Portfolio vs Niche

A founder with one year of runway faces two common paths:

1. Portfolio play — ship 10–15 small web products into crowded markets, cross-link them so they promote each other, configure SEO properly, and wait for compounding growth.
2. Niche play — pick one niche, find a messy problem an AI agent can actually fix, build a POC, get 2–3 design partners paying, then raise (or don’t) from a position of strength.

These aren’t comparable bets. They’re different games — and one is much harder than it looks.

The playbook the niche play follows

1. Pick a niche with an unfair advantage. Not greenfield — adjacency. Domain knowledge, infrastructure, or distribution the founder already owns. Picking a niche cold vs. one where the rails are already built is a completely different risk profile.
2. Find a messy workflow, not a feature gap. Agents earn their keep where the problem is ugly: cross-system reconciliation, compliance checks, document-heavy back-office work. Messy is the moat — a clean problem is a feature a competitor ships in a sprint.
3. Build a POC, not a product. Just enough to put in front of real users.
4. Get 2–3 design partners paying before heavy build. The paying part is the entire filter. Two partners paying even $500/mo tell you more than fifteen SEO sites with 200 visitors each ever will.
5. Develop with them for 6–12 months. Let their workflow shape the product. Workflow lock-in is the retention story for agent products.
6. Treat fundraising as optional, not a milestone. A lean stack and low burn keep design-partner revenue in charge of the clock. Raise from strength — or don’t raise. Agent products with real workflow lock-in can be profitable at tiny team size.

Why the portfolio play is harder than it looks

The “many small products” math worked when SEO arbitrage worked. In 2026 the odds are much worse:

• Organic search traffic is collapsing. AI overviews answer the query before the click reaches the site. The channel the whole strategy depends on is shrinking.
• Crowded markets make distribution slow and expensive. Ranking via SEO in a crowded niche takes 12–24 months — if it works at all.
• Cross-linking 10–15 thin sites is a penalty pattern. Google pattern-matches it to a private blog network. The “products promote each other” mechanic actively hurts.
• Attention split 15 ways ≈ 3 weeks of polish per product. Each one ships against incumbents with years of iteration, having received less than a month of work.

And the part that settles it: the people who actually ran this playbook — the “12 startups in 12 months” crowd — almost all converged on the same outcome. One product showed signal, they killed the rest, and went deep on the winner.

Even the success case of the portfolio play collapses into the niche play. So the rational move is to start there.

The one-line version

A year of attention spread across fifteen lottery tickets in a dying distribution channel loses to a year spent going deep on one messy problem, adjacent to existing expertise, with people who pay to have it solved.