2-year check-in:
claude is somewhat helpful in professional setting. It is good enough to help to crunch mid- sized codebases and suggest implementation.

The biggest weakness of LLM is contextualization in my opinion. It is often general and often lacks specifics about a company/silo/tribe.

How copilot-generated code influence your code reviews in a team setting? Does it feel better, same or worse?

It is post-filtering usually. However, it does not stop there — lookup "ontogpt" on github. They basically use LLM to produce ontology.

I would think there will be many applications at different phases with different goals, so exciting times ahead indeed.

#Unity rug pull is another reason to have a knowledge graph as part of your business strategy. I know it is hard to produce decent KG while managing asset pipelines, developing, writing, and managing thousands of other tasks at the same production cycle... but it would make things so much easier.

It is getting popularized again, thanks to LLMs... and many companies see linked data as a competitive advantage.