// verdict · checked 2026-08-11 · Jules GA (docs and changelog as of Aug 2026)
Jules
Google's async coder. Real loops, real tools, Google-real product lifespan TBD.
- CAPABILITY
- 5/5 AGENT
- PROOF
- OPERATOR INSPECTABLEYou can run it and read the steps yourself — the record exists for whoever runs it
- TRACE ACCESS
- OPERATOR-OWNEDWhoever runs it holds the record; nothing is published
- LEASH
- Not measuredno execution record scored — measure it yourself
| LOOPSruns until doneyes | Docs: the VM lets it "clone your repository, install dependencies, and run tests" and work its plan to done. |
| CHOOSESpicks its own toolsyes | Jules writes its own plan, then picks tools in the VM (shell, pytest/npm, Playwright, web search). |
| ACTSchanges the worldyes | Docs: in the VM it "clones your code, installs dependencies, and modifies files"; you click to push a branch. |
| RECOVERSfixes its own errorsyes | FAQ: on a failed step "Jules will retry automatically", marking the task failed only after it keeps failing. |
| UNSUPERVISEDnobody watchingyes | Docs: leave the page and the plan auto-approves on a timer — "no need to babysit"; the API auto-approves too. |
Capability decisions combine documented behaviour and execution evidence. The proof label on each row shows which of the two decided it, and Leash is reported separately, only when it was measured from an execution record. Primary source: jules.google. Tested against: Jules GA (docs and changelog as of Aug 2026). Our confidence in this dossier: 8/10.
Can you see it work?
the weakest of the five grades aboveOPERATOR INSPECTABLE — You can run it and read the steps yourself — the record exists for whoever runs it. Trace access: OPERATOR-OWNED — Whoever runs it holds the record; nothing is published.
activity feed shows each step and per-file diffs; no export documented
Where we looked: jules.google/docs and /docs/running-tasks/: plan approval, 'an activity feed as each step completes', inline per-file diffs. No export documented. · see the evidence →
- 2026-08-114/5 AGENT → 5/5 AGENTUNSUPERVISED corrected to yes: No per-action confirmation: web plan gate auto-approves on a timer, and API requirePlanApproval defaults to auto-approved.editorial correction — no case was filed
Badges
earned, never for sale
These render from the live dataset, so a badge cannot outlive the verdict it claims. If the verdict changes the badge changes with it, and an unearned one returns 409 rather than an image. No sponsor can buy one, at any price.
Think this is wrong?
bring evidence, not opinionName the criterion and link a trace, a doc, or a recorded run that shows the behaviour. A verdict changed by evidence is the best thing that can happen to this site, and the change goes on the public record with your reason attached. Vendors are welcome, and a vendor's own filing is labelled as such rather than buried.