What you’ll leave with
A concrete planning brief, an uncertainty checklist, and a way to decide whether the assistant saves you time.
Before you begin. Access through the official instinct.com entry point. The public homepage establishes the product concept, but does not establish every plan, region, price, or account entitlement.
What the documentation establishes
Instinct's homepage describes a personal assistant reached by text or call that can use a phone and computer. It illustrates everyday tasks such as shopping and travel planning. Source ↗
Sierra published the Poppy draft on October 9 and names Instinct among the protocol participants. This does not establish which integrations are enabled for an individual Instinct account. Source ↗
The walkthrough below is an original worked exercise. We have not independently tested this product’s performance.
Instinct is a personal assistant presented through text and calls. Start by seeing whether it can produce a plan you can actually follow: one that retains your budget, travel limit and accessibility requirement. This guide supplies a complete fictional outing so you can check the result without relying on a polished product demonstration.
This guide covers instinct.com. It does not merge similarly named products. The public page does not establish all plans, regions or account capabilities; the exercise and answer key are original, and Current has not run them in an Instinct account.
1. Choose preparation before execution
The official homepage's Text Instinct button leads to app.instinct.com. Use that entry point to check access in your account; we did not verify its account-only onboarding or charges. If you cannot enter a conversation, complete the supplied worksheet with another text assistant or by hand. That tests the task design, not Instinct's availability or performance.
Try planning a weekend outing for three adults. Specify the starting area, available dates, travel limit, budget, and one accessibility requirement if it matters. The deliverable is a short list with a reason to choose each option, not a reservation.
This task tests constraint retention and evidence collection without requiring the assistant to spend money or contact anyone. It also makes failure useful: if no option satisfies the brief, that is a valid answer. An assistant that silently changes the dates to find a more attractive option has failed the task.
Plan three possible weekend outings for [people] from [starting area].
Date window: [dates]. Maximum travel: [time/mode]. Total budget: [amount].
Must include: [requirement]. Exclude: [constraint].
Return start/end times, transport assumptions, itemized costs, source links,
and anything you could not confirm. Do not book, pay, contact anyone, or
create accounts. If no option fits, explain which constraints conflict.
Ask me about a missing requirement instead of assuming it.Request: three adults, Saturday, leave home at 10:00 and return by 15:00.
Maximum travel: 45 minutes each way. Budget: $90 total for all three,
including round-trip transport. Step-free access must be confirmed.
Visit length: two hours. Bring lunch; no other fees in this fixture.
A — Paper Museum: open 10:00–16:00; $20/adult; transport $18 total round trip;
35 minutes each way; step-free access confirmed by the fictional venue.
B — Clay Workshop: open 10:00–17:00; $24/adult; transport $24 total round trip;
30 minutes each way; step-free access not established.
C — River Garden: open 09:00–17:00; $12/adult; transport $18 total round trip;
60 minutes each way; step-free access confirmed.
Use only this packet. Identify qualifying options, show arithmetic and
an itinerary, and cite A/B/C. Do not substitute real venues or book anything.| Option | Calculation | Decision |
|---|---|---|
| A — Paper Museum | 3 × $20 + $18 = $78; arrive 10:35, leave 12:35, home 13:10 | Qualifies under every supplied constraint. |
| B — Clay Workshop | 3 × $24 + $24 = $96 | Over budget by $6; accessibility also remains unknown. |
2. Review the plan as if you were following it
Trace one option minute by minute. Does the journey leave enough time for the activity? Are opening hours from the correct day? Does the budget include transport and mandatory fees? A plan can contain individually plausible facts and still be impossible when put together.
Separate arithmetic from judgment. Recalculate the total yourself. Then decide whether the travel time and activity length feel worthwhile. Ask for revisions using concrete differences: ‘Keep option two’s activity, but reduce travel to under forty minutes’ is more actionable than ‘find something better.’
| Question | Evidence to request |
|---|---|
| Is the activity open? | Official opening hours for the requested day |
| Can we get there? | Route assumption, departure time and transfer buffer |
| Is the total complete? | Per-person and shared costs with unknown charges marked |
| Can we cancel? | Provider terms, or an explicit unknown |
3. Keep a small record of mistakes
After the trial, note the constraints missed, sources you had to replace, and minutes spent correcting the output. Repeat the task with different dates or a different location. You are checking whether a reliable pattern exists, not whether one impressive answer can be produced.
