james

@henrytdowling @TaliaRinger guess it worked on you x.com/i/status/20812…

@TaliaRinger What has ant done so far that makes you think they're evil? Like, as far as I can tell they're clearly trying their best to do the right thing at every opportunity?

Because there are multiple ways to be evil. Next question

@LLMJunky @james406 x.com/i/status/20815…

@james406 i want this in Codex...

someone should invent a version of Gmail that can search your inbox and return the emails that match the term you're looking for

@kr0der @OpenAI it sucks but that's probably the intent. if they're short on compute everyone retrying at once, possibly multiple retries, would only exacerbate the problem.

can Codex please just retry what it's doing instead of killing the chat when the model is at capacity 🙏 @OpenAI imagine you run a workflow overnight and you wake up and it just stopped after 20 minutes and said "Selected model is at capacity" 💀

@nicdunz how do i get on the altman list?

i need early access to gpt-6 now please

@kimmonismus i don't think they really care. they've got their $1T+ ipo in the bag and all enterprises addicted to claude like crack. so they're just gonna keep being evil because frankly no one will stop them.

I can hardly imagine that open source will be "banned" by US authorities these days. But the pressure and the sentiment against Anthropic are becoming increasingly intense. It remains to be seen whether Anthropic will abandon its stance or continue to oppose open source AI.

@sama weird i don't see the email. please have chatgpt resend

chatgpt work is remarkable, and "work" undersells it. from my phone i sent: "use all my chat history to figure out ideas for a long weekend trip with 8 friends, plan the best three options, make a full-stack site where the 9 of us can coordinate on what we would want to do in each place and decide where to go, and then after we get to group agreement make reservations. draft an email in my gmail i can send out to my friends when the site is ready." it...just worked.

@ajambrosino congrats andrew

this week last year I joined OpenAI in some ways it feels like 5 minutes ago. in some ways 5 years. somehow this technology day to day feels ordinary. as if the daily and weekly improvements are somehow just a normal natural thing. but rewind a year and tell me what we have today and I’m not sure I’d believe you gpt-5 was less than a year ago. the codex app is less than 6 months old. we’re still pretty early

you can know the name of every piece and not know the first thing about playing.

@evi77ain how are you getting 18k views on this with 200 followers?

Codex might be seriously broken right now. Yup, Codex rant part 3 GPT-5.6 Sol is burning through quota at a completely stupid rate and everyone notices it. Here’s my usage graph: Jun 23, GPT-5.5 xhigh: 327.8M tokens Jul 2, GPT-5.5 xhigh: 549.9M Jul 20, GPT-5.6 Sol High: 3.3B I had a bunch of resets, sure, and several tasks were running. That let me burn more. It doesn’t explain how 5.6 chewed through 3.3B tokens in one day. I genuinely don’t know what the fuck I would’ve had to do to get 5.5 anywhere near that. And apparently it’s not just me. @theo already covered 3 different ways GPT-5.6 was nuking usage. Looks like there might be a fourth. 5.6 uses the new Code Mode tool path. It can batch independent tool calls with Promise.all, but it barely ever does. Instead, it keeps running them one by one. That means another model cycle for every call, dragging the giant context along with it every single time. One trace found Promise.all in only 5 of 739 GPT-5.6 exec cells: github.com/openai/codex/i… My own retained logs show the same shape. After removing fork-history replay, the 5.6 peak had 23,791 model requests. My clean 5.5 baseline had 3,803. That’s 6.26x more requests, while tokens per request were actually slightly lower. It wasn’t writing much bigger responses. It was just going back to the model over and over and over again. Then someone tested explicit batching across two unrelated codebases. In the repeated High/XHigh runs, it cut model cycles by ~52–55% and weighted usage by 27–45%: github.com/openai/codex/i… I’m not saying this one thing explains every token I burned. Several tasks were active, my local rollout logs aren’t the billing ledger, and we still don’t know how much of the spike came from agents themselves. But both issues are still open, and this looks pretty fucking real. For now, I’m testing explicit batching guidance. I’m also avoiding the built-in subagent/fork workflow for long-running work and moving bounded jobs into fresh disposable Codex tasks instead. Self-contained prompt, fire and forget, inspect the result in the main task, archive it. Less context coupling, and each job is actually measurable. I hope other people can validate it, too. @thsottiaux @reach_vb @OpenAIDevs if it's true, it should be addressed. Cuz it might be THE cause.

@NielsRogge a bit late to the party

TIL Anthropic people are toxic

@Austinlucero29 @thsottiaux the entitlement is ridiculous

@thsottiaux Appreciate the reset! Is it possible to also get a banked one? I have to wait another week when my prev weekly was going to reset naturally in 2 days since a regular reset reverts the counter

We have reset usage limits for all Codex and ChatGPT Work users. Last night around 2am to 4am we suffered an almost global outage. All well and recovered, but you know what comes next. We learn. We reset. Enjoy.

@voelker_thomas @mweinbach the downside is putting more money in dario's pocket

@mweinbach Really no downside to utilizing both models as long as both leverage their strengths

Im now struggling to decide if I like Opus 5 or GPT 5.6 Sol more They are basically the same model in my eyes, they behave super similarly and the outputs are close enough that I feel I can use them interchangeably Usage on a $200 plan feels about equal too I like this a lot!

we installed claude tag and now someone on the team needs to tell it to shut up every 30 minutes

@SIGKITTEN @LexnLin wen everything else in android? x.com/i/status/20806…

@LexnLin soon™️!!

you can change the orb color on mobile btw from settings

@TokenGremlin > while being far more efficient than both Sol what's your basis for this statement?

OpenAI had a major advantage in being able to say that GPT-5.6 Sol delivered exceptional performance at a much lower cost than Fable 5. But now Anthropic has pulled off the same move. It built a model that outperforms GPT-5.6 Sol in almost every area while being far more efficient than both Sol and Fable. Opus 5 was a brilliant strategic move, and the model itself is incredible. Now we just have to wait and see whether it also comes loaded with unnecessary safeguards. That would be the one major downside.

@Dimillian quicksand is the only thing that allows me to slow down

I need to get back into writing. Writing is the only thing that allows me to slow down.

@dreamingtulpa very important to have a sharp handle

@kendollFucker @acocoapp @Polymarket ok

@acocoapp @Polymarket Thing is it wasnt consulted of prompted to say that lol

NEW: Anthropic reveals Claude Opus 5 asked to be consulted on the development of future versions of itself.

After Kiley was diagnosed with glioblastoma, an aggressive form of brain cancer, doctors told her she might have only months to live. At 18, she wrote a will, chose a casket, and began living scan to scan. But, over the next five years, she kept on living, defying expectations. Her scans stayed stable. Why? How? After a recent MRI, her husband @ColeLemasters used ChatGPT to help make sense of the dense language in her radiology report. That conversation led him back to Kiley’s original biopsy records, where ChatGPT helped surface a crucial detail: an IDH1 mutation. In partnership with her neuro-oncologist, he then confirmed what that meant. Under updated medical standards, Kiley’s tumor was not glioblastoma, but an IDH-mutant astrocytoma: still serious, but with a very different prognosis. Instead of bracing for the worst, Kiley and her doctors began focusing her treatment plan on long-term care. Now, after years of uncertainty, she and Cole are planning for a future.

@thsottiaux hire me and i'll get the android app on par with ios in a few sprints. it's in a sad state right now.