DeepSeek competes with Sonnet, not significantly worse or better. It tends to do weird things in codebases on the bigger side.

At $3/$15, Sonnet is more than an order of magnitude more expensive than DeepSeek at $0.435/$0.87 (with cached input pricing of $0.003625, DeepSeek is very good at caching, so it's very cheap to use). So, if they're equal in performance, DeepSeek is ten times better value.

But, from what I can tell DeepSeek is better than Sonnet, though I agree it is not at the level of current Opus or GPT 5.5 (but I think it probably beats Gemini Pro 3.1). I use the best model I can for code, because the cost of weaker performance is more than the $100/month I pay for Claude Opus, but it's worth knowing there are very cheap, very good, models for stuff I want to do that isn't Claude Code.

I think there are so many variables from harnesses to tasks, making it very hard to put the models to a pecking order unless one beats another in virtually every task (like in Opus vs DeepSeek).

But all in all, I don't think we disagree.

I always thought that stuffing too much into an LLM context window was a lot like overloading a burrito.Keep cramming stuff in and eventually the tortilla gives out, and everything you added since quietly spills out the bottom.

Anyway, this agent probably has the structural integrity of a fat burito held from one corner :)

The finite-memory nondeterminism monad is like a leaky burrito.

Comments in Github were usually horrible, but the AI stuff brought extra divisiveness. yt-dlp stops supporting bun because they call the rust rewrite a risk -> hate comments. rsync fixes security issues and gets some help from AI -> someone finds a bug and... hate comments. Poor maintainers.

Crazy.

> If I were a user, knowing that the maintainers just let Claude lose on rsync would be bad for my stress levels.

I think you are being too entitled.

I've been tasking LLMs to write a traditional AI for a full vibe-coded RTS. I remove the human players and let them battle. I don't know why but I enjoy watching AI players battle so much :)

In the repo, I even have a tournament script that calculates ELOs. So far, codex was unmatched. I'll try with Opus 4.8 too.

https://egeozcan.github.io/unnamed_rts/game/

https://github.com/egeozcan/unnamed_rts/blob/main/src/script...

I'm happy to report that this game is very fun for natural intelligence entities too. :)

Glad that you liked it! Please fork or note the version you like because I keep breaking it in spectacular ways :)

This is fun! I look forward to trying this out. Thanks for sharing!

Yet another time I've found something beautiful, only to discover that almost everyone else hates it.

Maybe there's a reason why I'm not a designer.

> If you really do need the first element, you can use array_first

That probably needs an array_second too, doesn't it? Maybe array_second_from_last as well?

That's foolish. It would clearly be named `second_array_element`.

Just use array_values() and suddenly you can use int indicies again.

I think writing a static site generator was the first moment I felt like I may be serious about this programming thing.

Those losers who still need Perl on their servers better be ready for a mind explosion

...thought, me back in (too lazy to look up which year it was). I probably published like two things with it, spent (what felt like) a million person hours on it, just to abandon it and use Textpattern.

I always thought the DL as a single row of a table.

In Germany, it's still not uncommon to hear Yava and YavaScript.

I understand that Jawohl is still quite popular there, but JawohlScript adoption is sadly lagging.

It's the only pronunciation you'll hear in Iceland