It was a different time when software was seen as something that was built together and everyone was interested in learning the best from one another.

No, it was really not. His tale is from mid-2000s, not from mid-1980s.

In mid 2000s these companies were already operating in the billions and their engineers were already well compensated, and it was known.

Hell, "Cracking the Coding Interview" came out in 2008. Getting a job at those companies at the time was already something coveted because of how well they paid.

> In mid 2000s these companies were already operating in the billions and their engineers were already well compensated, and it was known.

Perhaps in the USA, but in many other countries this does for sure not hold.

Really fun read. To be this seems awful close to my experience using these models to code. When the prompts are simple and direct to follow the models do really good. Once the context overflows and you repopulate it, they start to hallucinate and it becomes very hard to bring them back from that.

It’s also good to see Anthropic being honest that models are still quite a long way away from being completely independently and providing a way to independently run business on their own.

It's likely that the weaknesses have a shared foundation: LLM pre-training fails to teach those LLMs to be good at agentic behavior, creating a lasting deficiency.

No known way to fully solve that as of yet, but, as always, we can mitigate with better training. Modern RLVR-trained LLMs are already much better at tasks like this than they were a year ago.