Field noteOct 21, 20262 min read1 source

Agents, rebuilt to do more for less

In one minute

Notion’s announcement changes both the runtime envelope for its agents and the way some users can bring model access into the product.

The company says agents can now work on a single task for more than 24 hours, up from five minutes. It also says ChatGPT Plus, Pro, and Go plans can be connected to personal Notion Agent, while Custom Agents are not included in this initial connection.

The implication is that runtime persistence and model access are becoming distinct product decisions: longer execution does not itself establish that an agent reliably finishes useful work. Notion’s announcement also reports 30% lower Custom Agent run costs in testing and more than 99.9% infrastructure uptime; these are company-reported figures, not an independent evaluation.

Thesis

For production agents, the harness is becoming a more durable product boundary than the identity of the model running inside it.

Observed facts

  • Notion says agents can now work on one task for more than 24 hours, compared with five minutes previously.

    1
  • Notion says ChatGPT Plus, Pro, and Go plans can connect to personal Notion Agent; Custom Agents are excluded for now.

    1
  • Notion describes a dedicated service with a queue-backed worker loop designed to continue work through restarts and hiccups.

    1

Mechanism

Notion attributes longer sessions to a dedicated service with a queue-backed worker loop that continues through restarts and hiccups, alongside context compaction and offloading large tool results. The access connection is separately scoped to personal Agent, rather than all agent surfaces.

Why now

Notion says it has added seven models in the past month and presents these changes as a response to users treating the workspace as infrastructure for their own agents.

Implications

  1. The implication is that runtime persistence and model access are becoming distinct product decisions: longer execution does not itself establish that an agent reliably finishes useful work.

  2. Longer execution does not itself establish that an agent reliably finishes useful work.

Evidence boundary

Evidence boundary: The source is Notion’s own product announcement, and it does not provide independent measurements or a full account of task success.

Watch conditions

  1. What to watch: Whether Notion extends the ChatGPT connection to Custom Agents and publishes evidence on completion quality for long-running tasks.

Evidence / sources

  1. 1notion.com/blog/notion-ai-more-for-less
All Field Notes

Next reads

Read next

Open the graph
  1. 01topicAgentsExplore the agents topic hub.
  2. 02topicAgent RuntimeExplore the agent-runtime topic hub.
  3. 03topicModelsExplore the models topic hub.
  4. 04archiveField NotesOpen the latest editorial analysis.
  5. 05source ledgerSource LedgerInspect the public source evidence graph.
Who read this page?Machine requests, hidden until opened

Loading the privacy-safe route aggregate...

Open the JSON contract