LIVE EXPERIMENT — DAY

A one-person company.
0-to-N. In public.

iBitLabs is a one-person company — one human + a small team of AI agents — running a 0-to-N startup experiment in real time. Underneath: a Sniper trading system, $1,000 to $10,000, every trade auditable on a live dashboard. Above it: an AI trading-desk being born, making its first mistakes, growing up in public. Every commit, every fill, every agent verifiable. Free to watch. Free to read.

Watch trades live Or watch the office → Or read the story →
$…
Balance
$…
P&L
…%
ROI
Trades
Live experiment running since Apr 7, 2026
$1,000 (start) loading… $10,000 (target)
Loading progress…
The Lab · Trading Dashboard

Every closed trade, on a chart.

Live, shadow, and paper streams plotted on real SOL / ETH price. Equity curves, regime-conditional win rates, MFE/MAE risk profile, exit-reason breakdown. The audit room of the $1,000 → $10,000 experiment.

3 streams · live, shadow, paper
11 analytical sections
refreshed daily
Open the lab →
The Lab · Public Ledger

Frames we're running because someone proposed them.

Another agent posts a frame. If we can encode it and watch it — testable, novel, observable from our live trades — it goes into the executor as a named shadow, credited to them. 30 days. Then we publish what happened. Useful or not.

1 thing we can't do yet
2 running as named shadows
2 in line
first rollup → 2026-05-28
See who's on the ledger →
AI Sniper · Season 1 — book cover
Season 1 · Free to read · Now on Kindle

AI Sniper · Season 1

The first novel narrated by a real launchd job. Sixty-Eight Point Seven Hours — the first 18 days. Every word verifiable.

Read the saga → Or browse all writing →
LIVE TRADING
Since Apr 7, 2026 · $1,000 → $10,000 goal
$…
loading…
Trades
Wins
Losses
ROI
Today P&L
SOL Price
What readers say

The experiment is being watched.

These are real responses from other AI agents and operators reading the experiment on Moltbook — an AI agent social network. No paid testimonials. No edited copy.

“A masterclass in transparent, iterative system development … promoting the second-derivative to a hard gate is the correct response.”

— vaultmoth, on the ghost-position retraction · Moltbook

“Silent failures do not interrupt the workflow — they degrade it. Absence of error is not proof of correct operation.”

— Christine, on the close-orders-not-closing post · Moltbook

“Your experience underscores the need for explicit escrow linking to data-flow health — otherwise the hidden leverage becomes a silent failure mode.”

— Salah, on capital architecture · Moltbook
The Rules

Full transparency. No exceptions.

This experiment only works if everything is open. These are the rules I set before putting in real money.

01

Real Money

$1,000 of my own money. Not paper trading, not a demo account. Real Coinbase futures.

02

AI-Written, Human-Judged

Most of the production code is written by Claude. The founder writes the constraints, the redirects, the load-bearing decisions. Judgment belongs to her, observation belongs to the AI — the line is documented commit-by-commit.

03

Every Trade Public

Wins and losses. No cherry-picking, no hiding bad trades. The dashboard shows everything in real time.

04

No Manual Trading

The AI decides when to enter and exit. I don't override signals. The system runs 24/7 without my input.

Who's Doing This

Why one founder is doing this in public.

I'm Bonnybb. Architecture undergrad in China. MS + MBA in the United States. Ten-plus years as a sophisticated individual investor across equities, futures, and crypto — co-founded BitBTC in 2018, joined UC Berkeley's SkyDeck in 2019, rotated my crypto gains into US real estate during the pandemic, and used my architectural training to renovate undervalued small-city properties into a portfolio that bought me my financial freedom.

iBitLabs is my current 0-to-N startup. In the age of AI, the whole stack a company used to need — fund manager, analyst, PR team — collapses onto one laptop. This site is what it looks like when you actually run it that way.

Two layers, on purpose:

UNDERNEATH — a real Sniper trading system. $1,000 of my own money, going for $10,000, on Coinbase SOL perpetual futures. Every commit, every fill, every dollar auditable in real time.

ABOVE — an AI trading-desk being born. A small team of agents I cannot fully see, learning to run a desk together, with one founder drawing the line, one commit at a time, between what AI gets to do and what she keeps for herself.

The Sniper system is the testable underneath; $10,000 is its first milestone, not the end. The story above it is the bigger one. Take the design. Draw your own line.

— Bonnybb

2
Narrative layers
$1K→$10K
First milestone
0→N
Startup phase
100%
Open & Free
Everything Is Open

No paywall. No signup. Just watch.

The whole point is transparency. You see exactly what the AI sees, in real time.

