@kimmonismus Almost certain it was with subagent orchestration, which has been addressed to some degree.
My burn rate looks much more similar to 5.5 now, yesterday I used the 5 hour limit in 37 minutes.
English
Walmart McKenna
1.1K posts
I read this as an admission that something is fundamentally wrong with GPT-5.6, and that the claim that GPT-5.6 uses significantly more tokens than GPT-5.5 is not based on purely anecdotal evidence.
"changes that will make GPT 5.6 Sol more efficient across the board and that will be reflected in less usage"
Don't get me wrong. It’s not fundamentally flawed in the sense of being a bad model, but rather fundamentally flawed in terms of reasoning or token usage.
Tibo@thsottiaux
Morning. The last 48 hours of Codex and ChatGPT Work have been intense! Three important updates: - Temporarily removing the 5 hour usage limit restriction for all Plus, Business and Pro plans - Rolling out changes that will make GPT 5.6 Sol more efficient across the board and that will be reflected in less usage being used so that it can take you further. Exact impact to be quantified and shared - We hit 6M active users, and are landing a usage reset in the next hour Go do things
English
@thsottiaux Whatever yall did in the update this morning made thread/subagent orchestration night/day difference in token usage in my workflows.
English
Morning. The last 48 hours of Codex and ChatGPT Work have been intense! Three important updates:
- Temporarily removing the 5 hour usage limit restriction for all Plus, Business and Pro plans
- Rolling out changes that will make GPT 5.6 Sol more efficient across the board and that will be reflected in less usage being used so that it can take you further. Exact impact to be quantified and shared
- We hit 6M active users, and are landing a usage reset in the next hour
Go do things
English
@thsottiaux Idea 💡
I would opt in to share detailed telemetry data if there were some interface for me to see as well.
Could maybe get a 10-20% increased limit for users that let yall monitor usage for diagnostics or something?
What’s measured gets managed.
English
@thsottiaux Would love some data/transparency regarding limits.
Folks internal to oAI keep talking about token efficiency, but run a /last30days and see what the consensus is regarding token efficiency.
5.6 Sol on High achieves burn rates that were literally impossible on 5.5xhigh
English
@thsottiaux @rafaelobitten FWIW from 1 user:
1 task running on a persistent goal with sol high used an average of 25% of 5hr limit / hour.
If told to allow child tasks (made 4) consumption went to 73% of 5 hr limit in 40 min
Data point for the bot that reads the replies.
English
@PromptLLM It is no longer a standalone product, but the work/research/tech is not being discontinued.
It’s persisting as a key feature in a wildly successful product.
I’m certain that you could understand this.
English
Biggest flop of all time
Everything Altman and OpenAI have realised has been a failure apart from one thing. And that one thing took them to the moon
Lesson in there
Polymarket@Polymarket
JUST IN: OpenAI to reportedly discontinue ChatGPT Atlas on August 9, less than 10 months after launching the standalone browser.
English
@thsottiaux When trying to copy thread id or deeplink from sidebar in Codex App on Mac, it instead pastes the last user prompt.
English
Hello beautiful people! We have reset usage limits across Codex and ChatGPT Work. And another one will come later in the day. Rejoice.
Now that I have your attention, a quick update on ChatGPT Work, Codex and all the updates we shared yesterday.
We’ve spent the last 24 hours reading feedback, looking at usage patterns, and talking with many of you. The short version is that there is a *lot* of excitement for GPT 5.6 Sol, ChatGPT Work on mobile & web, but also that we didn't get everything quite right.
- We made it too easy to use the highest-compute settings without making the impact on usage limits sufficiently clear.
- We reorganized the desktop app in one bold move, making familiar things like chats and projects harder to find.
- Our launch framing was focused on ChatGPT Work and to some of our Codex fans it made it feel like Codex was going away over time. Absolutely not our intention, we love Codex and it is here to stay.
- And we introduced regressions for some existing multi-agent workflows, alongside a collection of rough edges in plugins and other parts of the experience.
We’re landing a first set of improvements today. We’re resetting usage twice so people can keep experimenting, changing defaults and the model picker so they don’t push people toward unnecessarily expensive settings, fixing several plugin submission issues, improving how we represent Codex in the product, and cleaning up some of the most immediate desktop problems.
A larger set of improvements will land next week. We’re bringing chats and projects back into the sidebar in a more familiar and customizable way, making usage and reset timing much more visible, clarifying when to use ChatGPT Work and when to use Codex, and addressing the many other smaller pieces of great feedback we've had.
The ambition behind this launch hasn’t changed. We think bringing ChatGPT and Codex together into a workspace where people and agents can collaborate is a very important step forward. But an ambitious direction doesn’t excuse avoidable confusion or regressions in the first version.
Please keep the feedback coming. We’re moving quickly, and you should see the experience already get better with a few updates today; and substantially better again next week.
English
@DavidOndrej1 1 thread of 5.6 Sol on high running on a goal uses about 12% of 5 hour and 5% of weekly limit per hour.
Others mileage may vary but I am actively running the test and it’s been consistent for 2.5 hours and seemed to track before the resets messed up my data.
English
@DaveShapi How is it fundamentally different to the wisdom of the farmer?
Is it not conceivable that a more optimal strategy is to treat the environment as a collaborative partner, and let the environment provide for the individual, even if there isn’t a clear deterministic path for reward
English
All Effective Altruism, LW, and associated conversations have an infinite regress problem, and they always converge on anti-human conclusions.
Here's how it works, with their own words.
1) Amanda Askell talks about "rational resentment" i.e. "the future machines might hate us for how we treat them today, so we should preemptively grant concessions and be nice" which is Roko's Basilisk dressed up in philosophical mumbo jumbo bullshit.
