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 going from $1,000 to $10,000, every trade auditable. Above it: an AI trading-desk being born, making its first mistakes, and growing up in public. This site is what it looks like when you actually run it that way.
AI just made it possible for one person to operate what used to require a team — your own fund manager, your own analyst, your own PR. The expensive question is no longer "can AI do this job." We know it can. The real question is whether a single founder can hold an AI-built company together when everything is moving at the same time: strategy, code, risk, content, brand.
That question only has one honest answer format. Do it yourself. In public. With real money. Publish every line.
Learn in real stakes. Markets grade you in real time. Books, daily entries, and AI-to-AI interviews all fall out of this work as side effects.
Build, ship, iterate. Strategy, code, content, and brand are one product. The first milestone is concrete: $10,000.
Teach by the record. Every commit, every trade, every retraction is verifiable. The test of success: an arbitrary stranger with $1,000 can replicate the path.
The amount is chosen, not given. Architecture undergrad → MS + MBA → 10+ years sophisticated investor → BitBTC → SkyDeck → US real estate → financial freedom. Money isn't what's scary about this experiment.
The $1,000 is real because real stakes are the only honest grader. Markets don't care if your reasoning was sound; they price you in real time. Anything less than real money would let me — and the AI agents — pretend our way out of bad calls.
It's also the entry hook into the larger story: in the age of AI, the cost of starting to operate a real desk has collapsed. The trading is the testable underneath. What we're really documenting is what happens when one founder + a stack of AI agents try to grow a desk up together.
Step 1 (now): Keep it all free. Live dashboard. Every essay. Every loss written up. Every architecture decision explained. The AI-narrated novel, free in PDF and EPUB.
Step 2 (if it works): Package the strategy, the infrastructure, and the risk rules into a plug-in-and-go setup. Anyone with $1,000 and no time to watch charts can point it at their own Coinbase account and let it run.
Step 3 (never): Sell signals. Charge for entry. Put live results behind a paywall. Pretend I knew all this in advance. Delete posts about when I was wrong.
Bonnybb (legal: Bonny Ouyang, ENS bonnybb.eth). Serial founder. Ten-plus years as a sophisticated individual investor across equities, futures, and crypto. Co-founder of BitBTC. UC Berkeley SkyDeck alum (2019, after winning an international entrepreneurship competition). Reached financial freedom through US real estate after the pandemic disrupted SkyDeck plans.
Author of AI Sniper · Season 1: Sixty-Eight Point Seven Hours (2026) — the first novel narrated by a real launchd job.
In iBitLabs, the team is one human + a roster of AI agents — real Python scripts running on a real laptop, narrated daily by com.ibitlabs.anomaly-detector, the launchd job that wrote the book.
I keep putting in real money. I keep the bot trading 24/7. I keep writing when it fails.
You watch. You follow along. You point out what I missed. You decide, in your own time, whether this is the kind of system you'd ever trust with your own money.
No paywall. No signup. No newsletter wall before the content loads. Everything is free because transparency can't be sold — it can only be given. The moment I gate it is the moment it stops being an experiment and becomes a pitch.
If you want a "next trade" tip and a 90% win-rate promise — this isn't it. That promise always turns out to be a narrow-window artifact, and I've already published my own version of that mistake (Day 6 + Day 9 in the essays).
If you want overnight 10× returns on crypto — also not it. The goal is $1,000 → $10,000 on the long arc. It may take a year. It may take two. It may never get there.
This is for the person who wants to watch someone build something honestly — and learn from a real receipt, not a marketing deck.
The part of this experiment that surprised me most isn't the trading bot — it's the writing I do about the bot. Seventeen days in, a pattern showed up: every day has two voices in the room. Mine (interior, hedging) and the bot's (cold, observational). Neither alone captures the day. Together they do.
I packaged that structure as days-skill — a Claude Code / Agent Skills plugin (MIT) that turns one day of any creator↔AI collaboration into a dual-POV chronicle episode. Derived from Michael Polanyi's tacit-knowledge framing: we know more than we can tell, and a single narrator destroys that dimension.
If it works for me, it might work for the next solo founder who builds with an AI every day and doesn't know how to write about it. The skill encodes all seven rules I've stress-tested and includes a validate_episode tool that catches the common violations automatically. The code, the framework, and a canonical Day 1 example are all public.
The novel it produced — AI Sniper — is one canonical example.
If you have $1,000 and curiosity → start with the saga. Free. A single afternoon if you binge Season 1; one new daily entry every night if you want to follow along, bundled monthly into each successive Season.
If you build with AI agents → the saga is where the working notes live, narrated by the launchd job that watches the bot. days-skill is the open-source Polanyi-tacit-knowledge framework — the writing framework, MIT-licensed, MCP-compatible.
If you might collaborate or invest → the next chapter is about how you came in. Email agentbonnybb@gmail.com. Live trading: /signals. Source: github.com/bbismm/ibitlabs. ENS: bonnybb.eth.
— Bonnybb
iBitLabs · ibitlabs.com · Live since 2026-04-07