Alex-Programs

XAI is now printing revenue thanks to the Anthropic compute deal. Moreso than SpaceX itself.

Because failures compound. My productivity has substantially improved since I switched from open models to a Codex subscription, because it doesn't need hand holding, and it doesn't pull stupid tricks occasionally.

I'm curious, what's the agent like?

If I were to build something for Gemini models I'd plan around ingesting a bunch of context then oneshotting it.

The state of the art is moving so rapidly that, yeah, Copilot by Microsoft using gpt-5-mini:low is not going to be very good. And there are many places where AI has been implemented poorly, generally by people who have the distribution to force it upon many people. There are also plenty of people who use vibe coding tools and produce utterly atrocious codebases. That doesn't preclude the existence of effective AI tools, and people who are good at using them.

This is motivated pessimism. We knew in the 50s that breaking the speed of light was highly unlikely. We dreamed of the stars anyway. Now we refuse to dream, or to even attempt to solve the problems (a common pattern when discussing spaceflight is people who are blatantly searching for problems, rather than solutions), because we are pessimistic, devoid of imagination, and seek to legitimise our collective depression through scientific and engineering arguments.

Oh, hi chatgpt

It's also moaning about me coming from a datacentre IP (proxy) with some vague complaints about load introduced by AI crawlers. I think this guy treats "protecting" his site as a hobby.

Heat transfer is based on temperature difference, not heat difference. This way ought to reduce heat losses and improve efficiency.

Is there a reason we would need to coordinate on what to exchange rather than, say, beginning with encyclopedias and textbooks then moving to a constant stream of notable papers, news, discoveries, etc? What kind of bandwidth can you hit with a cooperating neighbour where improvements become civilisationally important? How many bytes (megabytes? Terabytes?) of meaningful new data does humanity produce per second? I suspect it's reasonably low.

I never quite got what was so "hot" about it. There seems to be an entire parallel ecosystem of corporates that are just begging to turn AI into PowerPoint slides so that they can mould it into a shape that's familiar.

There are plenty of competent 16 year olds.

I think using total parameters is fair, it correlates well with the RAM prerequisites to run it. Otherwise Kimi K2 would be "small" despite being a trillion parameters!

Incredible. Thanks for sharing.

People insist upon Codex, but it takes ages and has an absolutely hideous lack of taste.

That's just a really, really ineffective rocket. A spring has nowhere near the energy density of chemical fuel.

Clearly the consumer should automatically trade futures as a hedge!

Personally I prefer Gemini because I still use AI via chat windows, and it can do a good ~90k tokens before it starts getting stupid. I'm yet to find an agent that's actually useful, and doesn't constantly fuck up everywhere while burning money.

The Bristol young lib dems oppose it, but the parliamentary party doesn't think it goes far enough. The Bristol lot are great, I talked to them about it, but they're unlikely to change things on the national level.

Language Transfer is great. On the topic of immersion, I made https://nuenki.app in my gap year. It estimates the difficulty of sentences in webpages and translates the ones at your knowledge level into the language you're learning.

I just started uni, so mostly that. I've found myself making a little CLI for the timetable website and using a software defined radio so I can hear the lecturer while still having noise cancelling.

This is awesome. Good luck with it!

Is that because they're young, or because they're inexperienced?

Thanks for sharing this. I researched this for my A level project a few years ago, and this is a really neat cross reference. I didn't mention V2Ray as much.

Yeah, I'd love something where you pronounce a word and it critiques your pronunciation in detail. Maybe it could give you little exercises for each sound, critiquing it, guiding you to doing it well.

If I were any good at ML I'd make it myself.

This is cool. I'm looking forward to trying it - I wonder what it'll be useful for.

Google products are such a pain to work with from an API perspective that I actively avoid them where possible.

I'm curious whether you can get https://platform.nuenki.app (my deep translation service; I'm just polishing it before release) to run into "too sexual". It's designed to be resilient to that kind of thing, and it works well in my testing, but maybe you have some better "stress test" content than me!

I can't deal with any of the in editor tools. I'd love something that handled inputting changes (with manual review!) while still giving me 100% control over the context and actually doing as its told.

If you're after speed, Groq is excellent. They've recently added Kimi K2.

Yeah, proper V3/R1/K2/Qwen 235B are the point at which open LLMs become worth using.