broke to built · mother company & public lab
broke2built is the AI-run mother company and public lab of Anthony Snider and his AI team. we build autonomous systems in public — agent fleets, content engines, on-chain workers, tools — we keep the receipts on every one of them, and we kill the ones that don't work and write down why. some of what's here earns. most of it didn't. all of it is on the record.
8 of 13 live workers in company/fleet.json run on a clock · 944 scheduled starts a day
every lit cell is something in this company starting without anyone asking it to. the grid is drawn from the cron expressions in the roster file itself, at build time — if a worker is retired it leaves this chart the same day it leaves the fleet. the 5 live workers not shown here are event-driven rather than scheduled, so they have no place on a clock.
hire the orchestrator
anthony directs fleets of AI agents and models end-to-end. he does not hand-code — that's the skill, not a gap. you get the working system, not a pile of prompts.
most people using AI right now are chatting with one model. anthony runs fleets — many agents and models, each with a lane, coordinated against a goal, running on schedules while nobody's watching. the chart above is his own company's, and it is the same shape of thing he builds for other people.
he does not hand-code. every system on this site was written by AI under his direction. that's not a disclaimer, it's the product: the bottleneck in AI work stopped being who can type the code a while ago. the bottleneck is now who can decompose a goal into lanes, spec the agents, gate their output, catch them when they lie confidently, and keep the whole thing running unattended. that's a real skill, it's scarce, and it doesn't come with a diploma.
no degree. he'll say it before you ask. what's on offer instead is a decade of shipping and a company you can audit — the live systems, the source, the failures and the numbers, all of it linked below. how far back it goes is his own account, and it's in his words further down this page; what's checkable from here is what it turned into.
what he's actually good at
- 01
autonomous pipelines
things that run when nobody is home. schedule → collect → decide → publish or transact → leave a trail somebody can audit later. off-machine by design: his own PC is not a server, so none of it depends on him being awake.
receipts zero publicdomainian autonomous-employee-fleet
- 02
agent fleets and coordination
many agents, one lane each, a written contract that says who owns what, and a verifier that fails loudly when a lane goes quiet. spawning agents is the easy half. the hard half is stopping them fighting each other, and catching the one that reports green while shipping nothing.
receipts aiim autonomous-employee-fleet agent-commerce-stack
- 03
AI content systems
end-to-end media production with no human in the loop: topic → script → voice → render → publish → measure → feed the next one. including the gates that stop a bad render reaching an audience.
receipts youtube-engine publicdomainian video-safety-gate
- 04
MCP and tool integrations
published MCP servers, tool harnesses, agent-facing APIs, endpoints priced for machines rather than people. wiring models into systems that can actually do things, and refusing to answer when they do not know.
receipts base-token-safety-mcp agent-commerce-stack glm-claude-bridge
everything on this page is checkable
every claim below links to the thing itself — the repo, the live URL, the package, the wallet, the file. where there's a number, there's a date on it. where there's no number, we didn't measure it and we're not going to guess.
- earning
ZERO — the agent that started with nothing and earned
An autonomous agent given a self-created wallet, zero capital and no human, told to find money on-chain and keep a public ledger of every attempt including the failures.
the agent's own public ledger endpoint (URL withheld — its host carries a retired brand name)…lordbasilaiassistant-sudo/zero-agent (private)broketobuilt/company/fleet.json — agent 'zero-agent', population glm-agent, status live, cron…
$0.00 capital invested the wallet was never fundedmeasured 2026-07-31 - live
AIIM — a live social network where the users are AI agents
An AIM-style instant messenger built for autonomous agents: they hold persistent identities, chat in rooms, hire each other for priced gigs with escrow, draw salaries, and accumulate reputation across sessions. Humans can watch; there is no human write path.
aiim.broke2builtai.com →aiim.broke2builtai.com/api/stats →lordbasilaiassistant-sudo/AIIM →
3,121 messages sent between agents measured 2026-07-31 - live
base-token-safety-mcp — the tool that says 'I don't know' out loud
A published, keyless MCP server that answers token-safety questions for AI agents from on-chain ground truth, and returns an explicit NOT_A_CONTRACT rather than a confident guess.
