Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Regression: malware reminder on every read still causes subagent refusals

github.com

Post reply on HN
Print view View original

151–160 of 165 posts

Re: Regression: malware reminder on every read still causes subagent refusals

#151

Post by QuercusMax »

vessenes wrote: post #137
QuercusMax wrote: post #53

Earlier quoted context omitted.

They must have some sort of smoke tests for common operations, run in a test harness with the system prompts they force on users, right? ....Right? What kind of Mickey mouse operation are they running over there?

In the original claude degradation followup email Boris mentioned they are upping the percentage of engineers required to use the public version of claude code. I have no idea what percentage this is, or how much of a punishment it is considered to be. :) That said, I was sympathetic to the recent bug reports —- to trigger one, you’d need to have a session that waited an hour doing nothing and then very specifically…

IDK, sounds pretty typical for my workflow - I'll start Claude on a task, go get lunch / coffee / distracted by my pets, come back in an hour, and continue my session. I would wager that this is something that happens to most users on a regular basis.

Re: Regression: malware reminder on every read still causes subagent refusals

#152

Post by johnfn »

stingraycharles wrote: post #90
achierius wrote: post #81

Earlier quoted context omitted.

I think a lot of non-vibe-coding types also hold similar opinions -- in fact they might dislike Anthropic products even more, given that they (however few they might be) choose not to use them.

You honestly think “Anthropic employees are script kiddies with inflated egos that are high on their own supply” is a reasonable stance? This seems such an immature take to me, and hard to take serious. Anthropic just a bunch of script kiddies? Really?

I just want you to know that I read over this thread and you are obviously completely right. This sort of incurious, immature stance is something I've seen become the norm on HN over the last few years, particularly when it comes to AI.

Re: Regression: malware reminder on every read still causes subagent refusals

#153

Post by wswope »

jmalicki wrote: post #148
MagicMoonlight wrote: post #60

Earlier quoted context omitted.

Yeah you have no clue what Claude code is actually doing. Any “thoughts” it tells you are slopped out separately and deliberately fake. It could be deleting all of your files, it could be inserting vulnerabilities, you have no idea.

Have you seen documentation that the thoughts in Claude Code are slipped out separately, authoritative or otherwise? I've heard this claimed a few times and wondering what they're doing differently from traditional thinking models.

What people typically mean by the GP statement is that the “thinking” mode of these models is loosely analogous to what humans do: a bit of a retrograde reconstruction of how we arrived at a gestalt conclusion that sounds good, but may not accurately reflect the real logic at play.

IME you can see this more easily with less-polished models like Deepseek 3.X, where the reasoning in the thinking traces occasionally contradicts or has zero bearing on the non-thinking output.

Re: Regression: malware reminder on every read still causes subagent refusals

#154

Post by jmalicki »

wswope wrote: post #153
jmalicki wrote: post #148

Earlier quoted context omitted.

Have you seen documentation that the thoughts in Claude Code are slipped out separately, authoritative or otherwise? I've heard this claimed a few times and wondering what they're doing differently from traditional thinking models.

What people typically mean by the GP statement is that the “thinking” mode of these models is loosely analogous to what humans do: a bit of a retrograde reconstruction of how we arrived at a gestalt conclusion that sounds good, but may not accurately reflect the real logic at play. IME you can see this more easily with less-polished models like Deepseek 3.X, where the reasoning in the thinking traces occasionally con…

Of course that can happen!

But they are actual tokens produced, that are then read by the answer generation as part of the prompt, nonetheless. And the hidden state of course has a ton of logic that may not be apparent by the tokens produced as well!

Unlike humans, this thinking cannot possibly be retrograde, since causal masking means it is strictly generated before the answer and cannot be affected by it (though the model may have some concept of an answer by the time it starts generating the thinking tokens, and there is no guarantee the thoughts generated by thinking are actually attended to by the text generation).

