An engineer has written a thread arguing that the biggest problem in modern software development is not that AI generates all the code — it is that nobody reads it. The tweet, titled “The problem is not the AI code, but nobody knows anything anymore,” presents a direct challenge to the assumption that AI-generated code is the future of engineering.
The engineer describes a workplace where entire teams rely on Claude, an AI language model, to write code, tests, specifications, and reports. Nobody on the team likes it. Nobody is resolving bugs. Nobody is thinking anymore. The thread argues that this reliance on AI is making engineering worse, not better.
The Engineer’s Case Against Claude-Driven Teams
The engineer opens with a description of their new role at a big company. It has been half a month since they started, and the state of engineering is terrible. Nobody knows anything about the system architecture or the intent behind the code.
They describe a cycle where middle management pushes AI hard, and teams are forced to ship as much as they can. Higher management says pushing code is not a bottleneck, so why are we slow? People work 12 to 13 hours a day just to press enter. Nobody is reading anything. Humans in corporate are doing nothing on their own.
Nobody on my team likes this. They are being forced to ship as much as they can. I have heard multiple times from higher management that pushing code is not a bottleneck, so why are we slow?
The engineer notes that everyone, from L1 to L7 engineers, is doing the same thing. Talk to Claude. There is no sense of victory. Everything is done by LLMs.
Data Engineering and the Knowledge Gap
Not everyone agrees that AI is making things obsolete everywhere. Hoyt Emerson, a data professional, argued that data people have always had to know everything about the product and business from day one.
Data people are different. We’ve had to know everything (or a lot, or involve domain experts) to figure it out, Hoyt wrote. But AI makes this obsolete, or seemingly obsolete.
The engineer responds that this knowledge is missing for people starting today. If you prompt away in a new field, that foundational understanding is gone. AI removes friction for data engineers, but it does not replace the knowledge that came before.
The Product Manager’s Dilemma
Sean Behan, a product manager, made a related point about what AI means for his role. He has always admired product people who cannot code but can manage a team to get the software they want.
“I’ve always admired product people who can’t code but can manage a team to get the software they want,” Sean wrote. “Knowing what you want has always been the hardest part.”
With AI, a good product manager could now build anything they want and find a market, make it look good, and do it quickly. But the foundation suffers.
“If you can’t code, you will essentially build a very bad foundation for a product that’s very hard to maintain,” Sean warned. “Although AI is getting better at that too, especially when you iterate often, but still, if you choose the wrong language or the wrong mental model, you have the wrong start from the get-go.”
It still helps to know the fundamentals, he concluded: for programming and designing a product, and for a good PM who knows what is needed but also understands system and architecture design.
Maintenance Is the Real Boss
The engineer argues that writing code by hand might be dead, but it helps. Having taste with AI is more important than ever. But the final boss is always maintenance.
The easier it is to generate a quick pipeline, app, or BI dashboard, the more you have to maintain. And if nobody knows a thing, that can get really hard.
Why Humans Are Still Needed
The thread closes on a point about direction and orchestration. The AI cannot prompt itself, so humans are still needed to direct and orchestrate the process.
Why do we even need humans? To me, it’s a clear sign that humans are still needed to direct and orchestrate it.
Intent, taste, design, and architecture are killer features in today’s world. But once these are absent, or worse, fundamental, get lost, it is really dangerous.
Self-Inflicted Problems
The engineer reads that this is a self-inflicted problem. If companies still hired juniors, then the problem wouldn’t be happening. But hiring juniors is not as easy as it sounds.
The thread references Kris Jenkins talking about the worry of middle management, not the vibe coding at Danger of AI or LLMs. It also points to the idea that if consumed by a human, should be written by human.
The Hard Numbers
- The engineer has been at the company for half a month
- Team members work 12 to 13 hours a day
- The engineer describes the state of engineering as “horrible”
- The tweet title is “The problem is not the AI code, but nobody knows anything anymore”
The Schedule Table
| Stage | Detail |
|---|---|
| New role | Started half a month ago |
| Daily hours | 12 to 13 |
| Management push | Middle management pushing AI hard |
| Team reaction | Nobody likes it |
| Knowledge gap | Nobody knows anything about system architecture |
| Foundation warning | Wrong language or wrong mental model |
| Final verdict | Maintenance is the boss |
Our View of the Thread
The engineer’s argument is a warning shot across the bow of a workplace trend. They are not arguing that AI is useless. They are arguing that AI is being used badly — that teams are abandoning reading, design, and intent in favor of letting Claude handle everything.
The result is undocumented systems, unmanageable codebases, and teams that cannot resolve bugs because nobody remembers why the code exists. That is a real problem, and it is getting worse.
The thread’s strength is its specificity. The engineer describes a particular company, a particular team, a particular workflow, and a particular failure mode. The failure is not that AI is broken. The failure is that humans stopped doing their jobs.
The engineer’s solution is not to ban AI. It is to stop forcing teams to ship without understanding what they are shipping. Read the code. Understand the architecture. Document the decisions. These are not new demands. They are the demands of professional engineering.
The thread is a reminder that tools are not enough. Tools amplify what is already there. If what is already there is ignorance, the tool amplifies the ignorance. If what is already there is curiosity and discipline, the tool amplifies that too.
The thread is a call to remember what software engineering is supposed to be. It is supposed to be a craft where people understand what they are building. It is supposed to be a profession where people care about the quality of the code they produce. It is supposed to be a practice where people document their work so that others can understand it later.
When those things disappear, the codebase becomes a liability. When nobody knows anything anymore, the system is brittle. When the team stops thinking, the product stops improving.
The engineer’s tweet is a warning. It is also a plea. The problem is not the AI code. The problem is that nobody knows anything anymore. Fix that, and the code will follow. Ignore it, and the code will rot.
The thread ends with a warning about maintenance. The easier it is to generate a quick pipeline, app, or BI dashboard, the more you have to maintain. And if nobody knows a thing, that can get really hard.
That is the final boss. It is always the final boss. It is the boss that the AI cannot fight for you.
The thread is a reminder that the future of software is not about faster generation. It is about slower comprehension. It is about the people who sit down, read the code, and ask why. It is about the people who refuse to let Claude do all the work.
The engineer’s tweet is a call to arms. It is a call to stop the madness. It is a call to reclaim the craft. It is a call to remember that the code is not the point. The point is what the code does, and who understands it.
The problem is not the AI code. The problem is that nobody knows anything anymore. Fix that, and the code will follow. Ignore it, and the code will rot.
Source material: “The problem is not the AI code, but nobody knows anything anymore,” ssp.sh.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.