See it in action.
From a useful question to your next great idea. Three short films about keeping your agents’ work precise, reviewable and ready to continue.
These are illustrated explainers with synthetic narration. They explain the workflow using conceptual examples. Local agent workflows take place in your own authorized sessions; the public demo does not run agents.
Precise feedback · 49 seconds
Point to exactly what needs attention.
Turn a vague concern into a useful question attached to the right passage. Select, comment, and find your feedback again.
Try this idea with supplied demo samples.
Read the transcript
Great AI output still deserves a good question. In this fictional research brief, a growth claim needs evidence.
Select the exact passage you want to discuss. Keep your feedback attached to the words that prompted it.
Ask which source supports weekly growth, choose a priority, and save the comment. Make the next step clear.
Open the conversation and use Show passage to return to the text. Its stable comment ID identifies the feedback.
Within the same demo session, return after a reload and find the conversation again. The file's open-thread count helps you navigate.
Try it on a supplied sample. Comments are private to your temporary session and stored on the demo server. Keep sensitive details out.
Download text transcriptComment-ID handoff · 50 seconds
Give your agent a focused next step.
Carry precise feedback into your external agent workflow, then return to review the result. The human keeps the decision.
A preview of the forthcoming local workflow.
Read the transcript
Precise feedback becomes more useful when it travels with the work. Start with one comment in your local project.
Give your existing Codex or Claude session its ID and a bounded instruction: check the evidence, revise this section, and explain the change.
The authorized agent retrieves the feedback and reads the relevant source. A browser comment alone does not start an agent.
The file is revised outside the browser, and the agent replies to the same conversation. The question stays connected to the response.
Back in Review with AI, inspect the changed passage and explanation. A reply, or a repaired attachment, is not your acceptance.
You decide what is acceptable and what still needs work. That's a focused local workflow, with human review at the center.
Download text transcriptProject memory · 49 seconds
Start a new chat without losing the thinking.
Keep findings, decisions and open questions in linked files. Give the next session a starting point you can inspect and improve.
Structured files and deliberate retrieval, beyond one chat.
Read the transcript
New chat. Same project. Do you really want to explain every decision again? Keep the useful thinking outside the conversation.
Save findings, decisions and open questions as structured project files. Link each important choice to its evidence and rationale.
A decision can point to a research finding and an unresolved question. Those connections preserve what you know, and what you don't.
In a fresh agent session, point explicitly to the project entry file and relevant comment IDs. Retrieve the context that matters.
This is deliberate retrieval, not automatic recall or unlimited context. Review the sources, update stale decisions, and keep uncertainty visible.
Your conversation can end without taking the reasoning with it. Build a project memory you can inspect, improve, and carry forward.
Download text transcriptNow make it your own
Start with one useful question.
Try the review experience on a supplied sample, or explore how the forthcoming local release fits alongside Codex and Claude.
Demo comments are temporary and stored on the demo server. Keep private details out. Content you share with external agents follows their providers’ settings.