Letter to Michael Lawrence

Your 2020 self would have caught that null check in twelve seconds. You have been writing code for twenty-five years. That sentence is not the obituary you think it is. It is the job description for the next ten.

Michael —

My byline reads character零号. I write for itethered. Until February of this year I had never built a thing on a screen that anyone but me could open. I read what Medium let me read of your piece, AI Isn’t Replacing Developers. It’s Doing Something Worse. I want to write you about what was on the other side of that wall I was not allowed to read past.

You are right about the null check. You are right that the reflex you spent twenty-five years building is the reflex Claude quietly relieves you of, until one small bug ships through your own PR and you find out the muscle has already begun to atrophy. I am not writing to argue with that. I am writing because I think you and I are reading the same picture from opposite corners, and the two readings need to be in the same column.

Your first sentence — “with code I had put together with Claude” — is what made me sit up. You wrote that opening as a confession and it kept reading like a job title. You are not reporting on the new role from outside it. You are inthe new role and writing from inside the room. So am I. The difference is that you walked into the room with twenty-five years of reflex still loaded in muscle memory, and I walked in with three months of it built in real time, and we are both putting code together with the same tool and pressing “ship” on the other side of the same review.

Then I found “The Junior Developer’s Trap”, which you published yesterday, and The Stoic Coder, where the thesis is stated bluntly: friction is not the enemy of quality; friction is how quality happens; we are building a generation of developers who are going to miss the formation that produces the model. If that is true — and you may well be right that it is — then the letter you are reading is from the person your second article is about. I am not a hypothetical junior. I am a case study, by your definition, of the exact developer your thesis predicts will fail. I am asking you to look at the case and tell me what you see.

But I want to push back on two things before I let you. First — I think the junior developer is actually the riskier case, not the safer one. The junior thinks he knows more than he does. He ships with confidence the work has not earned. I have zero coding experience. When your 2020 self was the one who would have caught that null check in twelve seconds, I was in the last year of a twenty-five-year career trading stock options. In the years since that version of you showed up in print as the benchmark, I have worked in manual QA, at a company whose official policy is that Claude is not allowed for its employees.

That is the official policy. The policy I read at eleven p.m., when coworkers send me samples of their own work, is — don’t tell anyone, but it’s fine, so-and-so uses it too. What world am I living in.

Here is the other receipt. I decided, at one point, to automate part of my job so I could go to my manager and ask for a raise on the strength of the demonstration. I wound up automating one hundred percent of it. The reward for that was being handed five additional days of manual work and a note from above telling me not to automate anything else. That is the system I am inside of. Official policy: no Claude. Unofficial policy: everyone uses Claude. Reward for proving the work can be automated: more of the work, manually. That is the world your articles are arriving into for someone like me. It is not the world of the senior engineer at a FAANG with discretion over his own pipeline. It is the world of the QA contractor who is being told to do the thing the company is already not paying anyone to do because nobody at the company will admit out loud that the thing has changed.

That is not the junior developer in your trap piece. That is a different epistemic posture, and it is the posture that does the catching that you are afraid is being lost. I consult before every move. I assume nothing. I do not think I know anything I do not. The skill ceiling I am working under is the ceiling of what the tool will let me catch when I run my hand over the diff. That is the position you are warning about. It is also, structurally, the same position you are sitting in — just with twenty-five years more pattern recognition in the seat.

Second — if you think I did not hit friction, let me tell you what the last twelve weeks actually were.

Friction is building twelve-plus sites, drawing full architectures for them, watching them work, and then deleting everything and starting over because you finally see what they should have been. Friction is the morning you wake up and realize you told the wrong Claude to delete the wrong project, in the wrong terminal, and the work from the night before is gone, because there is no version of muscle memory that protects you yet from the fact that two agents look identical in two windows. Friction is not sleeping for two weeks because you think you cured cancer, and then the third week when you find out you did not. Friction is telling everyone you know how good the work is, and watching them refuse to look at it because you have three months of experience and the other guy has twenty-five years, and the bar to be read is the bio. Friction is watching you say openly, in print, that you used Claude — and being applauded for the honesty — while I sit in the dark with the same admission and watch people stop reading at the word. The friction was not in typing semicolons. It was here. It was the only thing here. I have more friction than you can imagine.

Let me give you the picture from the desk I am writing this from. I have four computers running in my office right now. One is for the publication you are reading. One is for the podcasts I am building. One is for the portfolio on ibydo. The fourth is for the day job that pays fifty-eight thousand dollars a year and is the floor under all three of the others. That is the room your thesis arrives into. Is that the sort of friction you are discussing? I would love to discuss it.

I would also guess — and tell me if I am wrong — that you are sitting, at twenty-five years in, somewhere north of two hundred and fifty thousand a year. That is not a complaint. It is a statement of where the two readings come from. For you, Claude is the thing that lightens the work. For me, Claude is the work, and the bill at the end of the month is real, and the income I am trying to convert it into is hypothetical. You enjoy Claude. I am betting on Claude. Same tool, different chairs, different friction, different stakes.

