Hacker News: petercooper

New comment by petercooper in "Sleep research led to a new sleep apnea drug"

petercooper — Sat, 23 May 2026 09:20:29 +0000

There are some older studies that showed nicotine (not via smoking, which makes sleep apnea worse) had a positive effect on sleep apnea as well due to the stimulation increasing muscular activity in the airways. Obviously lots of downsides too so it never caught on, but this seems to have a similar mechanism.

New comment by petercooper in "NTSB pulls docket after AI recreates dead pilots' voices"

petercooper — Fri, 22 May 2026 21:33:29 +0000

I'm only an ardent viewer of crash investigation stuff, not a pro, but it seems to be a good way to show specifics of warning noises, engine sounds, unusual cabin noises (if relevant) and sometimes even structural failures happening over time in a more direct way from the cockpit voice recorder without sharing the actual "audio".

New comment by petercooper in "Node.js 26.0.0 (Now with Temporal)"

petercooper — Thu, 21 May 2026 17:54:03 +0000

Criticism noted! But it is experimental, unconstrained FFI is inherently unsafe, and not all JavaScript developers are "mildly educated web devs".

I doubt we'll see this go into mainstream dependencies likely to be used by such developers – it hasn't in Ruby which has had easily accessible FFI mechanisms for years.

Though, there is already a fun experiment using it: https://blog.platformatic.dev/destino-doom-terminal-nodejs-f...

New comment by petercooper in "Node.js 26.0.0 (Now with Temporal)"

petercooper — Wed, 20 May 2026 20:38:10 +0000

https://nodejs.org/en/blog/release/v26.1.0 is particularly cool as it added initial FFI support.

New comment by petercooper in "A few words on DS4"

petercooper — Fri, 15 May 2026 19:20:53 +0000

I agree philosophically about building more takes on ideas to flesh out ideas. I guess I was querying more the idea about an agent being part of DS4 specifically.

I'm 100% up for an "agent by antirez", but I'm intrigued why it would/might be part of DS4 itself. Is there something extra to gain from a tighter coupling between inference and harness? (My gut instinct is.. maybe? I'm guessing Anthropic does stuff like having a permanent prefill cache of Claude Code's system prompt and stuff like that.)

New comment by petercooper in "A few words on DS4"

petercooper — Fri, 15 May 2026 09:36:53 +0000

I've been using the Q4 version on my Mac Studio over my local network and it's been good. Indeed, I had the first ever experience where I was playing with it alongside my various other agents and forgot it was a local model as it was doing such a good job.

I do wonder, though, if another agent is really needed. I've been driving it with Pi (Claude Code's system prompt is far too heavy given the prefill speeds) and it's been great. OpenCode is another good option. Is there anything else to gain from another similar tool specific to Deepseek 4?

New comment by petercooper in "Show HN: Hallucinopedia"

petercooper — Wed, 06 May 2026 17:05:53 +0000

Give it a week and see what Google AI Overview has to say about the Great Pigeon Census of 1887!

New comment by petercooper in "Higher usage limits for Claude and a compute deal with SpaceX"

petercooper — Wed, 06 May 2026 16:56:24 +0000

I don't know if it relates to the same data centers, but this also comes hours after several still recent Grok models were deprecated at short notice. Grok 4.1 Fast is the cheapest way to do research on X (cheaper than the X API!) and it's gone on May 15: https://docs.x.ai/developers/models - freeing up compute to sell?

New comment by petercooper in "Using “underdrawings” for accurate text and numbers"

petercooper — Mon, 04 May 2026 11:21:00 +0000

Because image models at the basic level are just text tokens in, image tokens out. You'd need an agentic process on top to come up with a strategy, review output, try again, and so on.

I believe Nano Banana and gpt-image-2 have a little of this going on, but it's like asking a model to one-shot some code vs having an agentic harness with tools do it. Even the most basic agent can produce better code than ChatGPT can.

New comment by petercooper in "Using “underdrawings” for accurate text and numbers"

petercooper — Mon, 04 May 2026 11:19:18 +0000

This seems analogous to how a human would do it accurately. If you asked an artist to paint stones in a large circular arrangement with the numbers in order in one shot, with no fixes or sketching allowed, it wouldn't be surprising to end up with problems in the arrangement.