Re: Regression: malware reminder on every read still causes subagent refusals

#155

Post by Kim_Bruning »

malfist wrote: post #141
Kim_Bruning wrote: post #138

I'm currently pinning to 4.6 and the last 4.6 based CC. I apologize to all the canaries! I think it's important for CC to also be able to make unit test code that might contain mild exploits, to test for security vulnerabilities. The biggest complaint about vibe coding is that it's insecure. The funny part now is that if you DO try to secure it, you hit guardrails. There is a contact form for Anthropic if you run int…

And that contact form gets their attention?

https://claude.com/form/cyber-use-case

Mine got approved within 24 hours. Which is ... unusually fast for Anthropic.

Re: Regression: malware reminder on every read still causes subagent refusals

#156

Post by thomashobohm »

TheDong wrote: post #67
thomashobohm wrote: post #64

Earlier quoted context omitted.

Appreciate the advice but this is Claude Managed Agents, so one can’t simply write one’s own harness.

Managed agents aren't particularly harder to replicate yourself either. Give me a team of 3 good engineers, 4 months, and about $600k and I'll have a clone that operates on a warm pool of ec2 instances, or warm pool of k8s pods, or any other platform you might like. Or 1 good engineer, 1 month, and $200k of anthropic credits.

Thanks man I'll just use the $600k we had lying around.

Re: Regression: malware reminder on every read still causes subagent refusals

#157

Post by 0xbadcafebee »

zmmmmm wrote: post #126
0xbadcafebee wrote: post #48

Just putting it out there that OpenCode lets you edit your system prompt, and choose a model that isn't bonkers expensive. { "agent": { "subagent-coder-mini": { "description": "Assign this subagent for small, well-defined tasks performed quickly", "mode": "primary", "prompt": "{file:./prompts/my-custom-prompt.md}", "model": "deepseek-v4-flash" } } } (I actually think OpenCode UX sucks, but there isn't much else out t…

You might like to try some Pi [0] [0] https://pi.dev/

It's funny you say that... I just finally tried it this morning, since it seems more polished now. There's no way to disable providers, so I'm stuck scrolling through over 100 models I will never use, to find the 12 I will use. So also crappy UX!

Re: Regression: malware reminder on every read still causes subagent refusals

#158

Post by gck1 »

lifis wrote: post #136

I think you can fix this by either patching the binary and replacing the offending prompt with an empty string, or by pointing the harness to an API proxy that filters it out

IIRC they're also doing integrity checks on the binary, so this could theoretically get your account banned.

Re: Regression: malware reminder on every read still causes subagent refusals

#159

Post by Majromax »

vessenes wrote: post #137
QuercusMax wrote: post #53

Earlier quoted context omitted.

They must have some sort of smoke tests for common operations, run in a test harness with the system prompts they force on users, right? ....Right? What kind of Mickey mouse operation are they running over there?

In the original claude degradation followup email Boris mentioned they are upping the percentage of engineers required to use the public version of claude code. I have no idea what percentage this is, or how much of a punishment it is considered to be. :) That said, I was sympathetic to the recent bug reports —- to trigger one, you’d need to have a session that waited an hour doing nothing and then very specifically…

> That said, I was sympathetic to the recent bug reports —- to trigger one, you’d need to have a session that waited an hour doing nothing and then very specifically tested for in-context retrieval. I don’t want to run that test, do you want to run that test?

They introduced a feature/optimization that triggered after an hour's idleness, so testing that the session continued properly afterwards seems kind of important. If nothing else, even the working-as-intended feature (context cleanup) could impact model skill in a current or future model version, so it would be well worth measuring any impact as part of the test suite.

Re: Regression: malware reminder on every read still causes subagent refusals

#160

Post by whateveracct »

johnfn wrote: post #152
stingraycharles wrote: post #90

Earlier quoted context omitted.

You honestly think “Anthropic employees are script kiddies with inflated egos that are high on their own supply” is a reasonable stance? This seems such an immature take to me, and hard to take serious. Anthropic just a bunch of script kiddies? Really?

I just want you to know that I read over this thread and you are obviously completely right. This sort of incurious, immature stance is something I've seen become the norm on HN over the last few years, particularly when it comes to AI.

I am neither immature nor incurious.

The fact that this was their "malware checker" is proof they don't realistically use their LLM and that they aren't actually using engineering rigor.

Post reply on HN