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

Issue: Claude Code is unusable for complex engineering tasks with Feb updates

github.com

Post reply on HN
Print view View original

811–820 of 829 posts

Re: Issue: Claude Code is unusable for complex engineering tasks with Feb updates

#811

Post by ericpan »

robeym wrote: post #707
bcherny wrote: post #209

Hey all, Boris from the Claude Code team here. I just responded on the issue, and cross-posting here for input. --- Hi, thanks for the detailed analysis. Before I keep going, I wanted to say I appreciate the depth of thinking & care that went into this. There's a lot here, I will try to break it down a bit. These are the two core things happening: > `redact-thinking-2026-02-12` This beta header hides thinking from th…

This is confusing. ULTRATHINK is a step below /effort max? ULTRATHINK triggers high effort. /effort max is above high. Calling it ULTRATHINK sounds like it would be the highest mode. If someone has max set and types ULTRATHINK, they're lowering their effort for that turn. For anyone reading this trying to fix the quality issues, here's what I landed on in ~/.claude/settings.json: { "env": { "CLAUDE_CODE_EFFORT_LEVEL"…

The docs say that CLAUDE_CODE_EFFORT_LEVEL controls adaptive reasoning intensity, and CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING bypasses that entirely in favor of a fixed budget via MAX_THINKING_TOKENS. So setting both is contradictory. If true, disabling adaptive thinking would override what effort level is trying to do.

https://code.claude.com/docs/en/env-vars

Re: Issue: Claude Code is unusable for complex engineering tasks with Feb updates

#812

Post by cordwainersmith »

We rolled it out across ~1k engineers and the biggest issue wasn't the model quality, it was observability. Nobody could tell me if the agent was stuck in a loop, which sessions were expensive, or what the cache hit rate looked like. Without that visibility you can't distinguish "the model is bad" from "my setup is bad." Most of the complaints we got early on turned out to be config problems.

Re: Issue: Claude Code is unusable for complex engineering tasks with Feb updates

#813

Post by robeym »

ericpan wrote: post #811
robeym wrote: post #707

Earlier quoted context omitted.

This is confusing. ULTRATHINK is a step below /effort max? ULTRATHINK triggers high effort. /effort max is above high. Calling it ULTRATHINK sounds like it would be the highest mode. If someone has max set and types ULTRATHINK, they're lowering their effort for that turn. For anyone reading this trying to fix the quality issues, here's what I landed on in ~/.claude/settings.json: { "env": { "CLAUDE_CODE_EFFORT_LEVEL"…

The docs say that CLAUDE_CODE_EFFORT_LEVEL controls adaptive reasoning intensity, and CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING bypasses that entirely in favor of a fixed budget via MAX_THINKING_TOKENS. So setting both is contradictory. If true, disabling adaptive thinking would override what effort level is trying to do. https://code.claude.com/docs/en/env-vars

So if it bypasses, is the optimal setting for performance setting effort level to max, keeping adaptive on? I try to avoid letting the model decide what is unimportant and needs less thought

Re: Issue: Claude Code is unusable for complex engineering tasks with Feb updates

#816

Post by tracker1 »

monatron wrote: post #344
johndough wrote: post #331

Earlier quoted context omitted.

I think it is hilarious that there are four different ways to set settings (settings.json config file, environment variable, slash commands and magical chat keywords). That kind of consistency has also been my own experience with LLMs.

To be fair, I can think of reasons why you would want to be able to set them in various ways. - settings.json - set for machine, project - env var - set for an environment/shell/sandbox - slash command - set for a session - magical keyword - set for a turn

I tend to make a concerted effort to often make sure anything settable via cli is settable via environment variable... though, I often have a search-upward option for a .env file as well. Mostly so that it's easier to containerize/deploy an application in a predictable/reusable way.

Re: Issue: Claude Code is unusable for complex engineering tasks with Feb updates

#817

Post by burgerzzz »

lelanthran wrote: post #648
matheusmoreira wrote: post #545

Earlier quoted context omitted.

Literally two weeks ago it was outputting excellent results while working with me on my programming language. I reviewed every line and tried to understand everything it did. It was good. I slowly started trusting it. Now I don't want to let it touch my project again. It's extremely depressing because this is my hobby and I was having such a blast coding with Claude. I even started trying to use it to pivot to profes…

> People who depend on this to make a living must be very angry indeed. Oh cry me a fucking river. The people depending on this to make a living don't have the moral high ground here. They jumped onboard so they could replace other people's living, and those other people were angry too. They didn't care about that. It's hard to care about them when the thing they depend on to make a living got yanked, because that's…

Since when am I responsible for other people's living?

Re: Issue: Claude Code is unusable for complex engineering tasks with Feb updates

#818

Post by ccmcarey »

dostick wrote: post #355
bcherny wrote: post #253

Earlier quoted context omitted.

Yep, exactly

Mentioning ULTRATHINK in prompt is the equivalent to /effort max?

No, ultrathink puts it in /effort high mode. There's no kw for one turn of effort max

Re: Issue: Claude Code is unusable for complex engineering tasks with Feb updates

#819

Post by liamsfr »

So: 1/ lack of thinking in transcripts is not a decisive metric for determining if any thinking was done, but 2/ the reply does not address the qualitative aspects that Stella’s team observed and provided data for from what amounted to a bad qualitative experience with serious financial implications.

It’s a sidestep for explaining away the research, but does not address the underlying issue: has quality been degrading (selectively, intentionally or otherwise)?

Re: Issue: Claude Code is unusable for complex engineering tasks with Feb updates

#820

Post by liamsfr »

kouteiheika wrote: post #798
olejorgenb wrote: post #784

Earlier quoted context omitted.

So this means I can not resume a session older than 30 days properly?

I have no idea; you have to check their docs. AFAIK what they do is that they calculate a hash of the true thinking trace, save it into a database, and only send those hashes back to you (try to man-in-the-middle Claude Code and you'll see those hashes). So then when you send then back your session's history you include those hashes, they look them up in their database, replace them with the real thinking trace, and…

So we are paying the price for the cost of infra need to protect their asset which was trained on data derived from the work of others while ignoring the same principle? I need this to make sense.
Post reply on HN