But his point is interesting.

I recognize what he's saying (the comment stright below the main tweet is spot on about systems thinking). He's right, I think, but it shows the obstacle for anyone that has evolved as a not-systems-thinker. And I cannot use these bots to help others get better bots at this time, yet, because they come with massive vendor lock-in despite everyone and their stepmom's LLM platform having copied the OpenAI API. One of the most costly things is constant evaluation and prompt tuning. I've also found that arena.ai top lists are biased towards plausibly retarded prompts and do not send traffic to more obscure models often, so you gotta DYOR every time, for your own use-cases, across a wide set, not just the top list. That's not just hard work; building the right framework for judging the job at hand is expert work - it requires you to be a Chad in the field you're working in.

I don't know hardly anyone outside of my Bitcoin bubble who cares.

I think it's because the trend since the 80s has been to rely on regulation for privacy and it's been seeded with political bravado all over Cali, Canada, Europe, Mexico... I even see politicos in the Carib emulating this behavior. But it's virtue signaling. Yes, it's extra income for the state when they catch "bad orgs", but the real problems are caused by the places that they force surveillance upon: govt agencies, banks (especially of the "neo" kind), mobile providers, healthcare... and now social media and AI firms. These guys get hacked all the time. And the chance that your data isn't out there on the darkweb is minimal.

(And thus, when you sound the alarm, the politicians will not tell the public that their little laws don't help preventing criminals stealing your stuff; criminals don't care about breaking the law, that's why they are criminals)

You sound more down on coding than many writers are on writing.

If I get someone that spent a couple of weeks and then messes up their first PR, they're on their way to become a contributor. If I get someone that spent a couple of dollars on API creds from the smartest AI in the world and then still messes up their first PR, there is no way for me to guide them. This is why coding dies. It's a self-fulfilling prophecy, by everyone saying it will die.

I can't speak to coding, but as far as writing goes I'm pretty optimistic.

I last night had a bot solve a real security issue for me in 30 minutes (incl my human time) that would have taken me at least half a day by hand by having a smoke, saying, no to a boiling the ocean idea and giving it a very precise instruction (and rockets for the lulz.) So of course I'm optimistic on this tech, but it's not as good as advertised.

The bot at first did what any jr coder that (in a distant past) would work for me do though: ask me to lower the requirement. Tough luck lil bot, meet my resolve, and pretty please use that grey mass... oh wait.

But, there is a catch. The solution doesn't work in the field because the field is a spaghetti of shit built upon shit. In fact, I boiled both oceans telling it to use its solution because it ended up in an endless loop of npm install && npm audit. Upside is that we can all get our fish soup straight off the boat now. I'll have to replan it.