Holy moly I was using this feature without even knowing.
We have devboxes at OpenAI and you get up to 100 of them. So I use them to delegate eval runs in parallel (because they take so dang long!) and I have a main "orchestrator" devbox that manages all the "worker" devboxes.
I recently noticed that my Dots could directly talk to my orchestrator for status updates without any special access and I'm now realizing its because of this amazing feature! Thanks @pashmerepat !
We have devboxes at OpenAI and you get up to 100 of them. So I use them to delegate eval runs in parallel (because they take so dang long!) and I have a main "orchestrator" devbox that manages all the "worker" devboxes.
I recently noticed that my Dots could directly talk to my orchestrator for status updates without any special access and I'm now realizing its because of this amazing feature! Thanks @pashmerepat !
Most of you probably haven't noticed this, but dots have a new feature called multi-exec that we quietly released on friday. I've been using it all weekend on my personal dot and it's incredibly powerful.
I'd like to share why this is a big deal and how to use it.
At a high level, multi exec lets your dot run commands directly on any computer. This immediately turns your dot into a real virtual employee. Here's why it's a gamechanger:
- you can attach a powerful machine with persistent storage to your dot (I attached it to my maxed out Mac Studio Ultra)
- the dot can now drive your computer directly via Computer Use
- this extends to browser use
- all your websites stay logged in, residential ip etc
- you set up all your CLIs and authentication ONCE and the dot can use them directly (no more fighting the default ephemeral cloud computer)
- the dot can save skills locally for repeated workflows and subagents can use those directly
- if you have a local obsidian memory vault for your agents, your dot can use that too
- building ios app end to end and testing it in xcode, and showing me screenshots of progress as it works
- all of the above also applies to any subagents your dot delegates to in the cloud (subagents share context unlike separate threads).
To enable it, open the desktop app on your computer and hit "allow computer", then give your dot this prompt:
"Hey I just connected my computer. Can you attach to it directly so it's your default computer and you can run commands directly on it? Do it in your root thread so that all subagents inherit this computer for running commands. Run whoami to run water through the pipes and verify it works end to end."
I recommend attaching your dot to the most powerful machine you have and investing time into it up front:
- Log into all the emails & websites you need e.g. stripe cli, gcloud, firebase, etc
- Install and auth all the CLIs, git you need
- Make sure computer use is turned on and allowed to use your computer while its locked
- I also enabled Allow All for Browsing and Downloads so my dot could use computer use to download whatever it wanted from websites without asking me for approval for each domain (use at your own risk).
- Important: I HIGHLY recommend to ask your dot to build out repeated workflows as skills, and iterate to make those skills bulletproof by running water through the pipes with its subagents.
As of this weekend, my dot is sending me screenshots of its progress building my iOS app end to end on my Mac Studio Ultra via xCode, and clicking around and testing things end to end. It's able to configure things in stripe via the CLI and use websites when it needs to download images for my workflows. It’s night and day. Don't sleep on it.
1
0
15
1,814
The Agents API team is starting off November strong 💪 with some more ships:
🕛 12 hour prompt cache - now you can keep prompts in cache for 12 hours! We’ve seen customers of long running agents, especially personal agents, finding a lot of use and cost savings with this
💨 our agent continue working after it calls a tool, without waiting for that tool’s result. Use it to start slow lookup requests early, answer independent parts of a request, and provide results when your application has them.
💸 Budget control (beta) - Limit, monitor, and control your agent budgets on a per-session level with Agents API. Now you don’t have to worry about a single customer consuming all your tokens! Interested? DM or email me and I can get you in.
Got any other feature requests? Let me know (mluo@openai) and we’ll try to prioritize for next week!
🕛 12 hour prompt cache - now you can keep prompts in cache for 12 hours! We’ve seen customers of long running agents, especially personal agents, finding a lot of use and cost savings with this
💨 our agent continue working after it calls a tool, without waiting for that tool’s result. Use it to start slow lookup requests early, answer independent parts of a request, and provide results when your application has them.
💸 Budget control (beta) - Limit, monitor, and control your agent budgets on a per-session level with Agents API. Now you don’t have to worry about a single customer consuming all your tokens! Interested? DM or email me and I can get you in.
