How to lose all your dev team in one-shot

No one will read this, so it doesn't really matter. But this is how I feel...

In 1997 Netscape was the number one browser in the market. Then its engineers decided to rewrite the whole thing, from scratch. The rewrite took three years. For three years nothing was shipped from that team, and by the time the new browser existed, the market already belonged to someone else (Internet Explorer). Joel Spolsky called it "the single worst strategic mistake that any software company can make."

This is a story about a company that spent two years writing and rewriting their product until they lost all their developers...

"only one person's work was ever real"

There was a founder who, every morning at standup, told his developers that the module he wrote was the best thing in the world.

The developer who actually wrote that module, a week earlier, was sitting in that call. He had spent months on it. It did the exact thing the founder was bragging about but, because it wasn't written by the founder himself, the module was deleted and re-written as "the best thing in the whole world" - by the founder. The developer said nothing. Everyone on that call had learned that correcting the founder cost more than simply nodding.

This is just one example of how the team operated for two entire years.

The story always starts like this, with a man who was right once! He built the first version of the product alone, in a language he liked, and it worked well enough to make money. Now, he needed to modernize the product, so he hired 6 developers to do the job.

But the original product is his "baby-project" and it's "too hard", with "too many hours of thought put into it". So, no one could ever how the product actually works, except for the founder.

Let's create a new product and leave the old one alone...

The initial idea was to "don't touch if it's working"... And that's how the team of 6 devs went on this journey of writing a beautiful new software, with incredible new ideas that could revolutionize the industry... ...if it ever shipped...

But then, six months in, the founder decides that he didn't like the tech his team was using. Instead of creating tasks, splitting the work, and implementing new tech as a team, the founder decides to rewrite the whole codebase overnight, by himself, breaking everything that was tested and working.

And now his team has to work on his new codebase, and re-implement all the modules they already had up-and-running previously. But then, the founder decides that he doesn't like the modules and, once again, instead of discussing with his team, he decides to rewrite each module over-night.

He hired six developers for a job that never needed more than one, and since there was "nothing to give them", he gave them slices of his own and watched over every slice. Hire people you don't need, then make sure only one person's work is ever real. Run that prompt for two years and the output is six résumés updated in the same week!

That's called "AI pychosis"...

"The best thing in the world" is what a person says when he has stopped receiving information from outside his own head! After months of "yes sir" AI models only agreeing with whatever the founder wanted to do, he was completely entangled in "AI psychosis".

His development team had learned that disagreeing with him was completely waste of time. He would propose, six people would say some version of yes, and the yes went into his head as evidence - just like his AI agents. Two years of that and a man sincerely believes he is the next Steve Jobs and that millions of strangers are about to subscribe to his AI app, in the same two years that not one line of that app reached a single customer.

"four full rewrites in two years"...

What shipped to customers in the end was the same, old product again. Netscape threw away one rewritten application and it cost them the decade. This company threw away four, each rewrite with total conviction that it was the "best version of itself" and abandoned the week after, when a newer technology felt more exciting. Then, in the last month any of his team was there, the founder gave up and decided to patch the legacy PHP system he had been trying to replace since before half the team was hired. Two years of six people's lives, and the roadmap ended on a patch on software that was already dead!

"a college project with a payroll"...

None of this was a talent problem. The company was a college project with a payroll: good concept, endless refactoring, no ship date, and nobody in the building whose job was to say deliver this, by Friday. The founder was also the lead developer, which meant the person setting the priorities was the person building them, which meant there was no one managing the process. Whatever he decided in the shower that morning was the roadmap for the day, and six people built it. CB Insights studied 431 startups that died since 2023 and found 43% failed on product-market fit, but you cannot fail to find product-market fit if you never let the market see the product. This company did not fail the market test. It refused to try it, for two years, on purpose.

"Nobody will ever use it"...

The saddest thing in that codebase was not the wasted work. It was the good work. One dev, left alone for a few months, built a tool that was easily the best video production application supported by AI. A whole application, ready to ship and to take over the market, and that nobody will ever use. By the time it worked, the founder had lost interest in the problem it solved, and he was busy at standup explaining that the module he had taken off someone overnight was the best thing in the world.

In the end, each of the six developers decided by themselves, no group chat, no pact, that it was time. Each of them handed in their notice believing they were the only ones leaving. They found out afterward, one resignation at a time, that they had all reached the same conclusion in the same month by complete accident. All six of them, without any coordination, arrived at the same answer.

The founder is still there, patching the corpse, and he will tell anyone who asks that he built the whole thing himself.

He is finally right about that.

References:

1. Things You Should Never Do, Part I — "the single worst strategic mistake" (Joel on Software, April 6 2000)
2. AI-induced psychosis — term coined by Søren Dinesen Østergaard, Nov 2023 (Wikipedia)
3. AI Psychosis: What Clinicians Are Seeing in Heavy Chatbot Users — sycophancy mechanism (Simply Psychology, June 2026)
4. Micromanagement statistics — 79% micromanaged, 85% morale damage, 69% considered quitting (Speakwise, March 2026)
5. The Manager Factor — 70% of the variance in team engagement (Gallup)
6. Second-system effect and "plan to throw one away" — Brooks, The Mythical Man-Month (Wikipedia)
7. The Top 12 Reasons Startups Fail — 431 startups, 43% no product-market fit (CB Insights, March 2026)