@trikcode Did you know Codex will finish its current task even when you hit your limits?
English
Andrew Guenther
524 posts
Andrew Guenther
@codebrewed
Infra and Incidents @OpenAI. Previously @OrbitalSidekick, @AWS. I build stuff in the cloud and have strong opinions about it.
San Francisco, CA Katılım Şubat 2021
91 Takip Edilen269 Takipçiler
Got rate limited mid-task yesterday and had to switch from Fable to GPT-5.6 Terra.
6 hours of back-and-forth context entirely gone.Lost 90 minutes just getting a different model up to speed on the project goals.
Hitting a usage cap at the worst possible moment sucks.
We desperately need a universal way to share context across different models.
English
Anthropic: AI is gonna kill you and your whole family and only Dario can stop it.
OpenAI: Check out this chill grandma talking to ChatGPT.
OpenAI@OpenAI
Introducing GPT-Live, a new generation of voice models for natural human-AI interaction. Rolling out in ChatGPT starting today. You’ll want to turn the sound on for this one.
English
dario: we can’t just yolo a trillion dollars into compute
openai:
Tibo@thsottiaux
Hello. We have reached 8M active users across Codex and ChatGPT Work. We are once again resetting the usage limits for all. And we continue to not have the 5h rate limit as well, allowing everyone to explore the boundaries of GPT-5.6 Sol and discover how ambitious you can be. See you tomorrow for more updates on our growth!
English
This tweet, I'm just verified now.
Andrew Guenther@codebrewed
I don't shitpost nearly enough about my work @OpenAI
English
@codebrewed @almmaasoglu i'm so incredibly curious what your job/day to day looks like ngl
English
English
Nearly 1.5 years ago, I published a public letter about my health condition and requested expanded access to Deep Research in ChatGPT.
(At the time, we had GPT-4.5 as the best model.)
Few hours after, @farouke_ from OpenAI texted me and gave Pro access for 3 months.
Few days after, @sama reached out to me and assigned Pro plan to my ChatGPT account which took effect after Farouk's expired.
Yesterday, Pro 20x plan which I had since last year is expired and I subscribed to Plus.
If I have to be honest, this feels like a huge downgrade. It was better than a dream to use Codex and GPT models at their peak capability without thinking about usage limits.
I even got accepted to get scholarship from one of the top private universities in my country, all thanks to the projects I built with Codex.
I want to thank to everyone that offered me to run prompts, or even purchase Pro for me, when I sent the quoted post. I am really grateful that good people still exist.
From now on, I'll try to keep building with Plus, or subscribe to Pro 5x if my economy ever allows me to as a student. 😅
❤️🫡
BLCNYY@BLCNYY
Dear @OpenAI and @sama, I hope the team is proud of the success of the GPT-4.5 launch. And, I’m sure this is just the beginning, and you guys have a lot more to share with us in the near future! My name is Ömer, and I was born without kidneys. I’ve been living with a transplanted kidney since 2013, have low vision, and navigate life with several other diseases. Despite these challenges, I am deeply passionate about learning, debating, and exploring AI. I’ve been playing with the newly released Deep Research tool inside ChatGPT lately. I do also participate in debates being held in my school lately, and Deep Research helps me and inspires me to share my opinions in a more logical and well-structured way than ever. However, the current usage limit of the Deep Research tool for Plus users is quite low for me. I have a lot of questions going back and forth in my brain, like “Can we prevent the body from recognizing a transplanted organ as a foreign substance?” or “What do studies have to say about the long-term effects of infancy habits on adulthood?”. Unfortunately, I cannot afford the subscription fee for ChatGPT Pro as I’m a student and don’t have an active source of income as of now. If you grant me more usage for the Deep Research tool, I can learn even more about my diseases and maybe work on making the lives of people like me easier than they have ever been. I believe in my heart that AI can help people explore new possibilities and unlock their pure, real potential. I’m excited to do more research on this and share my findings with everyone publicly soon. Have a nice week, everyone! ❤️
English
@mronge I've never seen a company rally around a goal the way we have on the superapp experience. There's rough edges, but the vision is clear and we're all excited.
English
You can tell the new Codex app was a mandate from the execs and not the same product people that made ChatGPT and original Codex
English
Seriously... pretty tough to work with 5.6. It burns through your rates like nothing before.
Chubby♨️@kimmonismus
GPT-5.6 Sol is 54% more token-efficient for agentic coding tasks, according to Sam Altman. GPT-5.6 Sol is great. But token efficiency is the one thing I just can't seem to confirm. I've never used up my rate limits faster than I do now - even on the "Pro" tier.
English
Andrew Guenther retweetledi
FYI: 5.6 Sol medium is better than 5.5 xhigh.
As you go higher in reasoning levels on Sol, you will get insane levels of performance, but can burn through limits much more rapidly.
We’re working on communicating this better!
eric provencher@pvncher
If you're having issues with your usage limits, I really recommend turning down reasoning effort vs what you used in 5.5. 5.6 Sol Medium and 5.5 Medium are not 1:1, and 5.6 Sol Low is the default for a reason! You probably don't need to use xhigh except for very tough problems.
English
@thsottiaux Built in secret storage. I yearn to make Codex pay my bills. Also, making phone calls for me. Phone calls is a surprisingly common stopping point for non-technical tasks.
English
@stochasticarrot @jxnlco It works great. Slack connector plus computer use.
English
@ajambrosino Can't do that big dog it's a hipaa vio, insurance ting
English
install codex on your parents’ computers so you can fix stuff remotely
English
@ramimacisabird @dinodaizovi Heh, I was just talking about this today actually!
English
@dinodaizovi I always think of x.com/codebrewed/sta…
Andrew Guenther@codebrewed
@joshbressers I had some experiences with this at Amazon. I can't provide exact numbers, but imposing a one week delay on making package versions available internally combined with automated blocking of known bad versions led to a significant decrease in deployed vulnerabilities.
English
More NPM package compromises...
stepsecurity.io/blog/ctrl-tiny…
This is an intriguing idea, though, enforcing a minimum cool-down period for new package updates before allowing them to be used in your enterprise:
stepsecurity.io/blog/introduci…
English
What a way to end the week. But also massively proud of all the progress we've made over the past year.
Sam Altman@sama
chatgpt went down for 30 mins today :( we are much, much better at reliability than we used to be, but clearly more work in front of us. (it is now the 8th biggest website in the world according to similarweb--we have had a lot of work to do these past two years!)
English
My #SEV0 talk recording is live! Thanks again to the team at @incident_io for having me. Credit to my daughter for painting my nails the night before 🤣
sev0.com/sessions/maint…
English
I could be on the healthiest grocery trip of my life, but if I see these mfers in the checkout line I will fold IMMEDIATELY.
English
Advanced Voice is AWESOME. Has been an absolute delight to be testing over the last few months.
OpenAI@OpenAI
Advanced Voice is rolling out to all Plus and Team users in the ChatGPT app over the course of the week. While you’ve been patiently waiting, we’ve added Custom Instructions, Memory, five new voices, and improved accents. It can also say “Sorry I’m late” in over 50 languages.
English