The OpenAI insider Tibo has a warning for everyone offloading their thinking to agents.

Ominous

Watch the full interview:

@thsottiaux Yeah, this part bugs me too. Agents are fine for draft work. Judgment gets rusty fast when you stop wrestling with the raw problem yourself.

@thsottiaux I see AI as a tool, something to help a human, but not to replace a human. Sadly some other companies, looking to maximise profit, see AI as a way to replace a human workforce and cut costs.

@thsottiaux

@thsottiaux please refer to him as St. Tibo, giver of tokens resetter of limits

@thsottiaux This is a fallacy, humans have never truly understand any system in all of its layers. We use compression tools (what coding fundamentally is) to traverse spaces we can't directly interface. LLMs are just another layer of abstraction that uses language, stop the cope.

@thsottiaux

@thsottiaux @thsottiaux looks like a monk in this video.

@thsottiaux This is less a warning than a boundary. AI can help us think and execute, but it shouldn’t own the judgment. The human still has to define the problem, check the output, and make the final call.

@thsottiaux Wait I thought agents could just automatically fix everything

@thsottiaux that ship has sailed

@thsottiaux Fuck vibe coding

@thsottiaux C'est super bien dit

@thsottiaux water found in ocean (It is a tool after all)

@thsottiaux What he says is true, but unfortunately the advice rings VERY hollow. They well know that many companies are RUSHING full speed ahead without due consideration, only using various types of bottom line cost savings as their ultimate driver(s) and goal.

@thsottiaux no wonder one of the labs have a tool that everyone likes and the other lab has a tool that still tries to fix flickering after 12 months