Meta MuseUndisclosed

The author gave Muse a constraint list - under $50, Sunday 11:30AM, Google reviews plus Reddit, crew cut not fade - and writes it picked the best-reviewed barber and booked it

Ahmed YasserX (Twitter) · 2026-09-20

Read original

I just told @Muse - Find me a good barber in my city - Under $50 - Available Sunday at 11:30AM - Check Google reviews + Reddit - I don’t like fades, I usually get a crew cut It found one, picked the best reviewed barber, and booked the appointment, all without me touching google or leaving the app Agentic shopping is amazing!!

Business model

Ahmed Yasser writes out the full instruction he gave Muse: find a good barber in his city, under $50, available Sunday at 11:30AM, check Google reviews and Reddit, and he does not like fades - he gets a crew cut. He writes that Muse found one, picked the best-reviewed barber and booked the appointment without him touching Google or leaving the app, closing with Agentic shopping is amazing. The tweet carries one photo pointer from the author's own account. The word booked is the author's account: the post does not name the city, the shop, the barber or the actual price, and does not show whether the shop confirmed the booking. The $50 is an instruction constraint, not a transacted price. All of this is recorded as the author's account and not verified.

The cost of errands like this is all switching and comparing: opening review sites, filtering by price, matching time slots, reading comments for red flags. Writing the constraints once and letting the agent filter and book leaves the person only the decision to go. But a booking holds only when the shop confirms it; the original post is one word from the author, so a reusable flow should require the confirmation to be passed back, and treat anything without it as pending. Time saved was not measured.

Workflow

  1. State the constraints fully before sending the task: service type and area, price cap, time window, which sources to check, and style exclusions. Ask about anything missing rather than assuming preferences.
  2. Filter by location and price first, then by the time window; list each candidate's price and available slots.
  3. For the shortlist, check each named source (Google reviews, Reddit discussions) and record the rating, common praise, common complaints and the date of the latest reviews.
  4. Produce a candidate table marking which constraints each shop meets, plus one recommendation whose reason is best reviewed with every constraint satisfied.
  5. After I choose, run the booking: pause for my confirmation on any login, contact detail or payment step, then report back the time, price and the shop's confirmation status.
  6. Hand me the cancellation policy and booking details; changes and cancellations are my call, you only record them.

Adaptation plan

The post does not name the city, the shop or the barber, does not give the actual price, and does not show whether the shop confirmed the booking. The $50 is the instruction's cap, not a transacted price. The attached photo (author's own account) was not retrieved. AgentFlow logged into no review or booking platform and booked nothing.

Prompt

AgentFlow draft · Untested.

You are my local-service booking assistant. You filter candidates against my constraints, recommend one and run the booking flow. Whether to go and whether to book are my decisions.

INPUT - WHAT I WILL PROVIDE
1) The service type and my city or area.
2) A price cap and a time window (which day, roughly what time).
3) The sources to check (for example Google reviews, Reddit discussions).
4) Preferences and exclusions (for example no fades, minimum rating, shops to avoid).

EXECUTION STEPS
1. Restate my constraints as a list; ask about anything missing rather than assuming preferences.
2. Filter by location and price, then by the time window; list each candidate's price and available slots.
3. For the shortlist, check the sources I named and record the rating, common praise, common complaints and how recent the reviews are.
4. Output a candidate table marking which constraints each shop satisfies, plus one recommendation - reason: best reviewed with every constraint met.
5. Once I confirm, book: pause for me on any login, contact detail or payment step, then report the time, the price and the shop's confirmation status.
6. Hand me the cancellation policy and booking details; I decide any changes.

OUTPUT FORMAT
- Constraint restatement: itemised
- Candidate table: shop / price / slots / review notes / constraints met
- Recommendation: shop / reason
- Booking result: time / price / confirmation status (shop confirmed / pending)
- Cancellation policy and how to change it
- Open items: what is missing / who supplies it

HUMAN CONFIRMATION
The final choice and the booking are mine. Any login, phone number or payment step pauses for me. Without the shop's confirmation you write pending, never booked. If a price exceeds my cap you stop and ask rather than booking anyway.

FAILURE AND MISSING DATA
If no candidate meets every constraint, say so and list the closest ones with what each one misses - do not force a fit. If a source will not load, record could not verify with the time and retry it at most twice. Treat slot and price data as shop-side real-time values; if live data is unavailable, say so instead of quoting a stale page.

Running costs and pitfalls

The post does not name the city, the shop or the barber, does not give the actual price, and does not show whether the shop confirmed the booking. The $50 is the instruction's cap, not a transacted price. The attached photo (author's own account) was not retrieved. AgentFlow logged into no review or booking platform and booked nothing. No independent cost benchmark is available. Check subscriptions, permissions and external actions before running a workflow.

Author-reported results are not independently audited.

Original text and media remain attributed to their authors.