Product Brain

Source

Finn Mallery (@fin465, founder @origamichat) — YC-style method to compress a month of GTM research into ~3 hours with Claude as a tireless research partner.

Original post: x.com/i/status/2086939268136755285 (10 Aug 2026) — high bookmark signal at capture (~5.6k bookmarks).

Route note: primary path is /productbrain. /productresearch 301s here (same page; “product brain” is the product concept we care about).

This page does two jobs: (1) capture Finn’s method cleanly, and (2) map it onto Para Living (paraliving.com) and our ecommerce brain app theflysales.com — especially the design question: is research a disposable task, or a living object per product?

One-sentence TL;DR

Don’t ask the model to “research the market.” Feed it dense primary sources, force questions about unspoken consensus and broken assumptions, attack the idea like an investor — then store that work under a specific product so competitors, market signals, your listings, and reviews can be monitored forever (e.g. skincare as its own product brain).

Finn’s method (the 3-hour compress)

What you do not do

You don’t open Claude and say “research the market.” That’s how you get generic slides.

What you feed it

What you ask (in order)

1. “What does every successful player in this market understand that customers never say out loud?”
2. “What are the assumptions this entire market is built on, and what would have to be true for each one to be wrong?”

After ~15 minutes of that pressure, you want three outputs:

• Blind spots
• Consensus
• Ideas relevant to customers that nobody’s talking about

Then attack like a great investor:

“What’s the strongest version of this argument, and where does it still break?”

~3 hours later: a very initial GTM strategy that feels like decade-in-market intuition — still a hypothesis. Next step is real customer conversations to validate (Finn points at tools like Origami for building target lists).

Why this matters for Para Living

paraliving.com (Para Living) is a real brand with real SKUs, listings, reviews, and competitors — not a demo chatbot. The Finn method is the seed ritual for each line (skincare is the running example): dense corpus in → structured insight out.

The mistake brands make is stopping at a Notion page or a one-shot Claude project. That research rots. Competitors relaunch. Reviews shift. Your listing drifts. The Para Living team should treat each important product line as having its own product brain — a durable object the team and agents keep updating.

theflysales.com — our ecommerce brain app

theflysales.com is our app — the ecommerce brain web app for product intelligence (shown at the Hua Hin coworking week — see /huahin2026). Related field notes already on this site:

• Amazon market data APIs — SP-API + Sellerboard for your account; Jungle Scout / SellerSprite / Keepa / etc. for market, keywords, competitors, history
• Company brain — org-level single brain + specialist agents (Eric Siu framing)
• Agentic SEO loop — continuous ranking intelligence with memory, not one report

Product Brain sits one level below “company brain”: company brain = org workflows and shared context; product brain = everything true about one SKU / line / ASIN cluster over time.

The design question (open confirmation)

Mike’s question for the team — and the reason this page exists beyond a tweet summary:

Skincare research is a task in our brain — but is it tied to a specific product so we can continually monitor competitors, market, our own listings, and reviews to improve it as an independent product brain?

This repo does not have live theflysales schema access from here. Treat the following as target architecture the Para Living team should confirm or correct against production in theflysales.com.

Confirm with the team

1. Is every research pack linked to a product ID / ASIN / line in theflysales?
2. Can we open “Skincare” and see competitors + market + our listing + reviews in one place?
3. Do agents re-ingest competitors / reviews on a schedule, or only when a human runs a task?
4. Where does Finn-style corpus live — object storage + vector, or only in a chat history?
5. Who owns refresh cadence per product (weekly reviews, monthly competitive, etc.)?

If any answer is “it’s just a task,” the upgrade is clear: promote research to a first-class product brain object.

What a product brain holds

How Finn’s ritual maps into theflysales

One-time seed (Finn method)
  feed: competitors + earnings + reviews + Reddit
  ask: unspoken truths → market assumptions → attack idea
  out: blind spots, consensus, white-space, draft GTM
        │
        ▼
Product Brain object  (e.g. "Para Living Skincare")
  stores corpus + structured outputs + owner + cadence
        │
        ├── ongoing: competitor scrape / price / listing diffs
        ├── ongoing: review theme extraction
        ├── ongoing: our listing + rank / keyword health
        ├── ongoing: market demand signals (JS / SellerSprite / Keepa / …)
        └── agents: weekly brief, alert on competitor move, propose listing tests
        │
        ▼
Human: real customer / buyer convos + ship changes
        │
        ▼
Loop: results write back into the same product brain

Same closed-loop spirit as SEO loop and company brain — scoped to a product so skincare does not pollute supplements, and agents never confuse corpora.

Example: Para Living skincare as a product brain

Agent / build checklist (for the team)

Related on this site

• Amazon market data APIs — data feeds for market + competitor layers
• Hua Hin 2026 — theflysales shown as ecommerce brain
• Company brain — org-scale single brain
• Agentic SEO loop — continuous ranking memory
• AI-native GTM — agents + lead data + outreach
• Agent growth plays · CRO field notes
• Sourcing GPT — adjacent product intelligence

Primary method source: @fin465 on X · Brand: paraliving.com (Para Living) · Our app: theflysales.com