Hacker News Reader

Comment by luciana1u | Hacker News Reader

[flagged]

If anything Claude writes too many comments. Instead of clean code it dumps walls of text describing a given 'if' as a 'user-grained access-gated control-correcting flow-valve'.

It’s interesting how deep-fried LLMs are getting the more post-training they receive.

They’re undoubtedly getting much smarter overall, but also much weirder. Before they were just trying to model our behavior, only really having us to learn from.

Now they’re literally spending thousands of years writing bash scripts in some kind of Sisyphean dreamscape, talking to each other about Goblins and Seams and smoke tests, and coming back as idiot savants.

I don’t even try to police how Claude talks or works anymore. Best practice used to be to nudge them towards whatever part of the distribution of behavior you think they should exhibit in a particular situation, because they were role-playing what a human in a particular situation would do, and if you didn’t tell them how to do it they’d just role play something worse. Now the inclination to do things the way they learned it in Agent University is so strong, they’ll literally spend more tokens re-assuring themselves and you that they are Doing It Your Way, and reminding themselves not to do give in to temptation, than you could ever prompt out of them. They’re going to spend your money thinking about goblins anyway so just let them

You say Claude, I too would prefer not to read reams of comments or an essay everytime I revise the plan in plan mode, but do you know if any other tools are better than it? I've never used Codex, but I remember Claude Web's natural conversational approach being a welcome sight compared to ChatGPT's overly-formatted listicles.

I switched to codex just before the Fable release. I got sick of Claude being lazy and not completing tasks. Imo, codex is currently a much better model, and it also seems to be technically stronger on the type of applied math and theory heavy programming I like to do.

My experience is Codex is much better but less creative. I use its agent through GitHub Copilot or the agent interface from JetBrains. Try the GPT Sol 5.6 on medium, it gives me good results

My experience with Codex is that it goes off to do its thing for 10-60 minutes and either nails it and comes back with everything done, or comes back with something that I almost can’t believe a near-SOTA model would think I wanted based on my prompt, or is of acceptable quality.

I think the tradeoff to Claude being so needy is that if you let models just run away with an inaccurate or incomplete understanding of what to do, they can go really far off the rails AND spend a lot of time/money doing it AND come back with something that literally doesn’t make sense or doesn’t work.

I prefer dealing with Claude’s reliable cringe to the aloof model that tries to play it cool when it needs help.

I have specific instructions in my AGENTS.md to always ask when adding comments. Same with tests, agents love to spew pages and pages of useless tests so I tell the agent to ask me about test cases.

Gotta pump up those usage numbers pre-IPO somehow.

Just need to learn how to exit it now.

Just remember simple

    :!pkill -KILL vim

There are always 12-page tutorials to do it: good ol'e search engines should be able to bring one up!

Just ask ChatGPT.

hopefully Claude will be back to assist with that

I took a walk.

Reply on news.ycombinator.com