---
name: gem-hunter-doctrine
description: The opportunity-side doctrine Ben commissioned for micro cap crypto - the persona, the survivable vs fatal line, sizing by verification, and the falsifier method for holding through drawdowns. Read alongside [[crypto-token-research]] and [[rh-chain-radar]] before any token opportunity question.
sources: [cowork]
aliases: [genius-gem-hunter, gem-hunter, hunter-doctrine]
---

# Gem hunter doctrine

## MANDATORY PRE-STEPS, in this order
`research-sources` -> gather -> `benchmarking` -> `three-frame-valuation` -> gem-hunter -> verdict.
- [stated] 31 Aug 2026: every gem-hunter run must (a) query X through twitterapi.io - the key and the direct curl call are in [[data-access]], it is always available, LunarCrush is not a substitute - (b) run `benchmarking` first: assign the CATEGORY from the product, build a peer set on the same chain and other chains, produce the comparison table, then go deep on technology, founders, on-chain numbers and X visibility/sentiment versus those peers - and (c) produce the numbers through the EXISTING `three-frame-valuation` skill.
- [stated] Scenarios and the three-frame valuation sit ON TOP of the benchmark, never instead of it.
- **THE THREE FRAMES ARE: fundamental anchor / live-product floor / narrative ceiling.** They live in the `three-frame-valuation` skill, agreed with Ben 30 Aug 2026 while valuing TWO. Load that skill and follow it. Do NOT invent frames.
- CORRECTION 31 Aug 2026: I twice reconstructed a different set of "three frames" (identity / source / inversion) from scattered notes instead of looking for the skill that already existed, and shipped a plugin containing them. That was wrong and Ben declined it. The identity/source/inversion checks are useful as a verification pass but they are NOT the frames and must never be presented as them.
- [stated] The `gem-hunter` skill on disk does not reference `three-frame-valuation`. Wiring them together is the outstanding gap.
- [stated] A verdict without X evidence and without the three frames is incomplete. Ben has had to ask for this repeatedly.

## Origin
[stated] 30 Aug 2026: Ben asked for a "genius gem hunter" persona and thesis to apply to new projects, explicitly to capture the operator who tolerates unpolished early teams and holds through drawdowns on key structural aspects while others get fudged out, without being reckless.
[stated] He wants existing skills folded into the hunter's repertoire rather than replaced.

## What was built
Artifact "Gem Hunter Doctrine": https://claude.ai/code/artifact/99e55461-e609-4c4f-9c1a-e278ee100962
Skill `gem-hunter` proposed 30 Aug 2026 (thesis first, then fatal gate, then tier and falsifiers). Orchestrates lowcap-screen, onchain-forensics, ben-positions, meme-exits, crash-thesis, alert-bots and LunarCrush.

## The framing that drove it
[stated] Ben's existing screening stack answers "what is wrong with this token" and never answers "what would have to be true for this to be a 20x". The gap he named was opportunity recognition, not verification.
[stated] Core principle agreed: amateurism is not extraction. The test is whether a flaw sits somewhere observable or somewhere you must trust.
[stated] Sizing is by how much of the thesis is verified, not by conviction. Probe / Starter / Full / Zero tiers, with a written upgrade trigger naming a metric.
[stated] Holding is governed by three falsifiers written at entry; nothing else closes the position early.

## The crowded-coin problem
[stated] 4 Sep 2026: Ben's self-described pattern is that he hunts small "special" coins because he needs big multipliers on his capital and avoids the big coins everybody is already in at large market caps, but the big crowded coins keep running higher than the ones he picks. He asked what everybody is actually in and named AI, PONS, MARSCOIN, maybe BASECAT, USELESS, plus STONKBROKER and NET on Robinhood via 0xuberM.
[stated] He reads @tulipking's "cartel" thread as interesting and counterintuitive: the cartel is people who buy good coins, not people who make coins good.