New comment by petercooper in "Apocalypse Early Warning System"

petercooper — Fri, 01 May 2026 20:58:33 +0000

Bloomberg Terminal chat?

New comment by petercooper in "Reverse Engineering SimTower"

petercooper — Fri, 01 May 2026 14:49:36 +0000

I did something similar, though at a much smaller scale, with Windows 3.1 Solitaire a few weeks ago, both to see how possible it is (with LLMs) and also to learn about Ghidra for myself. Sadly, I discovered nothing new that wasn't already surfaced through previous reverse engineering, but did turn it into a behavior-equivalent (almost, Draw One by default, and I got lazy with Vegas scoring) Web port: https://peterc.org/solitaire/ (you can press W to trigger the win animation without winning, just for laughs)

I can certainly believe a much larger app like SimTower could overwhelm an LLM-based process without a lot of handholding as even Solitaire required a fair bit of that from me. I'm hoping to take on 1991's A-Train next (the game that inpired SimCity 2000's interface).

New comment by petercooper in "Claude Code refuses requests or charges extra if your commits mention "OpenClaw""

petercooper — Thu, 30 Apr 2026 18:37:04 +0000

I used the term "smuggling" in the casual sense of hiding something. I have edited it to "place such identifiers surreptitiously" to avoid making whatever implication appears to have been taken.

New comment by petercooper in "Claude Code refuses requests or charges extra if your commits mention "OpenClaw""

petercooper — Thu, 30 Apr 2026 17:23:16 +0000

I wonder if projects which are anti-AI could place such identifiers surreptitiously into docs or commits as a way to sabotage people using Claude Code. Your project isn't going to get many AI PRs if just cloning your project wiped out their quota.

New comment by petercooper in "Making UIs like text adventure games"

petercooper — Mon, 27 Apr 2026 18:53:59 +0000

This idea made me think that maybe point and click adventure gams could provide an interesting model here, too. You still have nouns and verbs and locations, but the possible nouns and verbs are made much more explicit.

New comment by petercooper in "Gte-go – my over-optimized embedding model"

petercooper — Mon, 27 Apr 2026 11:17:13 +0000

Nice work! I've been simultaneously working on something similar though a totally different approach. I'm using `snowflake-arctic-embed-s` (which I've benchmarked to be the most accurate small embedding model for my use case - source code embedding) and a pure Go approach, but unsurprisingly it is somewhat slower than your approach. Did you attempt a pure Go version at first before adding in the Go assembly?

New comment by petercooper in "GPT‑5.5 Bio Bug Bounty"

petercooper — Sat, 25 Apr 2026 17:41:19 +0000

It sounds like asking CS PhDs to do a world record speed run. I wouldn't be surprised if the people best suited to the task aren't the type to get onto "a vetted list".

New comment by petercooper in "AI chatbots could be making you stupider"

petercooper — Mon, 20 Apr 2026 19:42:06 +0000

I'll need to engage in conjecture over elementary school lessons from 35 years ago, but one thing that comes to mind is we were calculating circle circumferences and areas quite quickly following the formulas. We still learnt arithmetic techniques by hand (though never logarithms, for whatever reason - I guess calculators replaced the log tables!), but when we moved on to broader things like geometry and statistics, calculator use let us focus on the actual topics and formulas and not repeating the grunt work like generations past.

For anything beyond that, I'd need to take it up with whoever wrote our curriculum! But I know it was mildly contentious at the time, much as the use of even more elaborate technologies are now.

New comment by petercooper in "AI chatbots could be making you stupider"

petercooper — Mon, 20 Apr 2026 18:29:12 +0000

I remember when my school introduced calculators and my parents got upset about it: "They won't learn to do sums in their heads!" Yet it opened us up to working on more interesting, larger problems, at a faster pace. LLMs could atrophy skills if used solely out of laziness (like the cover letters in the post), but they could also help you punch higher, and learn more, and faster, if you're motivated and mentally integrate them properly.

New comment by petercooper in "Codex Hacked a Samsung TV"

petercooper — Thu, 16 Apr 2026 12:14:59 +0000

Not as cool as this, but I had a fun Claude Code experience when I asked it to look at my Bluetooth devices and do something "fun". It discovered a cheap set of RGB lights in my daughter's room (which I had no idea used Bluetooth for the remote - and not secured at all) and made them do a rainbow effect then documented the protocol so I could make my own remote control if needed.