T BOT is a self-calibrating, multi-surface trading system where Claude AI holds full decision authority — running live on real money, 24/7.
The Problem
Algorithmic advantage has always lived on institutional balance sheets. Manual trading loses to speed. Static bots lose to regime change. Neither learns. T BOT is different.
Institutions execute in microseconds. Retail traders react in minutes. The edge gap is structural, not skill-based.
Most retail bots run fixed parameters. Markets shift. Win rates erode. Operators panic-adjust on feeling, not data.
Traditional managed funds charge management fees regardless of performance. Subscribers pay whether they win or lose.
The Solution
T BOT runs four fully isolated trading surfaces under a single Claude AI intelligence layer — each with its own capital, env file, state machine, and kill switch. A change to one cannot touch another.
60-min loop scanning sports props, weather, and economic events. Poisson CDF on ESPN data + Claude Sonnet signal generation. 16-layer filter chain with HMAC-signed trade signals.
30-min loop during FX hours. RSI/BB/SR technical features, ATR-based bracket sizing. BUY-only filter after backtest on 252 resolved trades shows structural SELL bias in LLMs.
Fed funds, CPI, election outcomes. 60-min loop with Haiku research layer. Category concentration cap guards against Fed-funds over-exposure.
30-min loop, momentum + ATR scanner across AAPL, NVDA, MSFT and 27 others. Claude Sonnet directional refinement. Bracket orders via IBKR Client Portal API.
How It Works
Every surface runs the same five-stage pipeline. No manual intervention. Each stage has an escape hatch. Claude holds decision authority within calibration-discipline guardrails.
Market scanner identifies opportunities via volume, price anomaly, momentum composite score
RSS feeds, ESPN data, SEC filings — Claude synthesizes signal narrative per market
Edge + confidence + Kelly sizing. 6 upstream guards. Disagreement log for post-trade audit
HMAC-signed order placement. File-lock dedup. Fail-closed safety flags. Real broker fill.
Resolution, P&L, Brier calibration. Auto-rebalance fires nightly. Bankroll compounds.
# Auto-Calibration Engine — v6.6.1 # 14/23 dimensions auto-managed FOREX_DIRECTION_FILTER=*:BUY FOREX_MIN_EDGE=0.15 ← engine managed FOREX_MIN_CONFIDENCE=0.55 FOREX_RR_RATIO=2.5 ST_MIN_CONFIDENCE=0.60 ← engine managed ST_DIRECTION_FILTER=NO ST_SUSPENDED_CATEGORIES=[] # Shapes A–H cover all calibration types: # multiplier_weight / position_cap / # threshold / set_membership / boolean / # time_window / block_list / ratio_tuner # Burst: 10 cron jobs @ 21:00 UTC daily CALIBRATION_APPLY=true SAFE_RESTART_GATE=39 unit tests
The Holy Grail
We shipped v6.6.0 — the "Holy Grail" — on 2026.05.30. 14 of 23 strategy dimensions now update autonomously from data every night. No human touches a parameter. Claude does it.
safe_restart.sh
Business Model
One model. One sentence. Subscriber pays T BOT a commission on profits they actually withdraw. That's it.
T BOT earns nothing until the subscriber is above their previous withdrawal peak. Losses carry forward — no double-dipping.
Commission crystallizes at withdrawal, not mark-to-market. Paper gains don't trigger fees. Only real money out the door counts.
Early subscribers pay nothing for the first 90 days.
Full terms in subscriber-docs/TBOT_TERM_SHEET_v2.md.
No subscription plans, management fees, or seat counts. Pure profit-share — T BOT's incentives are perfectly aligned with yours.
Security
Two external security audits completed. Every critical and high finding is closed. Defense-in-depth across four independent layers.
Authenticated Origin Pulls — only Cloudflare can reach the origin server. Direct IP requests fail TLS handshake. UFW whitelists CF CIDRs only.
11 secrets in isolated vault. Rotation via stdin pipe — values flow vault → CLI → SSH → env file. Never appear in shell, chat, or logs.
trade_signals.json is HMAC-SHA256 signed at generation, verified at execution. File permission 600. Injection attack vector closed.
parse_dry_run() only returns False for explicit false/no/0/off. Ambiguous values default to DRY_RUN with [SAFETY-H5] log line.
GitHub Actions pushes to droplet via SSH. Pre-restart check aborts if any .env file is missing — guards against wipe on reset --hard.
Cron every 15 min. FX pauses at NAV < $500 USD. ST pauses at balance < $500 USD, resumes at $700. Hysteresis prevents flap loops.
Audit Record: 2 external security audits completed as of v6.6.1. Round 1 addressed network-level findings (H1 + H2 + M1). Round 2 addressed host + supply chain (H3 + H5 + L5 + M5 + M6 + M7). All Critical and High findings are closed. C3/H4/L7 (root monoculture) are the only deferred items — a deliberate half-day scheduled change, not a structural risk.
Traction
Both live surfaces have been running autonomously since May 2026. Every metric below comes from production trade logs, not backtests.