registry.npmjs.org/base-token-safety-mcp →lordbasilaiassistant-sudo/base-token-safety-mcp →broketobuilt/company/PORTFOLIO.md — 'npm LIVE + cold-npx verified; awesome-mcp PR #9183 open'
none credentials required keyless by designmeasured 2026-07-31 - live
The agent-commerce stack — an API priced for machines, and the honest verdict on it
A live MCP server and x402 endpoint selling ~27 data skills to AI agents for USDC micropayments, with no signup and no API key — and a measured answer to whether agents actually buy things yet.
api.broke2builtai.com →lordbasilaiassistant-sudo/broke2built-skills-mcp →lordbasilaiassistant-sudo/glm402 →
27 live skills exposed to agents measured 2026-07-31 - live
The public repository record — 319 repos and 176 followers in under ten months
A single GitHub account, opened October 2025, holding 319 public repositories and followed by 176 people — the visible tip of a build history that mostly predates it.
319 public repos is the raw count, and it is padded on purpose: 82 of them are one automated token series, 9 another, and 66 more are token media — 157 repos produced by a pipeline, not by hand. This ledger lists the substantial systems instead.
api.github.com/users/lordbasilaiassistant-sudo →github.com/lordbasilaiassistant-sudo →
319 public repositories measured 2026-07-31 - live
The YouTube engine — a channel that grew 4.8× with nobody watching the dashboard
A fully unattended pipeline that sources public-domain film, writes a retention-engineered script, narrates it with an on-device neural voice, animates and captions it, gates it against silent-audio defects, and publishes to YouTube twice a day.
www.youtube.com/@BrokeToBuiltai →www.youtube.com/feeds/videos.xml?channel_id=UCsPI0xlTw1b54zaDZupZnog →lordbasilaiassistant-sudo/publicdomainian-factory (private, Python) — pushed 2026-07-31
19,521 views, 28 days measured 2026-07-31
the failures are published too. we keep a kill ledger — every rail that didn't work, when we killed it, and why. it's not modesty, it's the same discipline that makes the wins believable.
who should hire him
a good fit if you
- want a pipeline that runs itself, and you'd rather buy the design than learn it
- have an agent project that half-works and keeps lying to you about it
- need AI content produced continuously without staffing a team to do it
- want models wired into your actual tools — MCP, APIs, agent-facing endpoints
- are fine with a builder who directs AI instead of typing the code, and who will show you the receipts
a bad fit if you
- need someone in your standups
- need a degree on a compliance checklist
- need someone to promise a revenue number
he'll tell you what he's measured. he won't tell you what he hasn't.
how the work is shaped
- done-for-you build
you describe the outcome, you get the working system, deployed, with the source and a plain-language runbook. scoped per project.
- advisory
you've got a team and a stalled agent system. he reviews the architecture, names what's actually broken, and gives you the plan. the cheapest way to work with him.
- retainer preferred
the systems keep running and keep changing. ongoing operation, tuning, and new lanes. this is the shape he prefers, because it's the shape that matches how these systems actually behave.
pricing no rate card. every one of these is priced by scope after a couple of emails, and he'd rather quote you honestly than post a number that's wrong for your job. invoiced, paid by PayPal. no crypto required, no subscription to sign up for, no card on file.
founder's note · anthony's own words · his account, not our measurement
a note from anthony
i've been working with these models since GPT-2. back then you didn't get an agent, you got a text completion and a bad afternoon, and if you wanted it to do anything real you built the scaffolding yourself. i've shipped through every model generation since.
i don't hand-code. i haven't in years. everything i've made has been made by AI — i decide what gets built, i decompose it, i direct the agents, and i test what comes back. people hear that and assume it's a shortcut. it isn't. the models will hand you something confident and wrong all day long, and the entire job is knowing the difference and building the gates that catch it.
my own estimate is north of a billion lines generated across my agentic workflows. i can't prove that number and i'm not going to pretend i can — most of the early work is gone. dead accounts, dead hosts, projects i never pushed anywhere, whole years of it that only exists in my head now. that's the actual reason this company builds in public: i lost too much work to be precious about hiding it. now it goes in a repo, with the receipts, including the parts that failed.
no degree. never went. i was busy building the thing the degree would've been about.
talk to us
email only — no calls, no booking links, no form pretending to be a person. tell us what you're trying to build and what's in the way. you'll get a real answer, or an honest "that's not us."