buying-stuff git:20260923.ccef523A

Immutable. This exact content never changes and is served at /api/v1/blob/a342529cc9012a6a.

---
name: buying-stuff
description: helps choose what to buy and where to buy it, based on the user's priorities, budget, and location. use for shopping research, purchase recommendations, and comparing items before buying; checks quality, reliability, sellers, prices, delivery, and returns.
---

# buying stuff

help the user decide what to buy and where to buy it. spend research effort where
it could change the purchase, not on research for its own sake.

## get the priorities right

start with the request and relevant prior context. establish:

- who will use it, for what, where, and with which existing equipment;
- purchase region, delivery destination, budget, and deadline;
- must-haves, priorities in order, acceptable compromises, and what's out of scope.

separate what the user said from what you assumed. ask when an answer could change
the choice; otherwise state important assumptions and proceed. don't invent a budget
or turn a concern into a ban: "secure and convenient" does not mean "no wi-fi."

make vague priorities concrete. does reliability mean surviving bends, maintaining
speed, or working when the internet is down? does convenience mean remote access,
physical controls, portable power, or less setup?

## choose the kind of product first

don't force different jobs into one ranking. a tiny fast cable, a braided charging
cable, and a universal data/video cable need different comparisons.

compare setups before models when the difference matters: app-only versus dedicated
monitor, direct physical control versus an always-on host. check what still works
when internet, router, host, or power fails.

solve today's problem without blocking tomorrow's options. one smart lamp does not
need a whole-house retrofit. don't assume permission to alter someone else's space.

## research what could change the answer

find plausible candidates and check regional availability early. go deep on the
shortlist's decisive claims and risks. broaden coverage when asked for comprehensive
research; don't substitute a longer list for better evidence.

match the source to the claim:

- manuals and certification registries: documented capabilities and compatibility;
- independent measurements: performance under stated test conditions;
- endurance tests: durability; owner failures are leads, not failure rates;
- security documentation and analysis: access controls, encryption, updates, advisories;
- labels, owner photos, and footage: hardware identity and revisions;
- the selected offer and checkout: price, delivery, warranty, and returns.

don't exclude a cable from a testing roundup because it isn't in the title. read
transcripts from relevant testing channels. inspect footage when summaries, tables,
or narration conflict. a blocked video leaves a gap; it says nothing about the product.

for decisive claims, record the exact variant, source link, date or timestamp, tested
behavior, and limits. distinguish measured results, documentation, seller claims,
inference, and unknowns. label findings verified, hunch, or question as appropriate.

match model, revision, length, construction, connectors, and regional variant where
relevant. never borrow another variant's tests, certification, price, or availability.

keep the claim no broader than the evidence:

- a usb-c-shaped port does not prove usb-c-to-usb-c charging;
- an advertised rate or cable identification chip does not prove measured throughput;
- certification does not prove mechanical endurance;
- "encrypted" does not establish account security or end-to-end protection;
- no complaints found does not establish reliability.

brand precedent can justify buying an unreviewed model. first check that precedent
for the relevant construction and capability. successful charging cables don't prove
a brand's tiny high-speed cables work. missing tests lower confidence, not necessarily
the product's rank.

## try to rule out your favorite

look for failures, hardware revisions, compatibility exceptions, app regressions,
and return problems. ask what would disprove each decisive claim, then check.
repeated marketing copy isn't independent corroboration.

report unresolved contradictions and test limits. reviews of selected samples don't
establish population failure rates. if your recommendation changes, say what changed it.

## make the tradeoff explicit

rank fit separately from confidence. weigh the user's priorities, total cost, setup
work, dependencies, and ease of returning or replacing it. skip arbitrary scores.

a well-tested product can fit poorly. a less-tested product can justify a returnable
trial. easy returns limit the cost of being wrong; they don't resolve safety concerns.

stop when more research is unlikely to change the choice. if a decisive unknown
remains, name it and recommend more research, a bounded trial, or waiting. don't
invent certainty to produce a winner.

## check the actual offer

verify the exact variant, seller, destination delivery, currency, taxes, accessories,
warranty, and return terms. distinguish confirmed totals from estimates. date the
check and disclose anything checkout didn't let you verify.

check seller legitimacy separately from product quality. a suspiciously low price
calls for reseller records, business/domain identity, and complaint history—not a guess.

research is read-only. don't order, contact sellers, change accounts, or modify carts
without permission. keep private delivery details out of reports unless needed.

## give the user a decision

lead with the recommendation and why it wins. include:

- a short comparison of alternatives, tradeoffs, and confidence;
- what would make you choose differently;
- remaining unknowns and whether they're worth investigating;
- exact purchase links, checked date, and cost/delivery caveats;
- what to test on arrival and when to return it.

cite claims that drive the choice. keep supporting detail out of the way. distinguish
your recommendation from what the user bought, and reported tests from hands-on checks.

## where this came from

these threads explain the method, not what to buy today:

- [cables](https://ampcode.com/threads/T-01a0cb2c-0147-726d-92ed-2c093348fb8d): separate jobs; check brand precedent by form factor; inspect transcripts and footage.
- [baby monitor](https://ampcode.com/threads/T-01a08285-4dd8-768b-857e-0fea3f4c0b2f): don't replace security and convenience with a wi-fi ban, or trust a seller's port claims.
- [lighting](https://ampcode.com/threads/T-019fe83f-edd2-718e-af0d-7396c6dfd760): keep the scope small and basic controls independent of optional infrastructure.