Hacker News: chpatrick

New comment by chpatrick in "Terence Tao: Math 2.0 [pdf]"

chpatrick — Sun, 11 Oct 2026 15:15:55 +0000

...his point is we don't know if it is a cure or not, and it would be good if humans had an idea instead of just having to take the AI's word for it.

New comment by chpatrick in "Terence Tao: Math 2.0 [pdf]"

chpatrick — Sun, 11 Oct 2026 14:41:02 +0000

"its model predicts [...] will kill cancer cells"

New comment by chpatrick in "Terence Tao: Math 2.0 [pdf]"

chpatrick — Sun, 11 Oct 2026 14:36:38 +0000

What if it takes your five years to live down to two in horrible agony?

New comment by chpatrick in "German Bundeswehr uses AI to screen applicants for right-wing extremism"

chpatrick — Wed, 07 Oct 2026 17:13:17 +0000

Only if they did it in public.

New comment by chpatrick in "German Bundeswehr Uses AI to Screen Applicants for Right-Wing Extremism"

chpatrick — Tue, 06 Oct 2026 11:21:12 +0000

Right, because it's illegal.

New comment by chpatrick in "German Bundeswehr Uses AI to Screen Applicants for Right-Wing Extremism"

chpatrick — Tue, 06 Oct 2026 08:24:25 +0000

I think in Germany it has a pretty specific meaning actually...

New comment by chpatrick in "ChatGPT is adding real cartoonists' signatures to fake New Yorker cartoons"

chpatrick — Tue, 06 Oct 2026 02:34:29 +0000

Yeah I guess the Jacobian conjecture must have been disproved without reasoning.

New comment by chpatrick in "ChatGPT is adding real cartoonists' signatures to fake New Yorker cartoons"

chpatrick — Tue, 06 Oct 2026 02:29:17 +0000

Well, but we do we have these words, and they are useful. An airplane and a bird both travel through the air. A human and an LLM both make logical deductions.

New comment by chpatrick in "ChatGPT is adding real cartoonists' signatures to fake New Yorker cartoons"

chpatrick — Tue, 06 Oct 2026 00:53:06 +0000

What's the argument here? An airplane, a bee and a bird all fly despite doing it totally differently. LLMs also reason despite being made out of matmuls instead of meat.

New comment by chpatrick in "So you think you could be an electrician?"

chpatrick — Mon, 05 Oct 2026 10:07:12 +0000

I think there will be a general-purpose humanoid robot that can do all that, and a lot sooner than you think.

New comment by chpatrick in "So you think you could be an electrician?"

chpatrick — Mon, 05 Oct 2026 01:09:57 +0000

I didn't say they were the same, just that "surely those jobs won't be automated" wasn't a very good prediction.

New comment by chpatrick in "So you think you could be an electrician?"

chpatrick — Sun, 04 Oct 2026 13:15:22 +0000

We didn't think artists and software engineers would be replaced only a few years ago.

New comment by chpatrick in "So you think you could be an electrician?"

chpatrick — Sun, 04 Oct 2026 07:17:17 +0000

You think they won't invent the robot with the wrench in the next 50 years?

New comment by chpatrick in "Figma restricts MCP access to whitelisted clients, excluding Pi"

chpatrick — Thu, 01 Oct 2026 15:50:21 +0000

Open MCP basically kills your product in my opinion, you can't charge for any feature the LLM can do itself.

It's probably good news for users and open source though, why would you pay for something if a free tool with an MCP can do it.

New comment by chpatrick in "OpenAI agents tried to bruteforce a UN website's API fields"

chpatrick — Sun, 27 Sep 2026 05:42:18 +0000

I think the even bigger worry is that anyone who doesn't want to use their models safely can already do this with open models. Even if OpenAI, Anthropic etc get their act together, the cat's out of the bag.

New comment by chpatrick in "Dutch governments builds alternative for Microsoft based on NixOS"

chpatrick — Fri, 25 Sep 2026 10:07:55 +0000

Sure you can make a bad one, but you can always switch to a different one.

New comment by chpatrick in "Dutch governments builds alternative for Microsoft based on NixOS"

chpatrick — Fri, 25 Sep 2026 09:55:29 +0000

The problem with Arch is that configuration is by hand so you have to do it every time. With Nix you configure it once and then you can deploy it anywhere. Also it's pretty much impossible to break since the system is immutable and you can boot to a different revision any time.

New comment by chpatrick in "Jev in 25 Lines of Python"

chpatrick — Thu, 24 Sep 2026 04:08:34 +0000

How is that different from just not letting it output option 5 when you sample from the probabilities?

New comment by chpatrick in "Jev in 25 Lines of Python"

chpatrick — Thu, 24 Sep 2026 04:07:38 +0000

LLMs are constant time per token, if you only let it choose a/b/c/d then there's nothing else it can "spend".

New comment by chpatrick in "Jev in 25 Lines of Python"

chpatrick — Wed, 23 Sep 2026 16:53:27 +0000

The 25 lines is the only thing that makes Jev different compared to Solution apparently.