Give a dot one changing responsibility, if you already have eligible access. Try a read-only coordination brief across named project sources. Ask for evidence, a responsible owner and the next decision. Use Chat for a bounded answer, Work for a deliverable, and Codex for repository changes. The first dot is included at no extra cost in eligible Pro and Business Premium plans. Delegated Work/Codex tasks still consume their usual allowance, and launch-period deeper-work limits are temporary. Official terms.
The most useful thing our owner’s dot did was tell them not to merge two fixes independently. That is a reason to try a specific workflow, not a reason to buy another subscription on faith.
A useful warning about two competing fixes
Owner-reported experience, October 3, 2026: a dot surfaced two concurrent fixes in a data project. Both aimed to make updates safer, but they did not necessarily agree on the rules for storing and recovering the data. Two changes can each pass their own checks while making incompatible assumptions about the same system.
The dot recommended comparing their guarantees and tests, selecting one approach, and integrating the useful differences. It also asked which coding session owned that consolidation. The useful output was a decision and a handoff: choose one responsible integrator before merging either change.
Repository-record corroboration, checked October 3: the later engineering record names one integration candidate that combines selected guarantees from both drafts. Its reviewers report reproducing material differences in update ownership and recovery checks. Those reproductions belong to the subsequent engineering review; the dot’s warning did not establish them, and we did not rerun the probes. At this check both proposals remain unmerged drafts, with one proposed to supersede the other. A candidate is not operational acceptance.
The owner described the intervention as proactive. The supplied screenshot does not establish the background trigger, monitoring frequency, reasoning effort, complete context or usage consumed. We have no measured hours, dollars or outage avoided, and no matched comparison with another tool. The case shows one useful warning followed by a documented consolidation. Private repository links and the screenshot are withheld from this public account.
Choose the smallest tool that fits
| What you need | Start here | Define the result |
|---|---|---|
| An explanation, quick comparison or short draft | Chat | One answer using the material supplied |
| A brief, spreadsheet, deck or supported recurring workflow | Work | A deliverable you can review and use |
| Repository implementation or code review | Codex | Named revision, diff, tests and review conditions |
| Judgment across changing work and follow-through between conversations | A dot | One responsibility, authoritative sources and an escalation rule |
These are starting choices, not exclusive capabilities. OpenAI’s Work guide distinguishes answers from deliverables; the Dots guide describes continuity and delegated tasks. Use the Codex guide for repository setup and review.
Space can hold shared Pages and files; Projects can organize chats, instructions and project context. Neither is a mandatory stage before using a dot or a promise of universal memory. Shared material keeps its own access rules: copying private information into a shared Page exposes that copy to its viewers. See Space and Projects.
When a simpler alternative is sufficient
- A known rule: use code or CI to calculate overlapping files or reject an invalid format. File overlap is a signal to review, not proof of incompatible behavior.
- A fixed check: use a scheduled or event-triggered task when the trigger and output are already clear. OpenAI documents recurring tasks and supported app events; ongoing work is not exclusive to dots.
- A one-off question: request one review of the two current diffs. Continuity adds little if there is nothing useful to follow up.
A dot deserves a trial when changing plans, source material and ownership need judgment together. A fixed script can detect overlap adequately; deciding what the overlap means may need a reviewer. This case does not show that only a dot could have found the conflict.
Give it one useful responsibility
Start with one project and one review milestone. Name the current brief, the proposals or drafts, and the decision log. Grant sufficient read access to those sources, identify the responsible person or coding session, and set the threshold for an alert: a changed requirement, incompatible assumption or blocked dependency that needs a decision. Finish with a source-backed brief accepted by that owner, or stop at the agreed review date.
Create a dot on desktop if your account has access, then check which connections it actually has. A messaging channel is not repository permission. Local files need the separately connected computer to be available; prefer approved remote sources if local access is unnecessary. Setup and connections.
OpenAI describes proactive research as read-only gathering from permitted sources. Assigned follow-up work can take actions subject to permissions and approval checks. Set this trial to read-only reporting; instructions add boundaries but do not override built-in safeguards. Dots safety.
Read-only coordination prompt
Replace the brackets. Review the first brief before extending the responsibility.
| |
Two proposed trials, not observed results
A publisher reconciling changed terms. Sources: approved drafts, the official pricing/expiry page and the publisher’s claim ledger. Deliverable: a dated list of affected claims, proposed corrections and the owning editor. Boundary: no edits or publication until review. A scheduled diff can flag a changed page adequately; a dot may help connect that change to several drafts with different owners. Neither proves a checkout price.
A project owner preparing for a milestone. Sources: the current brief, dependency tracker, latest handoff and accepted decision log. Deliverable: a short list of blocked dependencies or handoff statements contradicted by newer evidence, with an owner and next check. Boundary: report only; do not reassign work or contact contributors. A fixed reminder is enough for a known deadline. The proposed dot trial asks whether it can reconcile changing status and ownership across sources.
Access and what the trial could consume
Vendor-documented capability; last checked October 3, 2026. OpenAI identifies dots as GPT-6 Astra-powered. Its introduction includes the first dot at no extra cost for eligible Pro and Business Premium plans. This does not establish equivalence to a separately configured Astra session or a quality advantage over Sol 6.1. Introduction · Product overview.
The access page lists gradual rollout for Pro 100/200/500 users over 18 outside the EEA, UK and Switzerland, Business Premium worldwide, and Enterprise worldwide with administrator enablement. Plus is not listed, and an eligible plan does not guarantee immediate access. Current access.
| Usage category | Documented treatment |
|---|---|
| Conversations with your dot | Excluded from ChatGPT usage limits |
| Deeper work | Included allowance; extended limits for the first month after launch |
| Work/Codex tasks it starts or manages | Normal product usage; Work and Codex share usage |
These are accounting categories, not three published numerical token pools. Dots terms · Shared Work/Codex usage.
The September 29 release note broadly excludes dots usage from eligible allowances for the next month and says later terms will follow. The specific access guide retains the delegated-task distinction. Do not read the broader wording as free delegated coding. Numerical Dots quotas, exact benefit expiry and lasting post-launch treatment remain unresolved. October 29 belongs to a separate Pro 200 transition, not a verified Dots expiry.
Use the offer record for detailed launch terms and the October buying guide for renewal choices. This trial may require no new purchase if your existing eligible seat and allowance cover it. Record actual extra spend; include the full future subscription bill when deciding whether to renew. We did not test signed-in entitlement, usage meters or checkout.