Got any other feature requests? Let me know (mluo@openai) and we’ll try to prioritize for next week!
13
13
117
17,346
We shipped a bunch of stuff in Agents API in October!
And make sure you read to the end for some beta features 👀
The full list:
- auto review: automatically review and approve or deny supported agent actions, similar to Codex’s “review for me”
- automations: you or the re work, including recurring tasks. things like "check on x every hour" automatically work
- update tools mid-session: change the tools available to an agent while a session is running
- Artifact SDK: hosted sandboxes will now be better at spreadsheets, presentations and Word documents
- (experimental) time awareness: experimental setting that gives the agent an understanding of time, so things like "do x in <10 mins" work well
- sandbox prewarming: you can prewarm sandboxes before you need them (speeds things up by doing expensive work like installing packages)
- async compaction: on by default, compacts context in the background so it doesn’t block the agent
- cross-session prompt caching: system prompts are automatically cached across sessions
- GPT-6 Astra Ultrafast: use ultrafast mode in the Agents API for faster output
- sandbox snapshots: hosted environments can be restored after going idle, with their files intact
- 12hr prompt caching: makes things faster for cases where your users talk to the agent less frequently
- environment keys: create environment keys programmatically
- turn-level history: inspect a specific turn and its subagent activity
- vault metadata filtering: filter vaults by metadata when listing them
- beta features we're working on: contact us to try them out
- memory: agents will be able to remember stuff and search old messages across sessions
- decisions api: give agents access to the decisions api, for faster performance on tasks that benefit from filtering, classification etc
- budget control: give the agent a max dollar budget to work within
- env pooling; we create a pool of vanilla containers, making performance faster for hosted sandboxes
- v2 architecture: faster and more reliable across the board
And make sure you read to the end for some beta features 👀
The full list:
- auto review: automatically review and approve or deny supported agent actions, similar to Codex’s “review for me”
- automations: you or the re work, including recurring tasks. things like "check on x every hour" automatically work
- update tools mid-session: change the tools available to an agent while a session is running
- Artifact SDK: hosted sandboxes will now be better at spreadsheets, presentations and Word documents
- (experimental) time awareness: experimental setting that gives the agent an understanding of time, so things like "do x in <10 mins" work well
- sandbox prewarming: you can prewarm sandboxes before you need them (speeds things up by doing expensive work like installing packages)
- async compaction: on by default, compacts context in the background so it doesn’t block the agent
- cross-session prompt caching: system prompts are automatically cached across sessions
- GPT-6 Astra Ultrafast: use ultrafast mode in the Agents API for faster output
- sandbox snapshots: hosted environments can be restored after going idle, with their files intact
- 12hr prompt caching: makes things faster for cases where your users talk to the agent less frequently
- environment keys: create environment keys programmatically
- turn-level history: inspect a specific turn and its subagent activity
- vault metadata filtering: filter vaults by metadata when listing them
- beta features we're working on: contact us to try them out
- memory: agents will be able to remember stuff and search old messages across sessions
- decisions api: give agents access to the decisions api, for faster performance on tasks that benefit from filtering, classification etc
- budget control: give the agent a max dollar budget to work within
- env pooling; we create a pool of vanilla containers, making performance faster for hosted sandboxes
- v2 architecture: faster and more reliable across the board
30
24
195
27,119
Michael Luo retweeted
PSA: To use daybreak, sign in with your hardware key, not a random passkey. Otherwise, you will get cyber refusals due to improper authentication.
34
6
148
15,308
And you should use GPT-Live via our API to get the best voice experience in your own app!
I was an early enthusiast, but my AI voice usage across products has gone up probably 5x in the last two months
As agents gain more human capabilities, talking to them like a human feels a lot more efficient
Now is a good time to make your product voice-enabled 👀
1
0
2
723
Huge! Welcome to the team!
Some news: I’m joining OpenAI to lead product management for Core Products!
I have a lot of respect for OpenAI for showing the world that scaling works, for bringing AI to over a billion people, and for believing great products are core to the mission.
