Hacker News: joshuastuden

New comment by joshuastuden in "GitHub Copilot code review will start consuming GitHub Actions minutes"

joshuastuden — Fri, 01 May 2026 22:39:55 +0000

Semianalysis just argues for the industry case. They never go against the grain.

New comment by joshuastuden in "Uber torches 2026 AI budget on Claude Code in four months"

joshuastuden — Fri, 01 May 2026 22:37:09 +0000

The CTO is more creative than the engineer so who needs the engineer.

New comment by joshuastuden in "Uber torches 2026 AI budget on Claude Code in four months"

joshuastuden — Fri, 01 May 2026 19:17:18 +0000

I'm not asking to say "it's not because of LLMs" I am asking for evidence that LLMs are creating revenue for users.

New comment by joshuastuden in "Uber torches 2026 AI budget on Claude Code in four months"

joshuastuden — Fri, 01 May 2026 18:38:47 +0000

You're basically arguing for massive headcount reductions.

New comment by joshuastuden in "Uber torches 2026 AI budget on Claude Code in four months"

joshuastuden — Fri, 01 May 2026 18:22:22 +0000

Exactly. At Cerebras I know of several people who burn tokens on completely USELESS tasks (randomly changing pixels in an image) just to keep them high up on the token leaderboard.

I suspect the other tokenboard leaders are doing the same. They made the metric "token usage" (which is just a proxy for LOC) so that's what they're gonna get.

New comment by joshuastuden in "Uber torches 2026 AI budget on Claude Code in four months"

joshuastuden — Fri, 01 May 2026 18:10:22 +0000

How can that be attributed to any code an LLM wrote?

>$8 billion of net income was the result of a tax benefit the company realized in the first quarter of the year.

So exactly how much of their revenue is because of any code LLMs wrote vs. just structural tail winds?

New comment by joshuastuden in "OpenWarp"

joshuastuden — Fri, 01 May 2026 14:49:44 +0000

How do you know they didn't?

New comment by joshuastuden in "GitHub Copilot code review will start consuming GitHub Actions minutes"

joshuastuden — Wed, 29 Apr 2026 03:29:13 +0000

No, because Anthropic can't serve their models unless they train them.

Training is akin to the cost of building the software/product. Inference is selling the product.

New comment by joshuastuden in "GPT‑5.3‑Codex‑Spark"

joshuastuden — Thu, 12 Feb 2026 21:49:41 +0000

Um, no. There's no limit on model size for Cerebras hardware. Where do you come up with this stuff?

New comment by joshuastuden in "OpenAI is unsatisfied with some Nvidia chips and looking for alternatives"

joshuastuden — Tue, 03 Feb 2026 21:31:56 +0000

Says Nvidia tried to buy Cerebras, but Cerebras refused? I wonder if there are other deals beyond the OpenAI one, then.