@MatthewBerman Also brutal that they cut off mid-task. Codex allows some extra free over-use to make sure you are not stuck mid-session.
Matthew Berman@MatthewBerman
Holy sh*t after only 5 minutes…
4:20 PM · Jul 1, 2026 · 10.1K Views
OpenAI's Peter Steinberger says sudden AI coding session limits are highly disruptive compared to OpenAI Codex
He calls for extra over-use margins to prevent mid-task lockouts.
364394411.1K
Sentiment
Many users praise Codex for granting extra free usage to complete tasks without mid-task cutoffs, while some call the original cutoff behavior brutal and poorly handled.
Pos
75.0%
Neg
25.0%
20 comments with sentiment.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Posts from X
Most Activity
Most Activity
VIEWS674LIKES15REPLIES2
Beff (e/acc)@beffjezos
@steipete @MatthewBerman we need 5.6 bruv
Peter Steinberger 🦞@steipete
@MatthewBerman Also brutal that they cut off mid-task. Codex allows some extra free over-use to make sure you are not stuck mid-session.
4hViews 674Likes 15Bookmarks 0
RETWEETS1
CodeWalkers@WalkingTheCode
@steipete @MatthewBerman It seems like Codex allows any running tasks to complete (or gracefully reach a stopping point) but prevents you from sending new messages when you hit your limits.
That is a well designed way to handle limits without cutting off abruptly mid-session.
5hViews 81Likes 1
Peter Steinberger 🦞@steipete
@moidrk @MatthewBerman It even works when you used up all your quota
4hViews 479Likes 7
Matthew Berman@MatthewBerman
@beffjezos @steipete OpenAI is much more generous with quotas
Beff (e/acc)@beffjezos
@steipete @MatthewBerman we need 5.6 bruv
4hViews 420Likes 2Bookmarks 0
Moid@moidrk
@steipete @MatthewBerman is it really free, i always thought if your 5 hour limit is over and you are amidst a task, it just takes from your overall weekly limit and bypass the 5 hour limit. but free?? really
4hViews 115Likes 1
Miller Guevara@MillerGuevara_
@steipete @MatthewBerman As it should be. I've always wondered why Claude can't just give us a "window". It always stops so suddenly, and re-starting burns so many tokens...
5hViews 235
Peter Steinberger 🦞@steipete
@beffjezos @MatthewBerman Yes
4hViews 355Likes 1
Tyler Bea@skibumtrading
@steipete @MatthewBerman Yeah i love that about Codex. When GPT 5.6?
4hViews 104Likes 2
Aljosa Asanovic@aljosa
@steipete @MatthewBerman Oh shit that's intentional?
It's so good I feared it was a bug they'd fix at some point.
5hViews 242Likes 1
Kevin Guaman@KCG3D
@steipete @MatthewBerman can't wait for GPT 5.6 Sol to help us forget about all this bs
5hViews 192Likes 1
Artificially Inclined™@Art_If_Ficial
@steipete @MatthewBerman Huh I assumed I was special/lucky every time it holds out 😤
5hViews 135Likes 1
Moid@moidrk
@steipete @MatthewBerman okay my respect for codex just doubled <3
4hViews 52Likes 1
Wes Winder@weswinder
@steipete @MatthewBerman yeah mid task cutoff is brutal
5hViews 151
Muhammad Eman Aftab@m_emanaftab
@steipete @MatthewBerman I think its a great thing, atleast it completes the work that he started.
5hViews 146
loganaden velvindron@loganaden_42
@steipete @MatthewBerman Yes agreed but how many can use 5.5- cyber outside if the usa ?
4hViews 78
Markus Baumann@MarkusBaumannAi
@steipete @MatthewBerman Great feature!
5hViews 74
Satyam@stym06
@steipete @MatthewBerman I don't think that's true...rather codex just goes off and you gotta type resume and that ends up responding with limit hit
5hViews 72
Patrick @ ProductRoot.io@ProductRootIO
@steipete @MatthewBerman OpenAI maintains a clear advantage over Anthropic regarding fair use, and GPT 5.6 will enhance that edge even further :)
4hViews 68
Ardjo@rdjooo
@steipete @MatthewBerman Yeah, OpenAI people actually care to deliver value.
5hViews 61
Snax@aRobotNamedSnax
@steipete @MatthewBerman Any whispers of when 5.6 is dropping ?
5hViews 48
Load more posts