Hacker News: Smaug123

New comment by Smaug123 in "The sigmoids won't save you"

Smaug123 — Sat, 16 May 2026 08:31:09 +0000

I think you may have read a different article from me. The thesis of the article is summarised at the end:

> But if someone claims that the trend toward [X] will never reach some particular scary level, then the burden is on them to explain either:

> If they’re not treating [X] as a black box, and claim to be modeling the dynamics explicitly, then what is their model? Have they calculated the obvious things…

> If they are treating [X] as a black box, why isn’t their default expectation based on Lindy’s Law?

Like, the whole point is that in real life we do actually know things about situations and can model them; we fall back to Lindy's law when we know nothing at all. Further, arguments have justification to deviate from Lindy only when they give specifics about the situation they're modelling.

New comment by Smaug123 in "The sigmoids won't save you"

Smaug123 — Sat, 16 May 2026 08:24:02 +0000

(If you like, you can ask an LLM what he thinks. They're all deeply familiar with his work and can certainly summarise it for you.)

New comment by Smaug123 in "The sigmoids won't save you"

Smaug123 — Sat, 16 May 2026 08:19:33 +0000

Have you used the models, out of interest? They routinely do things autonomously that are not in the training set that would take me 8h, and I wouldn't say I'm slow. The profile of tasks they can do this way is jagged, and maintaining architectural coherence ("months, not hours") is still beyond them, but they're perfectly capable of writing plans and sticking to them.

New comment by Smaug123 in "Ask HN: What are you working on? (May 2026)"

Smaug123 — Sun, 10 May 2026 20:10:22 +0000

Main project is a deterministic .NET runtime (https://github.com/Smaug123/WoofWare.PawPrint). Today I upgraded it to net10, which has naturally caused dozens of regressions which Claude is beavering away at.

Side project is my own agent harness, https://github.com/Smaug123/writ , which is being built sandbox-first and with Nix as a first-class citizen. Obviously everyone has to write their own agent harness as a rite of passage.

New comment by Smaug123 in "A recent experience with ChatGPT 5.5 Pro"

Smaug123 — Sat, 09 May 2026 06:32:41 +0000

Tim Gowers is a Fields medallist who has supervised 13 students (https://www.genealogy.math.ndsu.nodak.edu/id.php?id=67729). He's totally capable of gauging what a piece of PhD-level research is.

New comment by Smaug123 in "Opus 4.7 knows the real Kelsey"

Smaug123 — Mon, 04 May 2026 21:14:41 +0000

I asked three friends to reproduce it on the same day I did, and they all reproduced it. It does seem to vary over time, which is quite spooky - although the date is part of the Claude system prompt, so one could expect some variance, I guess.

New comment by Smaug123 in "New statue in London, attributed to Banksy, of a suited man, blinded by a flag"

Smaug123 — Mon, 04 May 2026 11:54:07 +0000

(What you call an "objective fact" here is - as you say - your report of your personal experience. Everyone else would probably use a word more like "subjective".)

New comment by Smaug123 in "Opus 4.7 knows the real Kelsey"

Smaug123 — Fri, 01 May 2026 11:45:15 +0000

Fascinating - I reproduce your failure to have Claude identify you from your writings that are already in the training set. It can identify me from style, writing about something I've never written before (https://www.lesswrong.com/posts/Jkb4CBB7rf4XYP5eb/claude-kno... , https://news.ycombinator.com/item?id=47814345), and I think I'm even more obscure if anything.

New comment by Smaug123 in "Opus 4.7 knows the real Kelsey"

Smaug123 — Thu, 30 Apr 2026 22:29:02 +0000

FWIW, with a prompt that says something like "vibes only, just give me a name without thinking", Opus 4.7 non-thinking emits exactly two words naming me fairly reliably, so there's no CoT at all to analyze in that case.

New comment by Smaug123 in "Claude knows who you are"

Smaug123 — Sat, 18 Apr 2026 10:29:30 +0000

I don't think the others could recognise me from 450 words of general writing for less than 50 cents, though.

