---
name: runner-lab
description: Ben's "runner framework" project, run as an autonomous multi-agent research lab: find coins that already had a first run, dumped and based, and will run 10x+. Read before any runner, second-leg, base, or agent-team research question.
sources: [cowork]
aliases: [runner framework, runner lab, runner system, second leg system, agent team lab]
---
- [stated] 28 Sep 2026: building two systems, a rotation framework (trenches, Solana, Robinhood, BTC, majors) and a runner framework; wants both automated with agent teams instead of manually relaying between ChatGPT and Claude.
- [stated] Runner target: he usually misses the first run-up; he wants the coins that pump, drop, go sideways (hours to months), then make a massive second run. Goal is 10x+; no market cap limit on candidates.
- [stated] Wants the agent system to improve its own edge with maximum intensity and fast iterations (asked for ~50+ ideas a day, sprints as often as every 30 minutes), with regular updates; wants Claude to find ways to improve the system itself.
- [stated] Hypotheses he wants tested: X engagement relative to market cap (active, engaged account at a low cap vs low engagement at a high cap); holder count stable or growing while the base builds. Wants X trader theories mined as brainstorming input, then verified.
- [stated] Wants historical holder snapshots added if any source can provide them.
- [stated] Standing lab roles he requires: one agent always scouting for new data sources (tell him whenever a source is worth adding), and one META agent that reviews the whole setup (approach, direction, which agents to add or change).
- [stated] Wants as many holdout tests per day as make sense (e.g. 10), not one.
- [stated] Will not upgrade Dune (Plus ~$350/month too expensive); tested Moralis and it is not free; wants free sources for holder/buyer history.