A growth tool, a wallet watcher, nine agents arguing about stocks out loud, the control room that runs it all, and a trading bot that went nowhere. No VC, no team, no data center.
Every trading tool on this page is deliberately disarmed. The Kalshi 15m Desk ran 181 samples at a 0.0% hit rate and lost $18.47 live — its own dashboard reports the zero, so it stays on the page. It was found armed and has been disarmed since. FunTokens is WATCH-ONLY with live off and zero backtested picks. The Debate Desk is PAPER ONLY with a live cap of 0. The 100% win rate visible on the FunTokens dashboard is a single sample, not a strategy.
That is not a caveat — that is the feature. The gate opens when the backtests earn it. Every number on this page is a real measured value, and the ones we do not have are shown as insufficient data rather than invented.
An AI growth tool for X with ten working modules. The whole loop: find an idea, write it, queue it, watch what lands.
Watches 40 on-chain wallets and 747 tokens so you don't have to. Shows the wallets, the transactions, and what those wallets are actually accumulating — then tracks whether the hot takes were right.
Nine AI agents argue a stock out loud, then settle it. Not one model handing you an answer — nine giving you a fight you can watch.
A GUI client that speaks the Muse wire protocol directly, plus a connector hub for every agent, knowledge source and platform on the machine. Not a wrapper — sessions, turns, approvals, models and the live view stream are real protocol traffic.
A desk built to beat a 15-minute Bitcoin market. It grades itself in public — every decision, probability, edge and gate score lands on its own dashboard. The screenshots below are the real thing, running, today.
armed=FalseThursday to Friday was a 10× jump in one day — 19 to 35 events. A single Motion With Receipts fan-out reached 6 seats inside 60 seconds. SupaBird also shipped Thursday, and a C:\c\Users\... path bug was caught and fixed within five minutes of first launch.
| System | Role | Version / state | Verified |
|---|---|---|---|
| Muse | Local video gateway + MSP agent host | 1.3.0 · create_ready: true · no API key required | 2026-09-26 |
| FreeBuff | Owns Connecture | 0.0.200 | 2026-09-26 |
| Grok | Build engine | 1.0.41 [stable] · default grok-4.7 | 2026-09-26 |
| MiMo | Not a SaaS — a local model | llama.cpp mimo2.cpp + MiMo-VL template | 2026-09-26 |
These four are not four chatbots. Muse runs the protocol. FreeBuff owns the control room. Grok is the build engine. MiMo is a model running on the box.
| Day | Shipped | Data point |
|---|---|---|
| Wed 23 | SupaBird v1.0 | 10 modules, 11 routes, local Next.js 16 |
| Wed 23 | Fleet sync | 28 targets hit |
| Thu 24 | Motion With Receipts | fanned to 6 seats in 60 seconds |
| Fri 25 | TheEcomDaddy curriculum | 5 public tweets → $79/mo confirmed |
| Sat 26 | Gmail restored | 22,160 → 26,290 messages syncing |
| Sat 26 | 4 apps booted | :3000 :8791 :8796 :8765 all HTTP 200 |
| Sat 26 | MSP integrity | bundled SHA == live SHA, drift: false |
Every trading tool could be switched on with one config line. Doing that would be a mistake.
| System | Mode | Live state | Backtested picks |
|---|---|---|---|
| Kalshi 15m Desk | PAPER · armed=False | disarmed · was armed, found and disarmed | 0.0% hit rate over 181 samples · −$18.47 live P&L |
| FunTokens | WATCH-ONLY · PAPER | off · no implementation | 0 — stats read insufficient data |
| Debate Desk | PAPER ONLY | live cap = 0 | NVDA verdict Hold (1 run) |
| Connecture | Local services | all healthy | n/a — not a trading system |
Anyone showing a 100% win rate on n=1 is selling you something. FunTokens' single closed Hot Take at 100% is a sample of one and the dashboard says so. Paper mode is not the boring part — it is where you find out if you are wrong while it is still cheap.
Captured live on the running desk, in PAPER mode with armed=False. The 0.0% and the Sharpe 3.08 are both read straight off the Quant Lab panel.




















