ldraw-nova: AI agents build LEGO models by writing Python geometry
In ldraw-nova, models write a plan and a Python generator instead of raw coordinates, but nothing checks physics or stability.
OpenAI Dots agents: revoking app access doesn't erase their memory
Disconnecting an app from an OpenAI Dots agent stops new reads, but pulled-in context stays, so revoking access and forgetting need separate steps.
Huseyin Kaplan · 1 min read
You can disconnect OpenAI's new always-on agents from an app in Settings. What they already learned from it stays.
Dots launched yesterday: agents with their own cloud computer, connected to your apps, working while you're away.
The safety write-up is more candid than the launch page. Disconnecting an app stops new reads. Anything the dot already pulled into its context remains until you reset that context.
Two more lines worth copying if you build agents:
- Background research runs on read-only tools, with the limits enforced in code.
- The action check runs outside the environment the agent controls, so it can't switch it off.
If your agent keeps memory, revoking access takes three steps:
- Remove the connection.
- Purge or tag what came through it.
- Write a test that proves step 2 ran.
That's the part I'd check before connecting a work inbox to any agent.
Revoking access and forgetting are two different features.
