The New AI Superpowers: Focus and Followthrough

Good points.

First use of AI code generation only recently - I think the key is to throw away code until you can one-shot the app. You wouldn't employ a dentist and tell them half-way through what they must do next.

That's just my first impression.

Interestingly, I started with Gemini and switched to ChatGPT using the initial creativity of Gemini as the starting point.

So like search results, why go to page 7 if you can just retry/rewrite repeatedly and never move passed page 1 because at one point, you'll find the right page at the top. As they say "too many cooks spoil the broth" so just get the menu selection right and the perfect meal will be delivered, and if it's not, don't send it back, throw it away and reorder! The cook won't care - it's a robot - no feelings.

If you can't one-shot it, or almost one-shot it - then you f##### up, not it.

Given the limits of LLM's regarding context, why is zero dependencies the goal ?

Shouldn't something like maximum reuse/abstraction, minimal written code, should be a better goal ?

Possible leading to more llm-scalable systems ?

But AI is only able to get you through those already easy first 90-99%. The hard and most time-consuming part of getting to 100% is still there, and AI does not help or make us more productive achieving that, in my opinion.

At least in Enterprise IT, where I work, the challenge is not how you scale technology, but how you scale trust. Big IT problems are always trust problems. That messy CRM systems everyone complains about? The messiness is likely due to low organizational trust, departments working in silos, and disagreements on definitions (what is a customer?).

LLMs can scale technology, but they are terrible with scaling trust. You know what scales trust in Enterprise? Good leaders, good culture, good conversations, good meetings, good narratives, good insights, good decisions. Exactly the things that scaled trust 50 years ago, and exactly the things LLMs cannot directly impact.

But now see what is happening. Rather than doing the hard work of gaining trust and consent, employees are using LLMs to build their localized agents. Of course they are happy! But the Enterprise, as a whole, is not better off, the real problems have not been addressed, the consensus has not been reached.

Two things.

1. Deps are an organizational hassle. They need review, get flagged as security and legal risks, and (mostly incorrectly imo) have been somewhat tarnished by the idea that deps are all just "leftpad" - simple things that are over abstracted. This is the "incorrect" argument that I don't really think should carry much weight, but often does win the discussion internally in organizations.

2. Deps can be misaligned with your precise needs, and you often end up "owning" them anyways the second you patch/fork/modify it in any way. This one is legit. It's a real thing that happens and is painful. Sometimes being honest that you should just be responsible for the thing in the first place is the right call. This one I buy as an argument, but it just doesn't apply to that many deps.

But the combination of the two makes larger organizations somewhat adverse to dependencies on average.