Hacker News Reader

JacobAsmuthon GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price

I don't know why anyone was saying that when Anthropic clearly knew Opus 5.5 significantly outperformed Astra at the time of Astra's launch. I think it might be a good exercise to go back and find out who called Astra a "gut punch" and lower your credence in their future claims.

JacobAsmuthon California farmers are struggling to sell grapes as demand for wine drops

Corresponding jump in life expectancy and decrease in liver cancer coming in the next decade or so I imagine.

Good to hear. Sad we weren't able to regulate alchohol out of existence but maybe Gen Z can kill a lot of it off.

JacobAsmuthon GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price

6.1 Sol is slower than 6 Sol: https://artificialanalysis.ai/models/releases/gpt-6-1-sol

However, that doesn't say much. You can just run a smaller model at a larger batch size to get higher throughput but lower interactivity.

JacobAsmuthon GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price

The new hardware (TPU v8 and VR) are more expensive but they are significantly cheaper per flop. e.g. many multiples more performance for only 2x the price.

If I have some ML workload to run I can buy $x of Blackwell chips or I can buy significantly less $ worth of Vera Rubin chips to get the same performance. That's the key thing to keep in mind when you're talking about financials.

JacobAsmuthon GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price

Very true, the sharp increase in difficulty (as measured by human passrate plummeting from 1->2 and again from 2->3) gives an even more stark view of AI capabilities over time.

JacobAsmuthon Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-Astra

Good god this criticism is completely groundless. Complete, massive misunderstanding of the task distribution and goals of TB2.1 and TB4

JacobAsmuthon Formalizing Fermat's Last Theorem

Except there's 10 trillion gears

JacobAsmuthon Headlong: A microharness for persistent agents

How can I be more specific with no information?

JacobAsmuthon Gemini 3.8 Flash and 3.8 Flash Cyber

Sure I do. You can just apply for access. What's your use case?

JacobAsmuthon GPT-6 Astra

LOL

JacobAsmuthon GPT-6 Astra

Astra is clearly able to acquire new knowledge in context and apply it. It was the whole thing that his ARC-AGI benchmarks have been measuring. It's a direct refutation of the original comment.

JacobAsmuthon GPT-6 Astra

All those stupid Bell Labs researchers not inventing Uber or Tinder. How come they didn't just build the obviously popular and profitable businesses that became possible once they invented the internet?

JacobAsmuthon Ask HN: Why were OpenAI, Claude, and Grok simultaneously down?

It could also mean that Google can absorb essentially unlimited demand spikes by load shedding.

JacobAsmuthon Gemini 3.8 Flash and 3.8 Flash Cyber

You're telling me for only 5x the cost and 1/10th the speed I can use a Chinese model which performs worse than Gemini 3.8 Cyber? And I get to do all the hosting and setup work myself instead of just using a model and framework which is already integrated with GCP? Dang!

JacobAsmuthon Gemini 3.8 Flash and 3.8 Flash Cyber

I'm confused. This doesn't mean they trained on it.

JacobAsmuthon How accurate have Ed Zitron's AI skeptic predictions been?

Will someone flag this comment please.

JacobAsmuthon GLM-5.3-Flash

They'll be able to buy them without paying NVidia's 80% profit margin

JacobAsmuthon Headlong: A microharness for persistent agents

The Googlers must be vague posting about something internal.

JacobAsmuthon OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

"No evidence" -> Large lab saying that they use distillation when training their models

JacobAsmuthon OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

https://techcrunch.com/2026/04/30/elon-musk-testifies-that-x...

Make sure to stay updated!

JacobAsmuthon Claudette: Make Claude stop talking like a BuzzFeed article

Unlikely. Much more likely is that Opus 5 was trained in an RL environment with subagents, and it learned to talk this way when reporting progress to the invoking agent.

JacobAsmuthon Anthropic IPO filing will show AI backlash as a risk factor, sources say

Not a risk.

JacobAsmuthon Vomit: Clean up Claude 5's token output with a separate LLM

https://www.yahoo.com/news/articles/ai-boss-trump-hates-beca...

JacobAsmuthon Qwen3.8 27B scores 52 on Artificial Analysis

It has double the active params.

JacobAsmuthon Nvidia dramatically reduces amount of OpenAI infra financing it may guarantee

You're suggesting that if a very good and cheap AI model came out tomorrow everyone would rush out to rent Azure instances to run batch size 1 inference on their model?

JacobAsmuthon Nvidia dramatically reduces amount of OpenAI infra financing it may guarantee

"eventually" is actually a function of frontier model capabilities. You only get Qwen6-27B when you have Opus 7 producing extremely high quality tokens for them to train on. So the market for local models is always significantly behind the frontier, by definition.

JacobAsmuthon Nvidia dramatically reduces amount of OpenAI infra financing it may guarantee

Ed Zitron has not yet made a single correct prediction about AI :)

JacobAsmuthon Nvidia dramatically reduces amount of OpenAI infra financing it may guarantee

Yeah but Apple pays Klarna to provide that service. Circular financing!

JacobAsmuthon Nvidia dramatically reduces amount of OpenAI infra financing it may guarantee

Wait until you hear about how airlines work, you'll start babbling about the rewards-points bubble

JacobAsmuthon Red queen hypothesis – A new way forward for self-improving AI

> The research team, which includes collaborators from NVIDIA and Flower Labs, have come up with a new method for recursive self-improving AI agents to continue improving themselves.

What happens if you apply the method to non-recursive self-improving AI agents? Can they continue improving themselves? Or does the recursive self improvement only recursively self improve AI agents which are themselves recursively self-improving?