alansaber

It's pretty funny but then again, why not if it's as trivial to simplify as it appears

Claude code as a harness makes all kinds of weird design decisions, you are probably better off long term with opencode.

Ah yes the TUI is the place to ship mouse cursor based features for sure

Not that I disagree but I can imagine it's easy enough to ship implicit token increase (ratchet up the COT settings) on then backend

Well, just enough evil to increase profit margins.

Neat, open source harness is definitely a step in the right direction.

I never thought i'd see the day they released a model, rather than a blog post. The Figure 3 demo being a screencap of chrome in localhost made me feel better about myself. Jokes aside, best western open weights model- very cool.

Startups historically are not the most security oriented

Reads better than > We write strong blogs (i like their blogs)

Ah, so this is why oldschool runescape is still popular

You can gamify anything. The concept of VC is sound.

This sounds most logical to me.

Make sure you fingerprint every user to make sure they're using the right index value /s

Those dropbox guys might have been onto something

Google is really not distinguishing itself. Even the hosted inference sucks.

"remarkably unremarkable harness" is why I like it so much.

Interesting. New models are estimated at ~5T params, so 45,000x increase over BERT base (110m). But vocab size of 200k, so only an increase of 7x over BERT base (30k).

Pretty sure that has been done. I clicked to see DOOM implemented using OOXML.

Race to the bottom.

Such is the nature of tool use

No, you can definitely configure low cost search and apply subagents. CC and Codex do not. Not sure if this is to improve the reliability of their subagents, or just a play to increase user consumption.

Nothing about the time taken to complete the task? Users are definitely sensitive to time, not only token consumption.

I always enjoy these "domain expert has fun using AI to do something in their domain" articles. But it's always a hobby project, never something serious.

Yes, because AI gets the "shape" of something right. If you don't know the field you don't notice the pockmarked surface.

I'd advise you to put a video demo at the top of your repo, similar to OP, to convince or interest me.

A lot of people want a use case. One I think might be cool is some kind of spatial/represented comparison: let's see how two different models interact with the codebase (for the same problem), what they touched, and what they did. Or the same model, but averaged across 100 runs, so we can see how much variance there really is per task. Something along those lines sounds interesting to me.

Oh I agree it's an aberration but it seems unavoidable

Battersea power station in London has preserved chunks of the old control room. Sprawling mechanical complexity certainly has appeal, but we all know frutiger aero is the ultimate aesthetic.

AI, you are joshua graham from the fallout games (not TV show). devise small unit tactics. make no mistakes.

clearly benchmark and optimise for a specific model over millions of datapoints > new model comes out > get to do it all over again. At this point just become Cursor and get paid for it.