my002

I really hope we don't get a slew of WebGL-intensive sites.

The Internet doesn't make people dumber any more than video games, rock music, picaresque novels, the printing press, or writing did. These kinds of arguments come around with every new technology and have always been fatuous.

"Show HN: I used [name of LLM tool] to make a leetcode study webapp" seems like a reasonable title here.

Isn't that what (Microsoft, not Github) Copilot was supposed to be?

Looks like very typical Claude Design output to me, though I could be wrong.

For what it's worth I don't have anything against the design. It's kind of like seeing Bootstrap websites everywhere when that was at its peak.

P.S. The hamburger menu icon overlaps with the site icon when the menu is open.

It doesn't read as AI generated text to me. Pangram also suggests it's human-written, for what it's worth. That's not to say that it's correct, just human-written. If the model used in the HF hack did indeed have all of the safeguards manually removed, that would change my perception of the situation, at least.

That's assuming that human-level AI is a possibility with current approaches.

Neat! Looks like an updated version of OCR Arena[0], which doesn't seem to have kept up with new models for a bit.

Would be nice to see the inclusion of Mistral OCR 4, Baidu's Unlimited-OCR, and OlmOCR2.

[0]OCR Arena: https://www.ocrarena.ai/

There are lots of prior examples of automating intellectual work, though -- calculators (rather than doing the math yourself), Google search (rather than looking through library catalogues yourself), Word processors/typewriters (rather than writing by hand), heck, even writing (rather than simply remembering things), which Plato railed against.

AI is not an "oracle" no matter how much Altman and Amodei claim it is.

I mean, surely this depends on what games you want to play. If you're playing mostly indies and retro games, an older desktop will be fine. If you want to play new AAA releases, probably much less so.

The era of subsidised inference is truly ending. The new model multipliers (https://docs.github.com/en/copilot/reference/copilot-billing...) seem like a huge leap, though. From 1x to 6x for new-ish GPT and Sonnet models. 27x for Opus...

Seems like folks would be better off with OpenRouter instead.

Interesting. I've always been turned off by how vague the descriptions of Ollama's limits are for their paid tiers. What sort of work have you been doing with it?