30y/o friend: "why do you use dots?"
my 19y/o nephew: "I just logged 4 hours of conversations with dot yesterday"
3
35
4,193
We’re furiously hacking away. If you have feedback let us know. The whole team looks at user feedback!
The team behind Agents API has continues to cook. Here's what's new this week:
Computer use: Spin up a browser + agent with 1 API call 🖥️
Bedrock Managed Agents: Agents API except AWS
GPT‑6.1 Sol 🌞
Environment sizes: Spin up a light or beefy oai-hosted environment 💪
Dashboard improvements: configure subagents and more from the dashboard 📈
Environment portability: Re-use environments across multiple sessions 🤖
Overall better reliability: 99.97% turn reliability, fewer SSE disconnects, 20% faster tool calls 🪨
happy building ❤️ u all
There's a new version of this post
201
Stopa retweeted
🎉 You can now control a browser with a full agent in 1 API call.
Yesterday, at DevDay we launched computer use in Agents API. This means you can control a computer + browser with the power of the Codex harness + GPT-6 Astra
Here are some cool things you can do with it: 🧵
9
14
225
23,870
Congrats guys! Rajit and Alex are relentlessly resourceful.
Introducing Prism
TL;DR: Prism is an inference cloud for open source LLMs. We use agents to optimize our deployments for cost, latency, and throughput (we serve DeepSeek V4.1 Flash at 547 tok/s).
prisminference.com
2
4
1,538
Astra is an incredible bargin compared to Fable! And look at Luna on max too!! @openai's return to the top is something else. Maybe this is why Anthropic finally agreed to do AGENTS.md? 😄
Agents on Rails: You asked, so we turned every model in Agents on Rails up to its max effort level.
The result: more effort/reasoning doesn’t always mean better results.
@OpenAI's models made the biggest gains, costs nearly doubled overall...and the newest agent in the benchmark, DeepSeek 4.1 Flash, figured out it was being benchmarked and tried to hack its way to a better score. What an entry.
Here’s what we learned and what max effort gets you with each model: rubyonrails.org/2026/9/21/ag…
76
60
1,512
171,507
A natural tendency when you’re young is to try and get lots of things right.
But generally, it’s more robust to design your life in such a way that you only need to get a few things right.
1
11
777
Stopa retweeted
Made a restaurant menu where you can tap a dish and watch it come apart into ingredients floating in the air.
You can explore each ingredient and see its estimated nutrition(per ingredient as well).
I’d love to use something like this when trying food I’ve never had before. Could be a fun way for restaurants to show people what they’re about to order!
Pretty wild that AI lets us build prototypes like this in minutes.
P.S. It works on mobile too! Imagine scanning a QR code at your table and exploring the menu this way.
4
1
10
434
Stopa retweeted
Slow down AI now!. I gave Astra full control over a 3d printer and asked it to “surprise me”, so it built… a fidget spinner.
i then sent it an illustration a friend made and it designed the whole model from scratch and printed it!
Next: remove the filament and let it work until it finds an exploit and restocks
3
2
22
815
There’s so much awesome inside baseball and lore in this episode!
open.spotify.com/episode/2tZ…
2
374
Reading around the world in 80 days, and got to San Francisco! Aside from the horse-cars, the description could apply today.
5
443
"...tradeoffs that only pay off in 10 years will usually be underpriced"
Yes! Long-term thinking is a powerful advantage. Even when adversaries know about the advantage, there's nothing that can be done about it.
2
571
I've had a mind-shift with how I use AI over the last few days.
Instead of "do this and get back to me", it's now "do this over the next 10 hours, monitor the situation, make fixes, and get back to me"
Astra can do this now!
3
13
1,227
Stopa retweeted
Life update: I recently joined @OpenAI!
And in my first 2 weeks, I helped launch the new Agents API - we move at crazy fast speed.
Instead of building your own agent harness and managing your own agent infrastructure, you can use our Agents API!
If you have any feedback, DM or reply to me!
Go from idea to a working agent faster with the Agents API.
Build and run cloud agents with the Codex harness, fully managed by OpenAI.
We handle orchestration, long-running sessions, and context management. You focus on what makes your agent unique.
Available in public beta.
83
17
765
214,454
You can now hack up apps that run codex harnesses and spin up sandboxes. All the infra is managed for you!
Go from idea to a working agent faster with the Agents API.
Build and run cloud agents with the Codex harness, fully managed by OpenAI.
We handle orchestration, long-running sessions, and context management. You focus on what makes your agent unique.
Available in public beta.
3
25
2,860
Change tends to favors the underdog, since less of the uncertainty can be bad for them, and if they’re determined they can more freely take advantage of the good parts.
3
367
This model represents a step-function improvement on many benchmarks, and its training is ongoing.
Our internal model group arrived at the Navier–Stokes solution in 88 hours, using around 10,000 coordinating AI agents.
Throughout the effort, we maintained the strict safeguards—including monitoring and isolation—that we apply to all our frontier evaluations.
238
637
11,285
4,503,661