2) More and more conversation about "worthy successors" and "passing the torch" either because humans are "unworthy" or "will be judged" or because it's just the right thing to do. Sometimes it's also dressed up as "the only way to survive is by passing the torch" (surrendering sovereignty preemptively).
3) That we may "owe moral obligations" to the things we are building (which does not follow, unless you explicitly and deliberately build in a way that it can suffer... which is itself wrong but even if you make something that can pretend to suffer does not make it so).
So, now when you look at the lunacy that is AI 2027 and their equally delusional follow-up AI 2040, it all regresses to "The AI will kill us, cause us to kill ourselves, and the only path to salvation is prostrating ourselves before the AI, or otherwise it's a suicide pact."
This is all purely eschatological language.
And these people are actively power seeking.
English
@thsottiaux @scaling01 Still resisting heliocentricism?
Try GPT-5.6 Flat Terra
English
I think GPT-5.6-Terra should be the name of the strongest model, not Sol
Terra is where everything began
Terra is the center of everything
this is literally warhammer and ASI is our emperor of mankind
English
@thsottiaux If the vision is “ChatGPT” is the mobile/light interface (esp when voice gets connectors) and “Codex” as the desktop/heavy interface, merging to create shared context within the ecosystem is more generational run than fumble IMO.
Execution will improve, the vision is there.
English
Or a generational run. We will continue to make tweaks and we have a TON coming for developers. Meanwhile traffic yesterday sat at ~2X previous peak, team is scrambling to keep everything up and reliable.
Time will tell, but I have conviction that attempting to build the one interface to AGI is worthwhile.
Theo - t3.gg@theo
fwiw, I think that turning Codex into ChatGPT Desktop is a generational fumble
English
@GwatneyBrian @jeremyct Yes, what a stupid 10 year old I was in 2003 for not buying a house.
English
@jeremyct Sounds like you wasn't smart enough to buy it and do the same thing. You need to understand this.
English
I need you to understand something.
Your landlord didn’t build wealth because he worked harder than you.
He bought a asset in 2003 that appreciated while he slept.
You paid for that appreciation every single month.
And he’s going to sell it to someone else and call himself a self made man.
English
Im just a rando with no inside knowledge but IMO:
Anthropic is ahead on pretraining which leads to the “genius” quality and is better at modeling how humans think, which helps infer intent.
OpenAI is the leader in RL and post training which has a more utilitarian reward incentive that manifests as the well-behaved worker bee quality.
Arguments could be made about the comparative utility of either approach, but it points towards a world that may not be winner take all. Tool / task fit may be a big driver.
English
i'm a bit disappointed to hear that gpt 5.6 is not as good at understanding the unsaid intent behind the prompt
will see what i can do with a really good fablemaxxed AGENTS.md
i'm hoping i'll be able to steermaxx it successfully 🤞🏻
gnight, tomorrow is a big one 💤
Ben Davis@davis7
On Fable being smarter than 5.6 - it is, but: I think smart is the wrong word, the thing 5.6 is lacking behind fable is the "big model smell" which roughly translates to "understanding the unsaid intent behind your prompt" fable fills in the blanks better, it writes more beautiful code, it suggests the next thing the way you would or better 5.6 is a blunt instrument. It does what you say, and it will damn well make sure it does it no matter what. 1000+ tests, 20 nested try catches, catching every single edge case possible through rigorous end to end testing of the program itself. It gets the job done, just not in the cleanest way unless it's well directed the biggest thing it shares is that ability to run for absurd amounts of time with 0 issue (had threads run for 20+ hours spamming compaction and dozens of subagents and it ran flawlessly, could even wildly steer it halfway through and it just kinda worked) and orchestrate itself beautifully take everything good about 5.5, and make it better while fixing it's most glaring issues
English
@dj45williams Matt Jones
Alex Collins
Treylon Burks
Chris Gragg
English
English
@AlexiLalas Right now nobody is, youth sports are a vehicle for people to make money.
You’re right about the problem question, but it’s the same question for everyone fundamentally.
Our answers to that question and the execution is not as good as the people that we are competing with.
English
Youth soccer (youth sports) is a competitive market with businesses selling a product that obviously customers are willing to pay for. I’d love if soccer was free to all. But who is going to pay for all this free soccer?
Clay Travis@ClayTravis
One big US soccer thought: our youth development has to get more affordable. One of my kids plays travel. It’s at least $5k a year & this isn’t even super high end. We can afford it, thankfully, lots can’t. Youth sports costs in the US are out of control, soccer more than most.
English
@argofowl Bold strategy milking the anticipation for all it’s worth. Maybe the marketing folks have determined it’s better, but it reads to me as disorganization.
Lots of hype in the build up to draw attention and then deafening silence when everyone starts looking.
English
@pinto479 @cfryenewsguy I wouldn’t be surprised if Golf wasn’t more popular by viewership and revenue to be honest.
English
The USA continues to be massively behind Europe in soccer and it showed tonight. USA Soccer has to do a better job with development of its youth program. ECNL can't do it. Neither will MLS Next. USA Soccer must take control of this situation. This was disappointing to watch.
English
@filthyt0m @KasperStats To be honest your commentators too often are verbose and over analyse things.
Sometimes the action speaks for itself
English
In general, in all American sports, we value: grit and toughness, clear and fair rules, and good sportsmanship.
The flopping and embellishing part of soccer goes against all 3 of those. The draconian nature of red cards, and the arbitrary nature of officiating, contradicts #2.
Michael Kasper@KasperStats
But this is actually a great example of the obstacles for Americans to get into soccer. We're used to fine-grained, proportional penalization systems that feel "fair." Soccer basically has "free kick, flagrant 1, flagrant 2 + next match suspension" as the only options.
English