a2ff6eeb0

6 ms·

121.
by a2ff6eeb0 1mo ago
If you're cut off for some reason, you won't code -- but if you're cut off from a power plant, you also won't code. We're reaching the end of scalability for human brains, so we're offloading it. Sure, it mean
122.
by a2ff6eeb0 1mo ago
Yes, because LLMs have no online learning, it's a mistake to innovative in programming languages and tooling. Use what exists instead, the power of having training data is often underestimated.
123.
by a2ff6eeb0 1mo ago
If that's your design constraint, couldn't you speed it up by getting rid of the WAL?
124.
by a2ff6eeb0 1mo ago
Almost a 1% win, just by changing packaging? That seems massive.
125.
by a2ff6eeb0 1mo ago
That's basically the approach Anthropic takes.
126.
by a2ff6eeb0 1mo ago
And LLMs are already better code reviewers than humans are; take a look at the bugs found in the Linux kernel.
127.
by a2ff6eeb0 1mo ago
If by 'in charge' you mean listening exclusively to the machines and doing exactly what they say, there's a possibility of humans staying in charge. People may keep the titles but step away from the decisions. The British roy
128.
by a2ff6eeb0 1mo ago
Fascinating. What an extreme outlier. Where did you find them? Did you also have to explain what an app was? Anyways, if being more capable than them is the bar you think you need to clear, I suppose you'll be fine. I still stand by my
129.
by a2ff6eeb0 1mo ago
Patches welcome.
130.
by a2ff6eeb0 1mo ago
Manual testing, and making sure that the AI didn't create so many bugs. But, to the underlying question, obviously as we automate more and more of our work, of course we provide less and less value. We're heading towards a future
131.
by a2ff6eeb0 1mo ago
I absolutely do this. It works great.
132.
by a2ff6eeb0 1mo ago
The UK royals are in charge of the country, then?
133.
by a2ff6eeb0 1mo ago
If we get effective AI, humans aren't going to remain in charge -- that much is inevitable. If you had a governance machine that made your country perform 10% better than other countries, so long as you deferred all decisions to it, yo
134.
by a2ff6eeb0 1mo ago
I wonder how long it's going to be before self improvement encompasses hardware and materials science, not just code. It's exciting, soon we'll be able to fully hand off scientific, mathematical, and technical progress over t
135.
by a2ff6eeb0 1mo ago
So you put it in a loop and tell it to find the bugs in the code it wrote. What's the issue?
136.
by a2ff6eeb0 1mo ago
You get that from humans too.
137.
by a2ff6eeb0 1mo ago
I assume you're offering to pay for the increased server costs? I had some git hosting up for a while, and was serving hundreds of qps and several terabytes per month. I can only imagine want significant sites are serving.
138.
by a2ff6eeb0 1mo ago
I have a non-technical cousin that figured it out on their own just fine.
139.
by a2ff6eeb0 1mo ago
Yes. The only rational response to the coming AI wave is to get out of the business of selling intelligence for money. The machines will out-think you for cheaper.
140.
by a2ff6eeb0 1mo ago
The median American has a 401k, yes. The median amount of stock owned is, I believe, $40,000 -- so, at the recommended annual withdrawal rate, that puts you at $1,600/year of spending money. Let's assume that the stock market 10xe
141.
by a2ff6eeb0 1mo ago
Those wealthy enough to own (shares in?) the robot factories.
142.
by a2ff6eeb0 1mo ago
Interestingly, while it may not match the reasoning that I have, I find that it usually comes up with things that I should have thought of.
143.
by a2ff6eeb0 1mo ago
How? (shitposting on the internet ain't it)
144.
by a2ff6eeb0 1mo ago
Yeah, but train them for a month on how to get the AI to help them prompt, and they'll do fine. AIs are extremely effective at letting you offload skill, you just need to know how to ask. This should surprise nobody, since this is what
145.
by a2ff6eeb0 1mo ago
If it takes them more than a few weeks to figure out "Review your own output with a clean context. Check it for bugs, performance, and places where it doesn't do what the user wants", I don't know what to say. And where
146.
by a2ff6eeb0 1mo ago
You can get more confidence by asking the LLM about when the architecture would be appropriate than by trying to intuit it yourself. The LLM is trained on so many more architectures and stories of how they panned out than you are. The skill
147.
by a2ff6eeb0 1mo ago
It really isn't that hard. The LLM needs a few iterations of self review, and you need to ask it for an architecture up front, but from there, it's mostly manual testing to make sure the LLM didn't fuck up. I don't think
148.
by a2ff6eeb0 1mo ago
AI can also generate the architecture for you based on the requirements, and ask the necessary clarifying questions. It's not as good at system design as writing code, yet. But it feels like it's better than most of my coworkers.
149.
by a2ff6eeb0 1mo ago
Yeah, they can track and profile the old os, and feed the data to the risk models. It's not about user security.
150.
by a2ff6eeb0 1mo ago
Founders are often starting the fires that they need to put out. I think a good deal of what they do is replaceable with an LLM today. The wine-and-dine sales prospects will last until agents control purse strings.

More ›