Start with a wedge you understand

This is stage one of the five-stage founder journey — the Build Your Company overview maps the whole path if you landed here first. Stage one is a choice: what to build, chosen narrowly enough that one person can prove it, because every stage after this one runs on conversations with people you can actually reach.

One person can now cover what used to need hires

The work that surrounds a young product — outreach, content, support replies, research, reporting, the small coding jobs in between — was historically the reason for early hires. None of it is the hard part of the business, but all of it has to happen every week, and a solo founder used to drown in it or pay someone not to. Agents now do a real share of that work: they draft, research, summarize, follow up, and keep schedules that a person would forget. A company that once needed a founder plus a first marketing hire plus a part-time support person can now be a founder who reviews what agents produced that morning.

That changes what a starting founder should look for. You no longer need an idea big enough to justify a team on day one, and you no longer need to postpone starting until you can afford help. You need an idea small enough to prove alone, with agents as the capacity that used to come from headcount.

Agents lower the cost of execution, not the cost of being wrong

The honest version of that story has a second half. Agents make execution cheap, and cheap execution cuts both ways — you can now produce more of the wrong thing, faster, for an idea nobody wants. A hundred polished outreach emails about a product nobody needs are still a hundred wasted emails, only now they arrive sooner.

What agents do not supply is judgment about which problem to solve, proof that anyone will pay, or the trust of a first customer. Those still come from you, from conversations with real people, and from selling before you build. Everything in this stage and the next exists to keep cheap execution pointed at the right idea. Treat agents as capacity, never as a substitute for finding out whether the business should exist.

Run every idea candidate through the same filter

Good first ideas are rarely invented from nothing. They are recognized — a problem you have watched people struggle with, in a market you already understand, reachable through relationships you already have. Instead of brainstorming into a blank page, collect your candidates and run each one through a filter that biases hard toward what you already know:

1. Inventory the problems you have seen up close. Go through the jobs you have held, the industries you have worked in, and the tools you have been forced to use. Write down every problem you personally watched cost someone real hours or real money — the report assembled by hand every Friday, the handoff that broke every quarter, the software everyone paid for and cursed. Firsthand knowledge of a problem is an advantage no better-funded competitor can copy quickly.

2. List the audiences you can already reach. Former colleagues, clients, an industry where your name means something, communities where you are a known member rather than a stranger. Be literal and write down names and places. An average idea aimed at people who will take your call beats a brilliant idea aimed at strangers, because the next two stages of this journey run entirely on conversations.

3. Look where money already changes hands for a clumsy solution. The strongest demand signal is existing spend: people paying consultants, buying tools they complain about, or burning paid staff hours on a manual workaround. Where money already moves, you do not have to convince anyone the problem is worth paying for. You only have to be the better place for spending that already happens.

4. Cut anything that needs distribution you do not have. If a candidate only works with a large audience, an ad budget, viral spread, or an enterprise sales motion you have never run, cut it now, however much you like it. Early companies die of distribution far more often than they die of engineering, and no amount of agent capacity manufactures an audience you cannot reach.

5. Pick the wedge — the narrowest version someone would pay for this quarter. Whatever survives the first four steps is still too big. The wedge is not the platform vision. It is one sharp problem, for one specific kind of customer, solved well enough that a real person would pay for it within the next few months. The vision can wait in a drawer. The wedge is what you take into validation.

Put agents on the research legwork

Working that filter honestly takes research — what already exists, what people pay today, who the incumbents serve and neglect — and that research is exactly the recurring legwork agents do well. In Task Machine you install a playbook, a ready-made bundle of agents, workflows, and documents built for one recurring job, from the playbook gallery. The "Deep Research Brief" playbook turns a question like "how do small agencies handle client reporting today" into a sourced brief you can read in ten minutes. "Competitor Watch & Battlecard" keeps a standing picture of who else serves a candidate market, what they charge, and where customers complain. Once a wedge starts to firm up, "Positioning & GTM brief" drafts how you would describe it, to whom, and against what alternative.

Agents run on your own machine through the tama CLI and start out supervised, so their finished briefs land in your inbox for review rather than piling up somewhere you forget to look. The reading gets done while you do other things. The pick stays yours, because choosing the wedge is precisely the judgment call the previous section said agents cannot make for you.

Write the wedge down before you move on

A wedge you cannot state in one sentence is not a wedge yet. Force it into the shape "we do X for [a specific kind of person], replacing [the clumsy thing they do today]" and notice what the sentence demands: a named audience and a named alternative. If you cannot fill in who it serves, you have a technology looking for a customer. If you cannot fill in what it replaces, you have not found where the money already moves. The written sentence is also the artifact the next stage tests, so the discipline pays twice.

You are ready for the next stage when you have a written wedge statement that names who it serves and what it replaces, and a list of twenty or more specific people or places — actual names, communities, and forums — where you can reach those buyers yourself.

A wedge on paper is still a guess, and the worst thing you can do with a guess is spend months building on it. Stage two, validate the idea before you build it, is where you find out whether anyone will pay.