dsabanin

Why is a language model bad at video games? I think the answer is stated in the question itself.

That's why I prefer relying on a custom setup based on a Claude Code. Much less surprises, especially if you don't rush into the models prematurely unless forced.

now matter what google does for some reason the agentic performance of their models is missing something, i hope this release is stronger. we need more competition.

I love it. The RAM is not a problem because Apple knows a dirty secret – apps adapt to the environment both at runtime at design time. Give them less RAM, they'll allocate less and clean caches more sooner.

Counting 38 acquired companies as dead is misleading the reader.

Much love to you, brother.

Thanks for doing this. Looks interesting, I'm going to check it out soon.

I'm convinced that at some point looking like being productive and being productive becomes the same thing.

$5.5 in PA.

New distraction is coming in! All eyes on this now!

Man, they really don't think much of their voter base or ordinary people for that matter.

Turned out they just were the selfish assholes everyone always said they were, with everything they say just being poor attempts at rationalization of their deep lack of morals, including their self-serving primitive religion.

Are we essentially looking at the infrastructure for the first mass prompt injection-based worm? It seems like a perfect storm for a malicious skill to execute a curl | bash and wipe thousands of agent-connected nodes off the grid.

We definitely do. How else are the LLMs that are going to replace managers will learn that? /s

You really picked a wrong target for your ridiculous attacks. 2swap is a mathematical artist, not Khan Academy. There's plenty of boring lectures on youtube already on every topic.

What an annoying comment to read about such an incredible video. Both me and my son enjoyed it a lot. It's an educational art piece with beautiful and insightful visualizations.

Zod's validation errors are awful, the json schema it generates for LLM is ugly and and often confusing, the types structures Zod creates are often unintelligible in the and there's even no good way to pretty print a schema when you're debugging. Things are even worse if you're stuck with zod/v3

The whole thing Trump and his minions are doing is so very reminiscent of what was happening in Russia during its turn to fascism. Similarities go to such an extreme degree across the board that I now feel there has to be at least a consultant from Moscow working with the current admin, if not full blown department at the FSB. Blatant lying, complete disregard of any laws, open rejection of norms and mockery of process, justification of extreme cruelty , obsession with some made up concept of “liberals”, and blaming everything on the predecessors (Putin is still trying to blame all the problems on the collapse of the Soviet Union).

I sincerely hope that people of the United States reject being treated like mindless cattle and choose freedom instead of what appears to be a complete and utter national-fascist tyranny.

Really loving Bun these days. Was really pleasantly surprised using their shell scripting API in typescript - single file, shebang line, straightforward DSL, support for piping data here and there.

Really, really nice and ergonomic. Made a few utility scripts today and enjoyed it much better than bash.

https://bun.com/docs/runtime/shell

I've been getting lots of scam calls lately, especially AI voice generated once, and there was one particularly annoying. Very persistent call about being approved for some loan, no reference to any particulars, all very vague, and I kept ignoring. In my mind I had no doubt it was a scam. Well, long and boring story short, now I have a missed payment on my credit report for my new HVAC system..

I find it completely acceptable to outsource some of my thinking and tasks to LLMs. Not everything is my primary area of focus or expertise, and not all questions are equally important. For example, I could buy any random toaster oven, or I can kick off a deep research to pick a best toaster by my criteria and in 5 minutes I’ll have a much better context for this decision, without having to switch my mind from my primary objectives. I don’t want to think about toasters.

Every. Damn. Time. I've enjoyed reading The War of Art, that talks a lot about this mysterious "resistance".

I think it's pretty clear we're at this stage already.

Hm, that may be controversial in itself, depending on where you stand in the current cosmology.

Very well said. This is such an important point.

I believe that if you truly accept what Hemingway said, that writing is rewriting, you get less attached to the idea of reaching the best design on the first try, and feel better when starting with a suboptimal solution.

Of course this sometimes conflicts with organizational pressures, where that quick and dirty solution may be deemed as enough by some and you won't get to finish with the proper design. For me the trick is to consider first version just an internal stage of work on a feature, not even communicated outwards most of the times, until the appropriate design is reached.

Is there something she is not skeptical of or controversial about?

I’ve been using your app for years, you’ve done a great job with it! Thank you!

I haven't seen console ninja yet, thanks for mentioning it. It’s amazing!

I was skeptical at first, but I think Google’s catch up game with OpenAI has been going pretty well so far. Gemini 2.0 Pro and Flash models are really nice. Deep research feature is done really well. Context window is still the best in the industry. Integration with search, gmail, google office suite, google meet, android, etc.

They finally have good enough models that lets them leverage their existing portfolio of products, their cloud infrastructure and how embedded they are in the modern work life.

Plus, unlike Apple, they are not as restricted in their access to the training data, because of their much less principled privacy stance.

And why wouldn’t he? His name is such a widely loved brand right now.

You will. Cursor is much further along on the journey of building an actually powerful AI coding system. Since they are smaller, they can afford to iterate more quickly and experiment with a much tighter feedback loop.