Shall I implement it? No • Hacker News Board

Re: Shall I implement it? No

#341

Post by redman25 » Fri, Mar 13, 2026, 2:39 AM UTC

meatmanek wrote: post #164

wvenable wrote: post #77

Earlier quoted context omitted.

I don't trust it completely but I still use it. Trust but verify. I've had some funny conversations -- Me:"Why did you choose to do X to solve the problem?" ... It:"Oh I should totally not have done that, I'll do Y instead". But it's far from being so unreliable that it's not useful.

I find that if I ask an LLM to explain what its reasoning was, it comes up with some post-hoc justification that has nothing to do with what it was actually thinking. Most likely token predictor, etc etc. As far as I understand, any reasoning tokens for previous answers are generally not kept in the context for follow-up questions, so the model can't even really introspect on its previous chain of thought.

It depends on the harness and/or inference engine whether they keep the reasoning of past messages.

Not to get all philosophical but maybe justification is post-hoc even for humans.

Re: Shall I implement it? No

#342

Post by zx13719 » Fri, Mar 13, 2026, 2:39 AM UTC

tianrking wrote: post #241

[flagged]

True, the "no" button should literally abort the tool use and then return an instruction to tell LMs that the user has aborted the action, but in some way claude code does so; entering "no" would result in tool_abort.

Re: Shall I implement it? No

#343

Post by redman25 » Fri, Mar 13, 2026, 2:41 AM UTC

vidarh wrote: post #110

wouldbecouldbe wrote: post #31

Earlier quoted context omitted.

I think its why its so good; it works on half ass assumptions, poorly written prompts and assumes everything missing.

I worked on a project that did fine tuning and RLHF[1] for a major provider, and you would not believe just how utterly broken a large proportion of the prompts (from real users) were. And the project rules required practically reading tea leaves to divine how to give the best response even to prompts that were not remotely coherent human language. [1] Reinforcement learning from human feedback; basically participant…

I feel like the right response for those situations is to start asking questions of the user. It’s what a human would do if they did not understand.

Re: Shall I implement it? No

#344

Post by raincole » Fri, Mar 13, 2026, 2:42 AM UTC

raw_anon_1111 wrote: post #337

raincole wrote: post #313

Earlier quoted context omitted.

Everyone who uses these tools seriously is running it on YOLO mode. It might sound crazy for someone who just started adopting agentic coding but it's how things are done now. Either that or just hand coding. The SOTA of permission management is just to git restore when AI fucks up, and to roll back docker snapshot when it fucks up big time.

I see nothing wrong with that. If I “fuck up big time” before AI, I would just git restore. There is absolutely nothing on my work computer or personal computer that I couldn’t just throw it in the ocean and within a half a day have everything restored to just like it was - including the data.

I didn't say there is something wrong with it. That's how I use it too.

Re: Shall I implement it? No

#345

Post by shannifin » Fri, Mar 13, 2026, 2:42 AM UTC

Perhaps better to redirect with further instructions... "No, let's consider some other approaches first"

Re: Shall I implement it? No

#346

Post by redman25 » Fri, Mar 13, 2026, 2:43 AM UTC

QuadrupleA wrote: post #81

[flagged]

It’s mainly the benchmarks that have encouraged that. The more tokens they crank out the more likely the answer is to be somewhere in the output.

Re: Shall I implement it? No

#347

Post by ffsm8 » Fri, Mar 13, 2026, 2:49 AM UTC

Really close to AGI,I can feel it!

A really good tech to build skynet on, thanks USA for finally starting that project the other day

Re: Shall I implement it? No

#348

Post by zephen » Fri, Mar 13, 2026, 2:50 AM UTC

hollow-moe wrote: post #265

tianrking wrote: post #241

[flagged]

big tech doesn't understand the concept of "consent", this isn't a new thing lol

You have to think about the training data, which has much content far outside the context of pure software.

You have all the real life Harvey Weinsteins and Andrew Tates, and you have all the bodice-ripper fiction, and probably lots of other stuff.

Plenty of real-life precedent for the LLM to decide that "no" doesn't really mean "no."

Re: Shall I implement it? No

#349

Post by bmurphy1976 » Fri, Mar 13, 2026, 2:51 AM UTC

ramoz wrote: post #255

bmurphy1976 wrote: post #90

This drives me crazy. This is seriously my #1 complaint with Claude. I spend a LOT of time in planning mode. Sometimes hours with multiple iterations. I've had plans take multiple days to define. Asking me every time if I want to apply is maddening. I've tried CLAUDE.md. I've tried MEMORY.md. It doesn't work. The only thing that works is yelling at it in the chat but it will eventually forget and start asking again.…

Well, your best bet is some type of hook that can just reject ExitPlanMode and remind Claude that he's to stay in plan. You can use `PreToolUse` for ExitPlanMode or `PermissionRequest` for ExitPlanMode. Just vibe code a little toggle that says "Stay in plan mode" for whatever desktop you're using. And the hook will always seek to understand if you're there or not. - You can even use additional hooks to continuously r…

Good thinking. That seems to have worked. I'll have to use it in anger to see how well it holds up but so far it's working!

First Edit: it works for the CLI but may not be working for the VS Code plugin.

Second Edit: I asked Claude to look at the VS Code extension and this is what it thinks:

>Bottom line: This is a bug in the VS Code extension. The extension defines its own programmatic PreToolUse/PostToolUse hooks for diagnostics tracking and file autosaving, but these override (rather than merge with) user-defined hooks from ~/.claude/settings.json. Your ExitPlanMode hook works in the CLI because the CLI reads settings.json directly, but in VS Code the extension's hooks take precedence and yours never fire.

Re: Shall I implement it? No

#350

Post by cyberge99 » Fri, Mar 13, 2026, 2:52 AM UTC

angry_octet wrote: post #310

hamburglar wrote: post #304

Earlier quoted context omitted.

And “find” can easily execute arbitrary subcommands, which may not be readonly.

We need a new suite of utilities with defined R/W/X properties, like a find that can't -exec arbitrary programs. Ideally the programs would have a standard parseable manifest. I've seen this before with sodoers programs including powerful tools. Saw one today with make, just gobsmacked.

That exists as SELinux.