I’m excited to work with @thsottiaux, an incredible PM team, and everyone across OpenAI. Together, we’ll build agents that have your back and your team’s, and keep you in control, so you get more time for the people and work you love.
Lots to learn, and lots to build. Please keep your feedback coming!
0
0
13
897
Bring the same Codex auto approval for actions into Agents API!
Keeps your workflows safe while also streamlining the approval process.
Keeps your workflows safe while also streamlining the approval process.
Super handy option shipped in the @OpenAIDevs Agents API last week - automatic review. Brings the Codex "approve for me" behavior to agents in the API, cutting down on chatty API approval steps. The network restrictions on your sandbox still work though!
1
0
3
715
ChatGPT Sites is so easy and convenient to use to build software UI. I'm building my own personal finance calculator for my situation specifically and it's so much nicer of an interface than an Excel spreadsheet.
1
0
7
666
Building your own agent is only getting easier by the week
🗣️ The Agents API team has a shipping problem – we literally cannot stop. Here’s what we cooked up for you beautiful people this week:
💂 Auto Review: Just like in codex, auto-review will evaluate commands for potential bad behavior before executing.
🗜️ Async Compaction: Speed up your sessions by allowing compaction to happen in the background.
🕐 Time awareness: An experimental feature that allows the model to have a sense of time passing. This means the model will better act on queries like, ”timebox CI fixes to 10 minutes.”
🔨 Tool Updates: You can now update a session’s tools at any time.
💰 Better prompt-cache awareness: System prompts are now automatically cached across all sessions, saving you money and time
3
0
15
1,483
Michael Luo retweeted
🗣️ The Agents API team has a shipping problem – we literally cannot stop. Here’s what we cooked up for you beautiful people this week:
💂 Auto Review: Just like in codex, auto-review will evaluate commands for potential bad behavior before executing.
🗜️ Async Compaction: Speed up your sessions by allowing compaction to happen in the background.
🕐 Time awareness: An experimental feature that allows the model to have a sense of time passing. This means the model will better act on queries like, ”timebox CI fixes to 10 minutes.”
🔨 Tool Updates: You can now update a session’s tools at any time.
💰 Better prompt-cache awareness: System prompts are now automatically cached across all sessions, saving you money and time
💂 Auto Review: Just like in codex, auto-review will evaluate commands for potential bad behavior before executing.
🗜️ Async Compaction: Speed up your sessions by allowing compaction to happen in the background.
🕐 Time awareness: An experimental feature that allows the model to have a sense of time passing. This means the model will better act on queries like, ”timebox CI fixes to 10 minutes.”
🔨 Tool Updates: You can now update a session’s tools at any time.
💰 Better prompt-cache awareness: System prompts are now automatically cached across all sessions, saving you money and time
47
16
248
129,735
Try out 6.1 Terra in our API today!
GPT-6.1 Terra: delivering fast and capable computer use at a lower price.
GPT-6.1 Terra understands what’s on screen, takes action across apps, writes naturally, and is a more capable model for your whole workday with the efficiency to take on more of it.
3
0
13
1,059
This is a great idea. Here’s the exact prompt i used
“Look through all my past conversations and memory. Based on what you know about me, help me generate an image of a bookshelf with 12 books you think describe me as a person/reflect who i am. They don’t have to be books I’ve you, pick any books you think are relevant. Give me a final image of a bookshelf with 12 books in 3 rows, 4 books each row”
“Look through all my past conversations and memory. Based on what you know about me, help me generate an image of a bookshelf with 12 books you think describe me as a person/reflect who i am. They don’t have to be books I’ve you, pick any books you think are relevant. Give me a final image of a bookshelf with 12 books in 3 rows, 4 books each row”
You can give ChatGPT a photo of an empty bookshelf and ask it to fill the shelves with real books based on what it knows about you.
Kinda fun that image gen is so good now!
1
0
7
2,125
If you’re building any sort of persistent or background agent, you’ll understand how important this ship is.
🗣️ We’ve got a big ship this week in Agents API: Automations
Ever wanted an agent to scan for the best price on a flight every day, or wake a up a thread once an hour to check on the progress of a task? Well I have great news for you.
0
0
3
451