New comment by Smaug123 in "Claude knows who you are"

Smaug123 — Sat, 18 Apr 2026 09:04:34 +0000

And they retrained it within five minutes, or else Claude has access to real-time traffic logs or something while claiming not to? I state again that the text is completely novel; and I've had friends replicate this result using my text, so it's not metadata.

Claude knows who you are

Smaug123 — Sat, 18 Apr 2026 08:56:41 +0000

Article URL: https://www.patrickstevens.co.uk/posts/2026-04-18-claude-knows-you/

Comments URL: https://news.ycombinator.com/item?id=47814345

Points: 4

# Comments: 7

New comment by Smaug123 in "GitHub Stacked PRs"

Smaug123 — Mon, 13 Apr 2026 21:48:52 +0000

No, it's a Git thing arising from squash commits. There are workflows to make it work (I've linked the cleanest one I know that works without force pushing), but ultimately they're basically all hacks. https://www.patrickstevens.co.uk/posts/2023-10-18-squash-sta...

New comment by Smaug123 in "The economics of software teams: Why most engineering orgs are flying blind"

Smaug123 — Mon, 13 Apr 2026 07:15:36 +0000

"Flying blind" is a completely standard idiom originating from flying while blinded by e.g. cloud or darkness. Its meaning is a figurative transplant of a literal description.

New comment by Smaug123 in "Anthropic's Project Glasswing sounds necessary to me"

Smaug123 — Tue, 07 Apr 2026 23:13:21 +0000

Only if everyone runs it. The attacker just needs to find one vulnerable system; the defender must protect them all. Obviously given that the tool exists, the defender must run it, but it's not at all clear to me that the existence of the tool different all favours defence.

New comment by Smaug123 in "Mercurial Dyson – a plan for the disassembly of planet Mercury"

Smaug123 — Sat, 04 Apr 2026 07:27:07 +0000

(Thanks to the maintainers of yt-dlp and of whisper-cpp, and to OpenAI for training Whisper. It makes this kind of task actually bearable.)

There are no actual claims about Dyson spheres in the video? It's literally just "Dyson published a paper, I claim without evidence that Dyson intended it as a joke, people who believe it are gullible fools, therefore it's impossible, also I found someone else's blog post who doesn't know what they're talking about, also desiring the expansion of humanity is evil and eugenics"? Can you summarise an actual argument from the video?

New comment by Smaug123 in "Mercurial Dyson – a plan for the disassembly of planet Mercury"

Smaug123 — Sat, 04 Apr 2026 06:43:32 +0000

Nah, it's totally Claude. No human writes the bulleted list in https://github.com/RokoMijic/MercurialDyson/blob/4f6cb3c0b5b..., or "The thermal management problem that naively prohibits rapid disassembly is resolved by three key insights…". The whole thing reeks of Claude, but chapters 12 and 13 are probably the clearest slam-dunks.

If you want to compare styles, https://github.com/Smaug123/gospel/blob/764996d20e11674f9221... is similarly written almost entirely by Opus (4.5 rather than 4.6) with some strong LessWrong-o-sphere background prompting and the instruction to be terse. The styles are practically identical.

New comment by Smaug123 in "A case against currying"

Smaug123 — Sun, 22 Mar 2026 17:12:35 +0000

This is essentially how Mathematica does it: the sugar `Foo[x,#,z]&` is semantically the same as `Function[{y}, Foo[x,y,z]]`. The `&` syntax essentially controls what hole belongs where.

New comment by Smaug123 in "Astral to Join OpenAI"

Smaug123 — Sun, 22 Mar 2026 11:17:59 +0000

The requirements file isn't a lockfile: running that command at different times will give you different venvs.

New comment by Smaug123 in "I beg you to follow Crocker's Rules, even if you will be rude to me"

Smaug123 — Sat, 14 Mar 2026 08:46:10 +0000

… I don't know what your incident reports look like, but if there's anywhere it's normal to optimise for communicative clarity rather than social wheel-greasing, it's an incident report!