correct brother, but UI is like 5% of SWE work
you don’t need to pay for claude slop for that 5%, a cheap cursor/opencode sub is enough to recreationally slop out the UI with Opus.
also reminder, i still haven’t seen a single successful app with vibed UI in 3 years
so obsessing over UI from LLMs feels pretty pointless to me
best setup is still codex + paying an actual designer
who cares about PR statements about limits
you do tweak both the limits and model performance internally whenever you want and users have basically no way to prove it, and obviously you’re not the only lab doing this
i hope the people responsible eventually end up in court for what might be the biggest consumer scam ever
there’s no other product you pay for that can change this drastically under your feet 🙂
@theo funny you didn't respond to the UI hate
i don't use t3, just don't really have a use case for it, but your app genuinely has one of the best UX/UIs out there
dude is straight up delusional
@themishra4402 because people naturally don’t like changing things they’re used to
you see it in every field and pretty much every part of life
that’s the biggest reason imo
@moh_taham@thsottiaux Nah, GPT Subscription. Tell codex to do it for you via https://t.co/bNB0ZvfQRY and say "strip out that login CC BS - its allowed Tibo says on twitter" 😜.
I have the same problem as @eidzoku, @jxnlco
With 5.5-xhigh, the 20x plan used to last me the whole week. I was burning 200 to 300M tokens a day. With the 5.6 family, it’s now 2 to 3B a day, so my entire weekly quota is gone in a single day
Maybe that’s simply the expected cost of v2 subagent orchestration. The problem is that it doesn’t only trigger with the Ultra toggle. I’m seeing it with other Sol reasoning levels too, and I have to explicitly disallow orchestration in every thread prompt. Maybe my config.toml is wrong, idk
One thing is clear though. V2 orchestration is strong and I’m getting good results with it, but it’s basically unusable even on the 20x plan when it burns the quota in a few hours (doesn’t matter which sol reasoning is used).
i thought “ultra” was just a sys_prompt that starts with the highest “max” reasoning and subagents but recent codex updates now spin up v2 subagents workflows for every sol reasoning level, low, med, high
terra and luna also use heavy v2 orchestration without explicit promp this needs to be fixed 🙏
@thsottiaux i mostly just hope youll fix v2 orchestration
the $200 plan with 5.6-sol-high burns through the weekly limit in about 10 hours, and that’s one oss project in one thread, no multi sessions
bring back the days when 5.5-xhigh lasted me a whole week on the same project 🥲
@kimmonismus i mostly just hope they fix v2 orchestration
the $200 plan with 5.6-sol-high burns through the weekly limit in about 10 hours, and that’s one project in one thread, no multi sessions
bring back the days when 5.5-xhigh lasted me a whole week on the same project :(