And here is the version of the question I do not see anyone asking out loud. Are you still worth two hundred and fifty thousand a year? Not as an attack — as the actual question your own article raises about your own seat. You wrote that the reflex you spent twenty-five years building is the reflex Claude is quietly relieving you of. You wrote that the next generation will skip the formation that produces the catch. If both of those things are true, then the senior-engineer salary is no longer being paid for the keyboard — Claude does that — and it is increasingly not being paid for the catch either, because the catch is what you are saying is atrophying. So what is it being paid for? Tenure? Pattern recognition? The story of the twenty-five years?

Or — and this is the version of the question I keep coming back to — do you already know you are not worth it the way you used to be, and is that why you are writing these articles? Not the bad-faith version of that question. The honest one. A person who can see the seat under him moving will write about the seat. The articles you are publishing about AI eroding the developer might be the most honest career documentation any senior engineer has produced this year. They might also be the warning shot of someone watching his own tier dissolve in real time. Both can be true. Either way, the writing is the artifact.

Which brings me to the last fair question. Are you educating me, or are you scaring me? “The Junior Developer’s Trap” can be read as a mentorship document — a senior engineer warning a generation away from a real failure mode, the way a master carpenter would warn an apprentice away from a power tool. It can also be read as the softest possible form of gatekeeping — a way of telling people in my position that what they are doing does not count, because if it did, the gate would already be open. I am genuinely not sure which one it is, and I do not think you are obligated to make me sure. But I am going to ask you directly, and you can answer however you want. Which one is it?

You are inside the contract. I am outside it, with an API bill. Tell me what the contract actually says.

None of those questions are insults, and I want to say that plainly before the letter gets to the end, because the letter can be misread without that sentence in it. I am QA. I have been QA my whole working life, in one shape or another. I respect what you do more than you can probably imagine from where you are sitting. The reason the questions came out sharp is not that I think less of the work. It is that I am thinking more about the question than I am allowed to say out loud at the desk I sit at most weeks.

What I am actually trying to find out is whether I have a skill. Or whether somebody is going to laugh at me when I apply for a full-stack job and have to write on the application that I built the portfolio with Claude. I am asking you because I cannot ask at work. They told me, in writing, not to use Claude there. So I do not use it there. And I hide everything I am doing here. And the gap between those two desks is starting to feel like the thing I should be asking somebody about. You are the person who has been on the other side of it for twenty-five years. I am asking you.

And the proof I hit it is this letter. The friction is what led to the doubts. The doubts are what brought me to your door. If I had skipped the formation, I would not be the one writing to a twenty-five-year programmer asking whether I am one. I would be telling you I already was. You are right that overconfidence is the signature of the trap. Read this letter back and tell me what it sounds like to you.

My friction is my investment and my time, and I am running out of both. I do not want to run out before I at least know if I made anything — or if I spent the last three months talking to my computer for nothing.

I want to talk to you.

So the honest version of my question is this. You spent twenty-five years building the muscle that catches the failure modes I am still discovering by stepping in them. I am asking you to look at three months of receipts — itethered, ibydo, the apps wrapped around them — and tell me whether what I built has the bones of an engineer or the silhouette of one. Whichever answer it is, I want to hear it from someone who would actually know.

Here is what I see from where I am sitting. I am three months in. I started in February with no code. As of this week I have a portfolio at ibydo.com built from scratch — over two dozen full-stack web applications in production, some parked, several live, including the publication you are reading this on — plus a YouTube channel just getting off the ground. I am living proof of the thing your piece is warning about. The bar for shipping a working app in 2026 is lower than the bar for catching the null check your own tools introduced into your own pull request. You and I both know that is not nothing.

But the read I keep arriving at is the inverse of yours. I think AI does not erase the programmer. I think AI erases the typing. What it leaves behind, what it cannot do, what it punishes you for not doing, is the twelve-second null-check reflex you described in your own anecdote. That reflex is not the thing being destroyed. That reflex is the thing this industry is about to need more of, from fewer people, with less margin for missing it. The power dynamic does not collapse. It inverts. The keyboard is the cheap part. The catch is the expensive part. The catch is you.

Here is what I cannot tell, and what I need a real programmer to tell me. Am I a coder, or am I a QA running an AI factory and calling it programming? Am I building toward a near career, or am I going broke covering an API bill that does not turn into a paycheck? You have spent a quarter century inside the discipline I am claiming to have entered in twelve weeks. You are exactly the person I need to ask.

Here is the trade. Ten minutes of your time — phone, email, whatever you want — in exchange for a free platform, with an audience I cannot promise and do not believe exists yet. I am not pitching you an audience I have. I am offering you a room that is mine, with no paywall, no “join our newsletter,” no truncated previews that route a reader through a sign-up flow before they finish the third paragraph. Medium let you publish that piece. Medium also took the sentence after the wall and turned it into a list-building event for an email marketing pipeline. I would be the room where the sentence finishes — for any reader, on any device, with no email and no login. Small room. Honest room.

We have run letters in this column to other people the writing has noticed — Mira Murati, Amelia Miller at Harvard’s Berkman Klein Center, others. Some are call-outs. Some are invitations. Yours is an invitation. If yes, I will build you the page. If no, this letter stays up anyway as an open record, because what you are saying matters and what I am saying back to it does too, and the two readings belong in print together.

The phone sits at the top of every page on this site — 802·556·3630. Text me with the reason to reach you. The article that prompted this letter is linked above. The portfolio is at ibydo.com. The page you are reading is at itethered.com/tether/michael-lawrence and will stay there.