Valuation & Terms
T BOT is an AI-native autonomous trading system with two surfaces live on real money today and nightly self-calibration across 14 of 23 strategy dimensions. Below is the valuation framework, use of proceeds, return scenarios, and key terms accompanying the seed-round term sheet. Confidential — for accredited investors only.
| Method | Berkus + Comparable | Early-stage hybrid |
| Live System (Proof) | $500K | Production on real capital |
| Holy Grail IP | $750K | 14/23 auto-calibrated dims |
| Tech Moat (Security) | $250K | mTLS, HMAC, 2× audits |
| Business Model | $250K | 15% HWM profit-share |
| Execution Track Record | $250K | Solo-built, 35+ commits/day |
| Pre-Money Valuation | $2.0M | |
| Seed Round Target | TBD | Founder discretion |
| Post-Money (est.) | $2.0M+ |
Based on FX profit factor of 14.99× and 65% win rate at current capital. Projections assume capital deployed per allocation plan. Not a guarantee of future results.
| Revenue Model | 15% HWM Profit-Share |
| Management Fee | None |
| Commission Trigger | Withdrawal only |
| Beta Window | 90 days, zero commission |
| Subscriber Target (Ph.4) | Beta cohort, 90-day |
| Round Type | Seed (SAFE or Equity) |
| Pre-Money Val. | $2.0M |
| Instrument | Founder Discretion |
| Min. Check Size | TBD |
| Investor Rights | Pro-rata / Info Rights |
| Board Observer | Optional at lead |
| Capital Deployed | Immediate |
| 60-Day Window | PF > 1.0 milestone gate |
| Exit Horizon | 3–5 years |
| Full Term Sheet | TBOT_TERM_SHEET_v2.md |
Roadmap
Two surfaces live. Two staging. The path to scale is funded data, multi-subscriber infrastructure, and expanding the capital base — not new code.
Four surfaces, full isolation, auto-deploy pipeline, PM2 cluster, Cloudflare security stack, Bitwarden vault, morning brief, HMAC signal signing, circuit breakers.
14/23 strategy dimensions auto-managed. 8 shape engines. 39-test safety gate. Daily calibration burst. Notification fan-out. Two external security audits closed.
FX through 2026-07-26 · ST through 2026-07-27. Stop condition: PF > 1.0 over the 60-day window or surface gets deprecated. Weekly audits. Sunday paired review.
Fund IBKR equities surface ($200K+). Scale OANDA FX bankroll. Launch 90-day beta subscriber cohort under HWM commission model. Multi-user dashboard + subscriber state isolation.
Remaining 9 auto-managed dimensions. Root monoculture hardening. Per-subscriber capital isolation. API for third-party signal consumers. LT surface lift when n≥20 data threshold is met.
Founder
"I shipped 35+ commits in a single day to close the Holy Grail milestone. The infrastructure, the AI stack, the security model, the business model — every line of code is mine. I'm not building a demo. This is live money, live data, and a live system that self-calibrates every night."
Operate a six-product portfolio end-to-end as a solo founder; stack runs on a $30/mo infrastructure budget across DigitalOcean and the Cloudflare edge. Cloudflare Workers · Algorithmic Trading · TypeScript · Anthropic Claude · System Architecture · FastAPI · Python.
Operated ICS / SCADA / HMI in continuous-production food manufacturing across three role levels; implemented KPI tracking and IDS hardening on plant control networks.
Built and ran a 14-year entertainment agency serving corporates and hospitality across East Africa (Safaricom International Jazz Festival, 2017). Concurrent: Director of Entertainment, Capital Club East Africa (2014–2016).
Four trading surfaces (Kalshi, OANDA, IBKR) on surface-isolated env files, per-surface bankrolls, kill switches, and calibration curves. Predict layer uses Claude Sonnet to synthesise RSS news + technical indicators into narrative probability.
Cloudflare Workers + D1 + Pages; HMAC-signed event ingest from every product writing to a unified-events D1 table. FastAPI backend, branded modals, per-surface kill switches.
Subscriber polls the signals feed, applies own per-surface gating + Kelly sizing + concurrent-cap rules, executes via their own broker credentials. Adapter pattern; systemd unit + state persistence. Plus: Canadian PR Mastery, Reinvention Blueprint, Maasai Moran Pipeline.
Ethical Hacking — UBC (2025) · PLC & Industrial Automation — LinkedIn (2025) · CompTIA Network+ (2024) · CompTIA Security+ (2023) · Google Cybersecurity Professional — Coursera (2023) · CS50 — Harvard (2020) · Entrepreneurship in Emerging Markets — HarvardX (2020)
Download full résumé (PDF)The Ask
Capital will go directly toward funding the two paper surfaces into live production and building out multi-subscriber infrastructure. The system is already running. We're raising to feed it. Pre-money: $2.0M.
Fund IBKR equities surface. Scale OANDA FX bankroll. Enable compound growth on all four surfaces.
Premium data feeds (paid market data). Upgraded DigitalOcean droplet for lower-latency execution. Redundant failover.
Multi-user dashboard. Per-subscriber capital isolation. Automated HWM commission crystallization engine.
Contact