In partnership with
FROM OUR PARTNER
100+ Claude Code hacks to ship code 10X faster
Top engineers at Anthropic say AI now writes 100% of their code.
Are you using AI to write yours?
These 100+ Claude Code hacks show you exactly how. Sign up for The Code and get:
100+ Claude Code hacks — free
The Code newsletter — learn the latest AI tools and skills to code faster in 5 mins a day
TODAY AI BRIEF
3 things to know about OpenAI's always-on agents
1 OpenAI launched Dots, agents that keep working after you stop.
OpenAI announced Dots on September 29, 2026 at its DevDay conference. The agents run on GPT-6 Astra, take a goal, and keep working without step-by-step direction, then message their owner with progress and questions. Read SiliconANGLE's report.
2 Each agent runs on its own cloud computer you can open at any time.
You can look in on what the agent is doing, and with your permission it can also work on your laptop. OpenAI says more than 4,000 apps are reachable through its plugins. Those are OpenAI's own claims. See the cloud computer details.
3 Access starts with Pro and Business Premium, and some regions are left out.
The rollout reaches ChatGPT Pro and Business Premium subscribers, with some accounts waiting several days. Pro subscribers in the European Economic Area, Switzerland and the UK are excluded. Enterprise customers can join a beta once an administrator turns it on. Read the availability notes.
AI FORWORD
An agent that works while you sleep is only as safe as its approval rules
The shift. The chat window was the boundary of an AI assistant. A goal-driven agent that keeps working across your apps moves that boundary, which changes what you have to supervise. Read the launch report.
The safeguards. OpenAI describes an auto-review check before any action that could touch an account or share information, plus a monitor that can pause an agent. Idle research is read-only, so it cannot send messages or change anything. See the controls.
The admission. OpenAI's only direct caveat is that the agents "can still make mistakes," and it tells users to review consequential work. The approval rules are the part you still own. Read the caveat in context.
The reframe. Treat an always-on agent like a new employee on day one. Give it a narrow job, a short list of things it may never do, and a daily summary you actually read.
PRESENTED BY :
Analytics on Live Data Without Leaving Postgres
When analytics on Postgres slows down, most teams add a second database. Then come the pipelines, the sync jobs, and a copy of your data that's always a little behind.
TimescaleDB takes a different approach: extend Postgres instead of splitting away from it. Hypertables partition your data automatically as volume grows. Hypercore compression cuts storage up to 95%. Continuous aggregates keep dashboards live without re-querying everything.
CERN runs Postgres this way for sensor data from the Large Hadron Collider.
No split architecture, no pipeline lag, no new query language to learn. Same SQL, same drivers, same tools.
Start on Tiger Cloud and get $1000 in credits.
SOCIAL SNAPSHOT
What to check before you let an agent run unattended
One agent included. Pro and Business Premium include one agent at no extra cost, and agent usage is not counted against allowances for the first month, per OpenAI's help documentation. Terms apply after that. Read the pricing notes.
Identity for specialist agents. Employers can give a specialist agent its own identity and credentials for one job. Procurement and invoice processing were tested internally, according to the report. Read the specialist note.
Read-only when idle. While it has nothing to do, the research mode cannot send messages or change anything. Check that the mode matches what you expect before you rely on it. Read the controls.
Old setups still fail. A community guide on self-run agents warns that missing credentials can fail silently and that in-memory vector stores lose data on restart. Read the reliability notes.
PROMPT STATION
Write standing orders for an agent before you switch it on
An always-on agent acts on what you wrote before it started, not on what you meant. This prompt works in Claude, ChatGPT or Gemini and turns a vague job into standing orders you can paste into any agent's first message.
I want an always-on agent to handle this job: [DESCRIBE THE JOB IN TWO SENTENCES]. The accounts it may use are [LIST THEM].
Act as a careful operations manager. Write standing orders in four parts: what it may do without asking, what it must ask me about first, what it must never do, and the one daily summary it sends me, with its length and contents.
Finish with the three situations most likely to go wrong for this job, and what the agent should do in each one.Tip: Swap the brackets for a real job, such as chasing unpaid invoices or sorting support emails. Keep the never-do list short and specific, and check the first daily summary against what the agent actually did.