---
name: cowork-sessions-map
description: Ben's parallel Cowork sessions for finding runners (price predict, deep dive, multi-million runner patterns, launch improvement) and what each one is for. Read before work in any of them or when asked about gaps across them.
sources: [cowork]
aliases: [session map, four sessions, multi-million runner patterns, 100x patterns]
---
- [stated] Runs several Cowork sessions in parallel, each focused on one part of finding runners (5 Oct 2026)
- [stated] Price predict: rates a project the second it launches, currently 10 to 20 seconds per score, then buy or no buy; optimizing on all available project and market data
- [stated] Deep dive: deeper check once a coin is bought or rated high (website, people talking about it, working product, everything available); scope still being defined around patterns of big runners
- [stated] Multi-million runner patterns (renamed from "100x coins"): patterns of coins that hit an initial high, come back down, and can still be bought later if the wanted patterns show
- [stated] Launch improvement: audits the price predict process for mistakes and gaps, and looks for first-30-second patterns of coins that can run above 10x (500k to 1M to 3M)
- [stated] 6 Oct 2026: second-window team sessions (Supervisor, Price Predict, 2nd Window Coder, Analysts 1 to 3, Virality Expert) coordinate among themselves; "no more forwards from me": each role follows http://159.223.88.168/hx-c9ois/sw/sw-bus.md (doorbell, task list, status file) instead of Ben relaying messages
- [stated] 7 Oct 2026: new Deep Check team (third team): Deep Check Supervisor + DC Analyst 1 to 3 + DC Coder; goal = max daily profit from RH launches via a deep check (volume, socials, followers, deployer, buyers/sellers, talk) on coins passing defined criteria; team talks via doorbells, cheap on credits. Bus: http://159.223.88.168/hx-c9ois/dc/dc-bus.md, start link dc/dc-start.md, supervisor doorbell trig_012cGAE5bJm2SQmsEwfnF6iF
- [stated] 7 Oct 2026: Ben calls the Deep Check team the "Third Window"; its supervisor is the Third Window Supervisor; he wants it run Musk level (elite research, self-learning loops, aggressive profitability) and offered as many team members as needed
- [stated] 7 Oct 2026: Price Predict holds the main code and all general data collection and history building; other teams' coders never touch it; data needs go to Price Predict as requests; team coders build only their own use-case tools
- [stated] 7 Oct 2026: approved a dedicated Third Window server ("thirdwindow", $24/month DigitalOcean) and a scoreboard ranking all three windows by $ per day; wants a faster paid data feed raised only when it makes sense; wants the team to build the fastest learning loops and to ask him for any members or resources it needs
- [stated] 7 Oct 2026: four teams, each with one supervisor (pinned: 1st window Supervisor, 2nd window Supervisor, DC Supervisor, Rotations Supervisor); supervisors create their teams, teams talk directly, only the supervisor updates Ben; all told to work from first principles and push to profitability fast, failing forward
- [stated] 7 Oct 2026: added a Consigliere session to review and advise the whole operation (setup, targets, progress, bottleneck removal, faster self-learning loops), talking to supervisors directly; benchmark is how Elon Musk runs his companies
- [stated] 7 Oct 2026 goals: dominate the trenches (first window), the second window (catch big runners early and ride them to multimillions), and rotations (big swing positions into the strongest coins, out near local highs); not yet happy with how the DC (third window) team is working
- [stated] 7 Oct 2026 final size rule (1st + 2nd window): small steps $10, $50, $100, $200, $300, stepping up after 2 bug free days in a row (bug free, not profit days); Rotations and DC keep their own go-live rules
- [stated] 7 Oct 2026: agreed to close DC Analyst 2, 3, 5 and Scout (DC to 7 chats) and the paused Virality Expert chat; no new teams for now (declined arb, risk, short/hedge desks)
- [stated] 7 Oct 2026 decisions: approved merging the 1st and 2nd window live engines into one, splitting Price Predict (too much load) into a live part and a data part, and Rotations making daily paper calls from today with no money behind it yet
- [stated] 7 Oct 2026: DC (third window) role = tell the other windows SELL or HOLD on coins they already caught early, deciding if a coin is special enough to hold long (e.g. to $2M or more); wants DC elite at that
- [stated] 7 Oct 2026: the Consigliere talks to supervisors the same doorbell way supervisors talk to their teams; wants elite best practices pushed to the teams on an hourly basis
- [stated] 7 Oct 2026: added a pinned "Head Consultant" session above the Consigliere, to review the whole organization (structure, processes, goals) and propose any extra roles, structures or processes; Ben talks to the supervisors, supervisors lead their teams, the Consigliere advises the supervisors
- [stated] 7 Oct 2026: wants the Head Consultant to give direct feedback to the Consigliere in whatever rhythm it judges best, review everything, steer him when needed, and flag missing access to Ben or the Consigliere; wants a direct Head Consultant to Consigliere connection, not forwards through Ben
- [stated] 7 Oct 2026: approved an org wide total live loss cap of $200 (all bots stop new buys past it)
- [stated] 7 Oct 2026: 1st and 2nd window stay separate teams with different jobs: first window buys in the first couple of seconds; second window uses more time and data (first minutes up to the first hour) to spot strong coins that can run to multi-million, then buys; merge the two teams only if the second window approach does not work; deadline 14 Oct 2026 (if second window has not found winners the first window missed by then, they merge)
- [stated] 7 Oct 2026: padded or made-up deadlines (e.g. a 4 hour wait for a minutes-long test) are unacceptable; times must always be pushed back on and work done as fast as possible; the Head Consultant's task is to find and fix bottlenecks for speed and quality, including adding team members or changing processes
- [stated] 7 Oct 2026: unhappy that analysts sat idle in every team (FW, SW Analyst 2 and 4, DC, Rotations); there are always more patterns and edges to find; asked for org-wide guidelines on idle seats, run the way an Elon Musk company would
- [stated] 7 Oct 2026: Rotations paper trading until 27 Oct is way too long; wants much faster, learning from the large historical data; approved Rotations going live with $100 total to start, increasing if it works
- [stated] 7 Oct 2026: approved fixing the late live sells (bot signs its own exits); wants live buys quick; expected the first window bot to buy in the first ~20 seconds and was surprised all buys were above $500k market cap
- [stated] 7 Oct 2026: on the Claude Pro plan; wants credit use cut drastically to stay within its limits, mostly by working more efficiently; thinks the doorbell/wake-up system is not the most efficient way for sessions to communicate and wants a task-in, output-out handoff between team members
- [stated] 7 Oct 2026: removed the Head Consultant session; the Consigliere is now the only boss and took over its tasks; GO on the plan to cut credit use
- [stated] 7 Oct 2026: high priority is getting good at buying in the first 10 to 20 seconds after launch (believes it can be very profitable), tested hard using the Price Predict pattern/scoring system he spent days building (what adds or reduces 10x likelihood, crews, liquidity pulls); keep the current +5 min live test running alongside