my manager has been experimenting have AI first right the specs as architecture decision records (ADR), then explain how the would implement them, then slowly actually implementing with lots of breaks, review and approval/feedback. He says it's been far superior to typically agent coding but not perfect.

Any current question to an LLM is just a textual interpretation of the search results though; the use the same source of truth (or lies in many cases)

>> You can always make stuff up to trigger AI hallucinations

Not being able to find an answer to a made up question would be OK, it's ALWAYS finding an answer with complete confidence that is a major problem.

It gave me two answers (one was Borland sidekick) which I then asked "are you sure about that?" waffled and said actually neither of those it's IBM Handshaker to which I said "I don't think so, I think it's another productivity program" and it replied on further review it's not IBM Handshaker, there are no productivity programs that include Connect Four. No wonder CTO like this shit so much, it's the perfect bootlick.

dial-up penetration in the mid-90's was still very thin, and high-speed access limited to universities and the biggest companies. Here's the numbers ChatGPT found for me:

* 1990s: Internet access was rare. By 1995, only 14% of Americans were online.

* 2000: Approximately 43% of U.S. households had internet access .

* 2005: The number increased to 68% .

* 2010: Around 72% of households were connected .

* 2015: The figure rose to 75% .

* 2020: Approximately 93% of U.S. adults used the internet, indicating widespread household access .