those two look identical.

Kahneman's line for it: confidence is a feeling, not a judgment.

Which one is your work?

Credit: Daniel Kahneman, Talks at Google, 2011.

I wrote about what this means for the four types of skill you own, and why three of them are already expiring:

[mihas1o @mihas1o]

https://t.co/XJakQuFPLX

[mihas1o @mihas1o]

@xsmryze Convenient thread to post right before selling a course on how to grow your money