Hacker News: Art9681

New comment by Art9681 in "OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?"

Art9681 — Mon, 26 Oct 2026 00:53:48 +0000

They are selling inference at severely gimped capability. Even in openrouter the quality and speed of the model being served varies by provider. When you dig into the details (which most people don't do), not all providers are equal.

New comment by Art9681 in "OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?"

Art9681 — Mon, 26 Oct 2026 00:52:16 +0000

DeepSeek hasn't proven anything except what they say. You're inherently trusting a company's statements over others. When an independent third party corroborates the numbers then we know for sure. Until then, every company is just selling a story for investors and users.

New comment by Art9681 in "OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?"

Art9681 — Mon, 26 Oct 2026 00:50:25 +0000

OpenAI and Anthropic have more compute and the best frontier models to throw at the optimization problems. They have published articles on these efforts. For example, Dots, and Anthropic's upcoming "persistent agent" is because of these optimizations. Both companies often publish articles detailing a lot of their technical efforts. They just get buried among the flood of other news.

New comment by Art9681 in "Most Americans Say AI Is Bad for Society, Oppose More Data Centers"

Art9681 — Fri, 23 Oct 2026 00:47:41 +0000

I get these surveys in the mail and promptly toss them in the garbage. The people willing to spend their precious time on them are probably the ones whose opinions matter the least.

I encourage you to read "About the Study".

"Ipsos conducted sampling, interviewing and tabulation for the survey using the KnowledgePanel, a nationwide survey panel designed to be representative of the U.S. adult population. KnowledgePanel members are recruited through probability sampling methods using address-based sampling. Panel members who do not have internet access are provided with a tablet and internet service. KnowledgePanel members receive a per survey incentive, usually the equivalent of $1 (though for some it is $2) in points that can be redeemed for cash or prizes."

New comment by Art9681 in "How to engineer an AI harness for game development"

Art9681 — Thu, 15 Oct 2026 03:01:54 +0000

LLMs will absolutely take over game development. That does not mean the result will be a fun game that has value. Just like a group of expert humans can put together a technically and artistically impressive game but it has little appeal.

Taste remains king. Despite that. LLMs will absolutely be able to create elaborate games by this time next year.

New comment by Art9681 in "GLM-Legal"

Art9681 — Tue, 13 Oct 2026 14:55:09 +0000

As always, China has to wait for the Americans to release something, then distill it.

It is no surprise ZAI announced this a few weeks after Astra for Law:

https://openai.com/solutions/industries/law/

New comment by Art9681 in "Sonnet 5.5"

Art9681 — Mon, 28 Sep 2026 22:30:04 +0000

You have to actually spend some effort reading the article they published to answer your own question.

New comment by Art9681 in "Sonnet 5.5"

Art9681 — Mon, 28 Sep 2026 22:27:33 +0000

The usage numbers tell a different story. Not only is the great majority of AI users on American frontier providers, they are also willing to pay the premium price for the premium product. It's not like tech is oblivious to the Chinese models. It's that the industry is aware they are always 6 months to a year behind.

You didn't discover some new trick for cost performance. And the rest of the world isn't dumb.

You're just too broke to afford the supercar and justifying the hooptie. It gets you to the kindergarten class after all. And that's all you need.

New comment by Art9681 in "Coding is not solved"

Art9681 — Mon, 28 Sep 2026 19:53:08 +0000

I'm a Principal with 26YOE. It's solved today. Not tomorrow. The industry isn't changing. It already happened. It happened so fast a lot of people not paying attention blinked and missed it. The sudden realization gave way to fear. And many a blog was written lamenting the loss of a bygone era.

It's highly entertaining I think.

New comment by Art9681 in "Tells of a Slop UI"

Art9681 — Sun, 27 Sep 2026 14:59:20 +0000

Can we get some examples of what a good UI looks like then? I would like to see a UI design expert take a couple of these slop designs and show what each should look like according to their taste.

New comment by Art9681 in "Tell HN: OpenAI $500 ProMax plan listed in API"

Art9681 — Sat, 26 Sep 2026 02:38:10 +0000

There is always a limit unless you start scaling horizontally in the sense that you have multiple accounts via subscription or API. There is always some limit to any single account, even via API.

I'm thinking the persistent agent is something that can't scale to multiple agents. Imagine only one agent you can invoke as many times as you want on a single account, but it cannot invoke subagents or scale itself. It's just always there ready to do work when you need it. But it behaves as "one".

This would be trivial to implement. Of course, if you have the money, you can have multiple $500 subscriptions and have multiple persistent agents.

New comment by Art9681 in "'We hacked the FBI:' Hackers say they have data on all FBI employees"

Art9681 — Wed, 23 Sep 2026 03:23:08 +0000

Anomaly detection is getting better. A small delta can still trigger it. Wether a human reviewed it is another story. But the post mortem will reveal all the signs in hindsight.

New comment by Art9681 in "We write code by hand"

Art9681 — Mon, 21 Sep 2026 14:57:10 +0000

Yea every PR submitted before LLMs was top tier work. Humans are born with an innate instinct to code and build complex software. It's the great equalizer. There are not outliers in coding. It is with great pleasure we "LGTM" and rubber stamp contributions. Nothing ever went wrong until LLMs ruined it for everyone.

New comment by Art9681 in "Claude Code now reads AGENTS.md if there is no Claude.md"

Art9681 — Fri, 18 Sep 2026 22:43:54 +0000

Great. Now if they could support the `~/.agents/skills` path next like everyone else does that would be even better.

New comment by Art9681 in "How GLM built its own inference infrastructure"

Art9681 — Thu, 17 Sep 2026 16:01:45 +0000

They are making these optimizations. They publish these reports too if you care to look.

New comment by Art9681 in "AWS says it can't restore some data from mideast facilities struck by Iran"

Art9681 — Wed, 16 Sep 2026 21:31:33 +0000

No disaster recovery plan? No offsite backups? Someone failed to applied the most basic principles that have existed for decades.

New comment by Art9681 in "DeepSeek v4.1 Flash Is Now Our Best Hacking Model"

Art9681 — Wed, 16 Sep 2026 16:46:59 +0000

Good luck building something of value with "day to day coding". If you're going to compete in the future economy, you're going to compete at the frontier. The low hanging fruit will be of no value. If I can recreate your SaaS with an open weight model at a fraction of the cost, then your SaaS has no value.

New comment by Art9681 in "DeepSeek v4.1 Flash Is Now Our Best Hacking Model"

Art9681 — Wed, 16 Sep 2026 16:44:55 +0000

As opposed to what? Is enclave.ai signed up for GPT Cyber or Glasswing?

New comment by Art9681 in "We must pace the frontier"

Art9681 — Sat, 12 Sep 2026 19:27:21 +0000

Jobs are going to evolve because of AI. Some jobs will be automated. New jobs will be created. It's like saying we should have never created car factories and automated much of it because we should think of the horse handlers and maintainers.

There will still be a market for "trust". And its customers will never trust the machines. And within that market a lot of new startups will emerge for those who actually get to work and seize opportunities rather than lament a bygone era.

New comment by Art9681 in "OpenAI: "We use ... de-identified data to improve ChatGPT""

Art9681 — Sat, 12 Sep 2026 13:13:37 +0000

Read the fine print. Use the API.