Nomad List founder Pieter Levels reports a Claude Code CLI bug that resets his effort setting to medium daily
A research engineer suggests the issue affects outdated CLI versions.
Users voiced frustration with Anthropic forcing a medium effort setting due to GPU shortages and silent throttling that broke trust and expectations, though some praised medium for feeling faster without quality loss.
No Digg Deeper questions have been answered for this story yet.
Most Activity
@levelsio hmm that’s a bug will investigate
I know Anthropic has a GPU shortage but every day forcefully putting my effort back to medium feels...well....annoying
@levelsio My brother in Christ. Take this and bless your heart.
@levelsio Hmm it should default to xhigh for Opus 4.7. Are you on latest version of Claude Code (CLI)?
I know Anthropic has a GPU shortage but every day forcefully putting my effort back to medium feels...well....annoying
@levelsio Have you added the effort level in the settings.json?
Set "effortLevel": "max" in ~/.claude/settings.json
@levelsio bro u rich, buy like 10 max 20 plans
@levelsio They also bottleneck performance over ssh :)
@levelsio Try codex fella. It’s very good atm
@levelsio I only use max 1M
@levelsio at least make it a popup yep.
@ElkimXOC @levelsio It was a satire, on their gatekeeping ofr other Harnesses. Not true
@levelsio wonder how long you will stay loyal to these shenanigans from anthropic
moving over to codex was the best decision last few months
@ElkimXOC How this works? CLI or SSH?
@ElkimXOC @levelsio Yeah
I was hitting limits consistently on claude 2 weeks ago and I was annoyed, that's why I downgraded from max to pro, but the new acquisition of GPU and then increase of the limits is amazing. I sometimes feel (and then confirm that) I'm getting more from claude than codex.
Most of my code work is getting done by codex and anything with UI/UX by claude and this combination has been working nice.
Not a single point addressing your "stealing" claim.
Try installing a tool that shows you the effective inference in dollar you're getting from your plan. Both claude and codex give me at least 10-20x the amount I pay them.
I know it's subsidized, and I'm actually grateful to have it while it lasts. If that makes me less intelligent and a bad engineer, I don't mind.
btw, peak hour limits were removed 2 weeks ago, when they doubled the 5 hour limit and later weekly limit was also doubled.
@AJalomaki Oh so good will use thx
@AlibekovSanzhar @levelsio Was it?
@tavlean For sure and I agree. My experience - I have never got rate limited on codex's $100/plan, but would get it all the time with Claude's. Also, gpt 5.5 is a far superior model than Opus, and it doesn't tell me to go to bed all the time.
@Adi4x4 @levelsio Little bit faster next time 👊👊👊😂
@jarredsumner @levelsio There's a bug draining my usage.
I use Claude on the web. When a chat gets longer, Claude starts compacting it. After that, if I continue using the same chat, my usage runs out within a couple of prompts.
Even after the usage resets, it gets exhausted again in like 4 queries.