Hacker News: aoeusnth1

New comment by aoeusnth1 in "Sharing AI progress in mathematics"

aoeusnth1 — Wed, 07 Oct 2026 03:45:35 +0000

Seriously, what's special about oncology that convinces you there will be no progress?

New comment by aoeusnth1 in "High Diesel Prices Bankrupted 16 Trucking Companies in Just 30 Days"

aoeusnth1 — Tue, 06 Oct 2026 06:39:36 +0000

> But be practical: The guys running the machinery that literally makes society possible deserve to be compensated fairly and I don't see many trucking magnates living large at the rest of our expense.

Properly taxing trucks will be mostly passed onto the consumer, not reducing their compensation much, and it would significantly reduce the number of trucks on the road (a good thing). Out of equilibrium, it would cause many of them to go under and compensation would crater... so of course it would be wildly unpopular.

New comment by aoeusnth1 in "SvelteKit 3"

aoeusnth1 — Fri, 02 Oct 2026 01:14:31 +0000

What's the point of using a slow frontend framework just because you're familiar with it, instead of a fast frontend framework which the LLMs can write equally well (or better)?

New comment by aoeusnth1 in "Nissan's third generation e-POWER powertrain"

aoeusnth1 — Thu, 01 Oct 2026 17:03:27 +0000

The reason is fuel efficiency. Why would weight be your primary concern?

New comment by aoeusnth1 in "Nissan's third generation e-POWER powertrain"

aoeusnth1 — Thu, 01 Oct 2026 17:03:04 +0000

Gas is more expensive.

New comment by aoeusnth1 in "Nissan's third generation e-POWER powertrain"

aoeusnth1 — Mon, 28 Sep 2026 15:40:35 +0000

I disagree. Basically anywhere an ICE engine is used for whatever reason, it should be a hybrid instead. There are good enough reasons not to use EVs but there are zero practical reasons to not use a hybrid.

New comment by aoeusnth1 in "Contrastive Language Models"

aoeusnth1 — Fri, 25 Sep 2026 04:13:55 +0000

Oh, an LLM forward pass is also like system1. The action embedding just makes it more useful for tasks or classification. If it has a small fixed computation budget it's similar to system1 than system2 (like a thinking budget, or agent harness).

New comment by aoeusnth1 in "Contrastive Language Models"

aoeusnth1 — Thu, 24 Sep 2026 17:48:30 +0000

The difference is the the joint embedding of actions and state which are fined tuned for certain outcomes. The bulk of the weights are just a forward pass on an LLM (Nemotron).

New comment by aoeusnth1 in "Contrastive Language Models"

aoeusnth1 — Thu, 24 Sep 2026 17:46:09 +0000

System 1 is quite informative about what it's for. It means it's a system designed for bounded time quick classification, vs. agentic unbounded or slow thinking with scratchpads and tools.

New comment by aoeusnth1 in "Contrastive Language Models"

aoeusnth1 — Thu, 24 Sep 2026 17:45:06 +0000

Yeah, the github landing page is much less misleading and clearly says this. I don't know why they try to make the marketing page gloss over this detail... it speaks to the mindset of the authors.

New comment by aoeusnth1 in "Claude Opus 5.5"

aoeusnth1 — Tue, 22 Sep 2026 16:46:36 +0000

You were right to call that out, and the evidence makes a stronger case than you are stating.

New comment by aoeusnth1 in "The August 17 outage"

aoeusnth1 — Fri, 21 Aug 2026 18:38:17 +0000

No customer has access to an appreciable fraction of any data center, and if they do they don't have 10x overhead in reserved quota.

New comment by aoeusnth1 in "Beef and dairy drive 41% of biodiversity damage linked to global farmland"

aoeusnth1 — Wed, 12 Aug 2026 14:16:50 +0000

Wynes & Nicholas downgraded electricity sourcing to "questionable merit." Table 1 gives a range of under 100 kg to 2,500 kg CO₂e saved per year, depending on the default CO2 intensity of the state/country you are living in.

A plant-based diet is one of their high-impact lifestyle choices.

New comment by aoeusnth1 in "Can Intel finally beat ARM on performance per Watt?"

aoeusnth1 — Sun, 09 Aug 2026 17:21:25 +0000

It sounds like you agree with the criticism that your title is a bunch of baloney, but somehow think that's not your fault? You wrote the lies in that title, not an algorithm.

New comment by aoeusnth1 in "Discovery Loop"

aoeusnth1 — Wed, 05 Aug 2026 17:00:07 +0000

Sounds like you get your news from 2024 when people thought things were plateauing after GPT4?

New comment by aoeusnth1 in "When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation"

aoeusnth1 — Wed, 05 Aug 2026 16:48:31 +0000

Soccer is nondeterministic. Are you as good as Messi?

New comment by aoeusnth1 in "Google Discloses $94.1B in SpaceX Stock, Marking 6% Stake"

aoeusnth1 — Sun, 26 Jul 2026 15:17:11 +0000

The return on capital is also phenomenal. To be a bear, you have to assume that will crater.

New comment by aoeusnth1 in "Opus 5 is currently #1 on Artificial Analysis Intelligence Leaderboard"

aoeusnth1 — Sat, 25 Jul 2026 02:36:16 +0000

No, because Opus is smarter than those models if they’re all on medium settings. You should compare at similar levels of performance, which would be favorable to Opus.

New comment by aoeusnth1 in "Claude Fable 5: mid-tier results on coding tasks"

aoeusnth1 — Fri, 12 Jun 2026 05:44:11 +0000

If it's memorized your benchmark then your benchmark is bad, it's not cheating

New comment by aoeusnth1 in "AI is slowing down"

aoeusnth1 — Mon, 08 Jun 2026 21:23:33 +0000

What about all his other articles that had f-bombs and the predictive utility of used toilet paper?