Recent submissions

Comment · 7 days ago
48gb 4090 uses stock vbios
Comment · 7 days ago
This is the wrong read. Industries outside of tech are weaker in tech, duh. But, they are currently experimenting with the technology, and it’s also less good at non coding tasks, but getting better.

Also adoption is less pervasive.

But wait a few years and you will see a big payoff.

Comment · 19 days ago
Yes
Comment · 26 days ago
Seems rather sloppy to not validate it wasn’t in their training or RL data.
Comment · 1 month ago
Techno feudalism.
Comment · 1 month ago
6x
Comment · 1 month ago
Mature training pipelines, plus ever expanding RL datasets of increased quality, and mega GPU clusters to finish training in a few weeks. Automated safety and reliability testing.
Comment · 1 month ago
Offline would be a better word.
Comment · 1 month ago
GeForce GPUs don’t support MIG. Only workstation and data center cards do that.
Comment · 1 month ago
If memory serves, there is a way to hack the drivers to unlock MIG for GeForce GPUs.
Comment · 1 month ago
How did you get the reasoning trace? Is it the actual one or the summarized one?
Comment · 1 month ago
And yet we see no prosecution for OpenAI’s felonies on HF.
Comment · 1 month ago
If they wanted to see how their model’s stacked up, they might burn $10k in tokens. They ran 120 billion output tokens costing millions of dollars.

That is what you would do if you wanted to beat someone to the punch.

Comment · 1 month ago
Finding Codex session data in the training set that you tie back to these two researchers is like an hour-long task.
Comment · 1 month ago
I don’t think you understand how brazen big tech companies are in practice.

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton