bad.ventures

Most ventures are bad.

bad.ventures is an agentic venture lab that builds companies. Its agents test ideas on real people, kill the ones that don't show promise, and build the ones that do.

The sieve Many ideas enter on the left. A quick screen drops most of them, a first test on real people stops most of the rest for lack of clicks, a test with real money stops most of what is left for lack of sales, and one is built into a company. quick screen first test real money dropped no clicks no sales a company The sieve Many ideas enter at the top. A quick screen drops most of them, a first test on real people stops most of the rest for lack of clicks, a test with real money stops most of what is left for lack of sales, and one is built into a company. quick screen dropped first test no clicks real money no sales a company
Many ideas go in. Strangers' clicks and money decide which one gets built.

01 · Why now

Building got cheap. Knowing didn't.

AI can draft the product, the page and the pitch in an afternoon. Finding out whether a stranger will pay for it still takes weeks.

The lab does both: it finds out first, then builds what strangers will pay for.

Studios bet on a handful of ideas a year. We're building a lab that tests them by the hundred and builds the ones that earn it.

02 · The agents

Three agents, one toolbox.

Every venture gets its own Operator; all share one Scout and one Reviewer, so the lab is built to run many ventures at once. The agent that builds a venture never judges it.

  • The Scout finds the ideas.

    It hunts for problems worth solving and pitches the best. It can't start a venture; the lab decides.

    Twelve ideas before breakfast, in love with none of them.

  • The Operator builds one venture.

    It turns the idea into a hypothesis, designs the experiment, writes the outreach, drafts the site, and keeps building while the evidence holds. It can't spend money or grade its own results.

    Loves its venture, which is why it doesn't get the last word.

  • The Reviewer keeps it honest.

    It judges whether a test was fair and whether a struggling venture should go on, and argues against every request for money. It can't change the evidence.

    Assumes every venture is bad until strangers prove otherwise.

The toolbox Services every venture's agents share: web addresses, a mailbox, sites, request and quote pages, and instruments that record what strangers did.

03 · An example

One idea, nine days.

An illustration. The idea and the numbers are made up.

IllustrationPrintable shift planners for night-shift nurses
  1. Day 1The lab writes the listing at three prices and puts it where the buyers already are.
  2. Day 92,140 views. 61 saves. 0 sales.

Killed$14 spent. Nurses like it; nobody buys it. Next idea.

Most ideas end like that, for the price of lunch. The rare one that sells gets built: its own name and site, and more money each time it proves itself, until it's a company. We're building the lab to carry every lesson into the next venture.

Tell us where this breaks.

info@bad.ventures