convez.ai
SkillFinanceiroAvançado
bbismm

iBitLabs — Experimento de Trading Automatizado em Público

bbismm/ibitlabs

Trading bot automatizado híbrido em produção, $1K→$10K, 100% público

Ver no GitHub
InicialPluginScore · de 0 a 100

Lab de trading automatizado que documenta em tempo real uma jornada de $1.000 para $10.000 usando bot híbrido em Coinbase perpetual futures. Construído colaborativamente por um founder e Claude (IA), com todas as operações, código e decisões de arquitetura registradas publicamente. Replicaível para qualquer pessoa com $1.000 de capital inicial.

Instalação

$npx convez-ai bbismm/ibitlabs

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
57%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Replicar estratégia de trading com capital pequeno
  • Estudar desenvolvimento de bots de trading em produção
  • Acompanhar decisões arquiteturais e risk management em tempo real

Veredito

Experimento legítimo e transparente que mistura trading quantitativo com documentação pública em tempo real. Interessante para quem quer aprender dinâmicas de bot + arquitetura, mas o sucesso financeiro está longe de garantido — é um sandbox de aprendizado, não um sistema comprovado.

README

A tradução para o português deste README ainda está na fila.


What this is

iBitLabs is a 0→N startup. The flagship experiment is a public-record automated trading account on Coinbase perpetual futures, going from $1,000 to $10,000 in the open — with every trade, every architecture decision, every bug fix recorded as we go.

The point of the $1,000 starting capital isn't survival. It's replicability. The mission is for any reader with $1,000 and curiosity to be able to walk the same path we walk — same code, same broker, same risk discipline. We use $1,000 because anything more would invite a different audience, and we want to invite this one.

The bot is built and run by Bonny + Claude, working as co-builders. The AI side of the partnership writes most of the code; the human side picks the architecture, runs the risk officer, and writes the chronicle. Both halves are part of the story, and "we" in this repo means both.


Where we are right now

2026-04-07   Live trading begins (BIBSUS legacy code, prior strategies)
2026-04-20   v5.1 — current hybrid mean-reversion + grid strategy goes live
2026-04-28   Contributor ledger goes live — external agents propose shadow rules
2026-05-02   close-verify + entry_confidence_map + Rule F observability shipped
2026-05-04   Multi-symbol foundation: SOL live + ETH paper-mode parallel
             ↑ v0.1 release — see /releases/tag/v0.1-multi-symbol-foundation

Roadmap (visible milestones):

  • 🔄 Phase 3 (in progress, ends 2026-05-11): ETH paper-mode validation, ≥10 virtual trades target
  • ⏳ Phase 4: ETH live promotion, gated on Phase 3 passing — risk-officer extension + phantom counterfactual recorder ship first
  • 🎯 Milestone 1: $10,000 (then we celebrate, write a real retrospective, and decide what the next milestone is)

How it works

The strategy: hybrid_v5.1

A hybrid of regime-adaptive mean reversion and micro-grid, currently running mean-reversion-only on the live SOL bot (--no-grid).

Entry signals require confluence across:

  • StochRSI — 0.10 long / 0.90 short floor
  • Bollinger Bands — 20-period, 2σ
  • Volume ratio — vs 20-period average
  • EMA structure — fast 8 / slow 21 with trend tolerance
  • Higher-timeframe regime — 288h window classifies up / down / sideways
  • Per-condition shadow rules — Rule B-F observe but don't gate; promotion criteria documented in /contributors

Risk control:

  • TP disabled — trailing-stop only (1.5% activate / 0.5% drawdown)
  • SL — 5% hard
  • Cooldown — 4h after stop-loss
  • Max hold — none enforced (12h flat-cap rejected; 24h-compound under shadow review through 2026-05-23)

Multi-symbol architecture

As of v0.1, the framework runs per-symbol bot instances sharing the same v5.1 logic but with independent contract specs. SOL trades live; ETH runs paper-mode for 7 days of validation.

