Better doesn’t always follow. Nyc has had constant public transit demand but there’s been barely any expansion since public takeover of the private subways.
So your objection is that the Trolley problem is difficult to decide? But that's the point of it. The problem itself doesn't assume anything, it's you who just made assumptions about it and use them to make a decision.
I'm really interested in technical details behind Jev (not this), how it can work so fast and so cheap. It's probably large (must be since the performance is so good) but somehow still fast, so it must include some really non-trivial stuff. The price suggests it may be runnable locally, but who knows.
If it was possible to re-create it as an open-weight, it would be exciting!
It might be conceptually similar to a single-output-token LLM (sort of). LLMs output next-token probabilities. You can ask LLMs to output yes/no, or to output only a color, or only a digit or something like that.
In this case I would imagine that they probably embed your input data into a vector space, and they embed your questions/outputs into another space, and manage to predict probabilities/classes/scores for your outputs very quickly. Embedding the output classes/questions into a vector spaces gives you something you can reuse across runs cheaply, as opposed to an LLM where you can prefill the KV cache but this is an expensive operation in terms of memory.
Classical music concerts database. The interesting part maybe isn't what it is, but how it's done: I'm aiming for a fully automated database, where "crawlers" for the classical music websites are created and validated by AI.
Rough idea: get a list of urls with classical music concerts -> send each one to a coding agent to build a crawler -> validate and run on an infrastructure for "free". Classical music is an ideal use case, because the number of concerts is small enough on an average website that it doensn't require many workarounds, and the websites are usually pretty crawler-friendly.
Great - that's what I was looking for when I lived in London! All that existed was TimeOut mag (on paper and online, but pathetically incomplete, more for tourists).
Interesting. I actually prefer when agents don't commit on my behalf unless I explicitly say so, I even had to add a custom instruction for Claude to stop doing it (Codex never does it). Even if I don't read all the code line-by-line, I at least want to see the changes at glance and commit myself. Git Fork[0] is a great tool for that, by the way.
In general, I don't like when I have to prompt models to NOT do something. It's probably difficult for the AI companies to get this right, they should understand ambiguity but still not over-do simple instructions.
In some countries, it is illegal to make unsolicited calls to a number that is on an official "do not call" list.
Any number on that list is effectively a honeypot. I guess you could even distribute an app that automates the whole process: put your number on the list, log incoming calls, aggregate the data and report them in batches large enough to catch the government's attention.
reply