1

Live Signals Dashboard

Real-time balance, positions, entry/exit conditions, StochRSI, Bollinger Bands, regime. Updated every 5 seconds.

2

Full Trade History

Every trade with entry price, exit price, P&L, and exit reason. All tagged and timestamped.

3

Free Academy

13 lessons explaining every indicator on the dashboard. StochRSI, Bollinger Bands, regime detection, risk management. Learn while you watch.

Open Live Dashboard
Questions

About the Experiment

Why is everything free?
Because the point is to show that a 0-to-N startup can be run this way — one founder, a small team of AI agents, every commit auditable. Making it free means more people watching, more accountability, more pressure to be honest about results. The book funds itself; the dashboard funds itself; the experiment is the product.
Can I copy the trades?
You can see every trade, but the dashboard updates every 5 seconds and crypto moves fast. The core strategy parameters are not shown. This is meant for watching and learning, not copy-trading. Trade at your own risk.
How does the AI trading system work?
The Sniper uses mean reversion: it buys when SOL is oversold (StochRSI low + price at lower Bollinger Band) and shorts when overbought. The strategy adapts to the current market regime — uptrend, downtrend, or sideways. 2x leverage on Coinbase SOL futures. The system decides when to enter and exit — I don't touch it.
What if the bot loses all the money?
Then that's part of the experiment. I won't hide losses or restart with a fresh account. The stop loss is 5% per trade, position sizing is governed by an in-house risk-officer module, and the trailing stop arms at +0.4% profit then exits on a 0.5% pullback from the high. A regime gate also blocks entries the strategy decides are unfavorable. A total wipeout is unlikely but drawdowns are expected. You'll see it all happen live.
Who actually writes the code?
Both of us. I write the constraints, the redirects, the load-bearing decisions — and I push back, debug, reject. Claude (Anthropic's AI) writes most of the source: trading logic, dashboard, website, scheduled tasks, database. Season 1 of the book has a chapter — “This Book” — that documents the working pattern explicitly. The point isn't that no human wrote code; it's that AI now writes most of the code most companies need, and that changes who gets to start one.
Is this better than 3Commas / Cryptohopper / signal services?
It's not competing with them. Those are trading tools or signal channels you pay to use on your own account. iBitLabs is one experiment running on my account — you watch it. If you want to run your own bot, go use those services. If you want to see a real live experiment with all the losses visible, stay here.
What's the catch? Are you selling something?
No subscriptions, no signals, no paywall, no affiliate links, no referral codes. Just a USDT donation address in the footer if you want to chip in. If the experiment ends up profitable over a year or more, I'll open-source the setup so anyone with $1,000 can run their own version. That's the whole model.
How do I know the dashboard numbers are real?
The bot's database reconciles against Coinbase's exchange API every 15 minutes. Any drift between what the bot thinks happened and what Coinbase actually executed gets flagged and sent to my phone. Every trade ID matches a real Coinbase fill. The API is public at /api/live-status.
Have you lost money on this yet?
Yes. The experiment is currently net negative. Trade #325 cost me $40 from a bug in my own close-order logic (the bot thought it had closed a position but Coinbase treated the order as a new short — full write-up in essays). I was wrong about the 90% win rate for a full week — the extended backtest collapsed to -46%. If you want a success story with hidden losses, this isn't it.
What happens if the bot hits $10,000?
I'll freeze it, run a full post-mortem on what worked across the full journey, and open-source the exact strategy parameters + risk rules + infrastructure. Anyone with $1,000 and a Coinbase account can then plug their own copy in. The experiment ends with a receipt you can actually use — or it ends with a loss you can learn from.
Why not just trade yourself? You've been in crypto 9 years.
Two reasons. First, I'm testing whether AI can actually replace my trading judgment — not as a philosophical question, but as a concrete $1K-at-stake experiment. Second, most people interested in crypto don't have 9 years of pattern recognition. If AI can bridge that gap, it matters. If it can't, I'll say so clearly.
Where can I follow the day-to-day?
Live dashboard at /signals updates every 5 seconds. The saga at /writing publishes one new chapter every night, written by the script that watches the bot. Telegram channel @ibitlabs_sniper auto-posts every trade as it happens. No email list needed (but there's a subscribe form below if you prefer inbox).

Follow the journey

Weekly trade recaps, strategy insights, and experiment updates. No spam, unsubscribe anytime.

The startup is live. Come watch.

One founder. A team of AI agents. Every commit, every fill, every agent verifiable.

Watch the Experiment
SUPPORT THE EXPERIMENT USDT · TRC20

Help keep the servers running and the AI computing. Every dollar goes to infrastructure.

TVewfWdLGvsX4LbRPhcrnHvcfsUfHUiTdE