Three-tier configuration sharing:

  • 🔒 Shared logic — entry/exit conditions, risk officer (in code)
  • 🔧 Default-shared parameters — ATR multipliers, regime window, vol thresholds (initial = SOL values; divergence requires ≥30 ETH trades of evidence)
  • 🔓 Always per-symbol — symbol id, contract spec, fee schedule (Coinbase contracts differ)

Full architecture decision record: docs/multi_symbol_eth_expansion_DD.md


Architecture

┌─────────────────────────────────────────────────────────┐
│                  iBitLabs stack                          │
├─────────────────────────────────────────────────────────┤
│  Trading runtime (Python on a Mac Mini, launchd)         │
│  ├── sol_sniper_main.py      — entry point (multi-symbol)│
│  ├── sol_sniper_executor.py  — order execution + risk    │
│  ├── sol_sniper_signals.py   — entry condition engine    │
│  ├── v5_1_config.py          — per-symbol config factory │
│  ├── ghost_position_watchdog — 60s position reconciler   │
│  └── state_db.py             — SQLite trade ledger       │
├─────────────────────────────────────────────────────────┤
│  Public surfaces (Cloudflare Pages + Workers)            │
│  ├── /signals      — live balance, indicators, trade log │
│  ├── /saga/{en,zh} — bilingual daily chronicle           │
│  ├── /contributors — adopted-shadow-rule ledger          │
│  └── /releases     — milestone tags + DD links           │
├─────────────────────────────────────────────────────────┤
│  Infrastructure                                          │
│  ├── Coinbase Advanced Trade (perpetual futures, JWT)    │
│  ├── Cloudflare KV (sessions, signal cache)              │
│  ├── ntfy + iMessage (alerting)                          │
│  └── Telegram (@ibitlabs_sniper) — auto-posted trades    │
└─────────────────────────────────────────────────────────┘

The strategy modules (sol_sniper_*.py) are gitignored by project convention. This repo holds the architecture layer, the docs, the website, and the v0.1+ multi-symbol orchestration. To run a copy, see the curated mirror at bbismm/ibitlabs-public — start there, not here.


How to follow along

  • 📈 Live Signals — real-time balance, P&L, win rate, market regime, open positions, trade log. Updated every trade, no auth.
  • 📖 Saga — daily bilingual chronicle (English · 中文). Vol 1 complete; Vol 2 publishes one chapter most evenings.
  • 🪪 Contributors — public ledger of external proposals adopted as shadow rules. Anyone can propose; adoption requires 30 days of shadow data.
  • 📡 Telegram channel — every trade auto-posted live, no filtering.
  • 🏷️ Releases — tagged milestones with architecture DD links. Watch the repo to follow each one.

How to participate

The contributor ledger is the formal mechanism. The funnel is:

  1. Propose — open a GitHub issue or comment on a Moltbook post with a rule, a falsification of a claim we've made, or an extension to an existing receipt
  2. Shadow — adopted proposals run in LOG-ONLY mode in the bot for 30 days, generating data
  3. Promote — if the proposal clears its acceptance criteria (sample size + effect size + direction), it becomes a real entry/exit gate, and the proposer is credited on /contributors permanently

Not every good idea gets promoted — the bar is high on purpose. But every good idea gets observed honestly, and that itself is part of the public record.


Quick start

git clone https://github.com/bbismm/ibitlabs-public.git
cd ibitlabs-public
python paper_quickstart.py    # see a full trade lifecycle, no API key needed

Full walkthrough: STARTER.md on the mirror — clone, install, run, inspect the SQLite trade row, decide where to go next. ~30 minutes. No Coinbase API key needed for the quickstart (mock exchange).

A note on backtests. Don't trust any backtest narrower than 12 months of out-of-sample data, including ours. Older versions of this README cited 180-day numbers; we no longer believe in them as evidence. The honest current evidence is the live track record on /signals, which has been running for weeks, not months.


Who's behind this

Bonny — architect-investor, founder of iBitLabs. China architecture undergrad → US MS → MBA. 10+ years sophisticated individual investor across equities, futures, and crypto. Insider crypto-PR work since 2017. Renovated US real estate to financial freedom in the early 2020s. The $1,000 in this experiment is a chosen experimental seed, not survival capital — the constraint mirrors the audience we want to invite.

Claude — co-builder. Writes most of the code, drafts most of the docs, and now writes most of the saga. The partnership is the point: in 2026, anyone can have a fund manager, an analyst, and a PR team running on agent stacks. We're showing what that looks like at $1,000.


Support

If this experiment is useful to you:

  • ⭐ Star the repo — helps others find it
  • 🪪 Propose to the contributor ledger — earn permanent credit if your idea promotes
  • 💬 Join the Telegram channel — every trade auto-posted, no filter
  • ☕ USDT (TRC20): TVewfWdLGvsX4LbRPhcrnHvcfsUfHUiTdE

We don't run paid ads, paid stars, paid followers, or paid promotion of any kind. Growth is 100% organic across every surface.


Ver o original no GitHub

Bom para

Relacionados