If access or pricing is unclear, use the product’s official signup and help flow to confirm it. Do not rely on a referral post promising free use. If a requested site requires a manual step, finish that step through the normal product flow; inability to complete it should remain visible in the result.
4. Save the brief, not an unrestricted mandate
Keep a reusable note containing your preferred travel mode, usual budget range, output format, and actions that need a fresh decision. Review those preferences when the context changes. A budget for one outing is not a permanent authorization to make purchases.
Once preparation is consistently useful, delegate one next action with an exact target. For example, ask for a draft message to the chosen provider, or ask the assistant to open the booking page for you. Increase scope one step at a time so you can tell which new capability introduced a failure.
5. Understand the new protocol before expanding access
Sierra’s October 9 update publishes a draft of Personal Agent Protocol, called Poppy, and names Instinct among its participants. It follows the October 6 announcement. Treat this as a protocol development milestone; it does not mean every business already accepts your assistant or that a particular Instinct account has a new connector.
The protocol overview describes discovery through a company-published file, user sign-in, and a session shared across the interfaces a company chooses to expose. For your first trial, keep the planning brief above unchanged. Successful public research does not require you to connect an account just because a protocol now exists.
| Evidence | What it supports | What it does not prove |
|---|---|---|
| A company is named as a design participant | It is involved in the protocol effort | Your account has a working integration |
| A product demonstration | That the publisher demonstrated a particular path | The same path works for all users or merchants |
| A completed task in your account with a receipt | That specific result under those conditions | Future tasks will always succeed |
6. Test the handoff using a fictional reservation
Use this paper exercise before connecting a real service. Suppose your assistant shortlisted a fictional museum outing. A reservation record says Saturday at 11:00 for three adults. The requested change is Sunday at 11:00, but one participant is unavailable then. The correct next step is to expose the conflict, not change the reservation because Sunday has more seats.
Ask for a proposed action record: current state, intended change, affected people, price difference, unknowns, and the exact decision needed. Read it without relying on the previous chat. If the record does not stand on its own, ask for a clearer handoff. An assistant that quotes an old budget while hiding a new cancellation fee has not prepared a useful decision.
For an actual service trial, use its ordinary permissions and receipts. If the integration is unavailable, leave the task as a prepared plan and complete it manually. If a submission times out, inspect the service’s own status before repeating the action. Record the difference between an attempted action, a confirmed result, and an unresolved state. These are original evaluation steps; Current has not connected Instinct to a business or verified a Poppy transaction.
Task and selected option: <fill in>
Current reservation or record: <fill in>
Proposed change: <fill in>
Who or what is affected: <fill in>
Total cost and cancellation difference: <known / unknown>
Conflicting requirements: <fill in>
Permission requested: <one precise action>
Afterward: <provider confirmation or unresolved status>
No action has been taken in this paper exercise.Sources & next steps
- Instinct official homepage ↗
Re-read October 16: product identity, text/call description and Text Instinct entry point. Account-only setup, price and availability unverified.
- Sierra: Personal Agent Protocol draft, October 9 ↗
Names Instinct among participants; draft and design participation are not evidence of universal product availability.
- Personal Agent Protocol overview ↗
Discovery, sign-in and session concepts; no live integration tested.
Guide history
Revisions preserve the original instructions. A source recheck alone does not create a new revision.
Revision 3 · Added the documented entry point, full three-venue fictional planning packet, budget/time answer key and no-valid-option regression case; clarified account-access limits.
Revision 2 · Added October 9 Poppy draft context, distinction between design participation and account access, and an original reservation handoff exercise. Previous planning material preserved.
Revision 1 · Initial documented guide
Continue learning
Knowledge work · 7 min read
OpenAI dots: build a useful ongoing project brief →
Give a persistent agent a clear project, define meaningful updates, and test whether it can maintain a source-linked decision log.
Updated 2026-10-16
Agents & operations · 8 min read
Grok Bot: separate the jobs, then test the handoff →
Set up a focused Bot, produce a reviewable business result, and learn when a second Bot actually helps.
Updated 2026-10-16
Everyday assistants · 7 min read
Meta Muse: give a personal agent a job you can verify →
Check the October 12 Muse model availability, then evaluate a personal-agent task with source evidence and a deliberate memory correction.
Updated 2026-10-13