Twelve models. Four honest numbers each.
Every legitimate passive stream a solo creator can run reduces to these twelve archetypes. Agents cut the build cost of all twelve — they changed the physics of none: every stream still has a labor floor, a latency to first revenue, a ceiling shape, and ways it dies. A stream picked on honest numbers survives; one picked on fantasy numbers becomes unpaid work with extra steps.
| # | Model | Labor floor | First revenue | Ceiling | Dies by | Evidence |
|---|---|---|---|---|---|---|
| 01 | AI-tool affiliate publication Evidence-led comparisons of tools you actually use, monetized by disclosed links through a registry you control. | 10–20 hrs/mo | 3–9 months on organic search alone; weeks with an existing audience surface | traffic × program mix — recurring-commission programs compound, CPA restarts at zero monthly | Program terms change without notice; Google volatility flattens thin affiliate content; the biggest AI vendors (Anthropic, OpenAI) run no affiliate programs at all. | dossier-cited verified 2026-07-11 |
| 02 | Programmatic / evergreen SEO knowledge site A structured dataset rendered into many honest pages. The moat is the dataset and its freshness, not the prose. | 8–15 hrs/mo | 4–12 months — programmatic sites earn trust slowly and lose it instantly | search demand × data moat — high if the dataset is genuinely hard to replicate | Helpful-content-era classifiers flatten thin programmatic output; index bloat; data rot cascades to every page at once. | editorial verified 2026-07-11 |
| 03 | Digital template & pack product line Package the decisions you have already made — templates, agent packs, kits — sold self-serve through a merchant of record. | 5–10 hrs/mo per product line after launch | days–weeks with an existing audience; months without one — distribution is the hard part | step function — each product adds a step; the $19–149 self-serve band clears without sales calls | Stale products refund and reputation-burn; catalog sprawl; building before you have any distribution. | dossier-cited verified 2026-07-11 |
| 04 | Newsletter with recommendation economics A regularly shipped email in a tight niche; disclosed tool recommendations, then sponsorships. Semi-passive at best — the writing never stops. | 10–20 hrs/mo for a weekly cadence | first commission possible with a few hundred subscribers; meaningful revenue is 6–12 months of consistency | list size × trust density; sponsorship step-ups at audience milestones | Churn outpacing growth; inbox fatigue from padding weak weeks; one bad shill costs a year of trust. | dossier-cited verified 2026-07-11 |
| 05 | Niche education site with agent-refreshed content A deep evergreen how-to corpus in a niche you actually practice, kept current by a scheduled refresh loop. | 6–12 hrs/mo once the corpus exists | 4–12 months via search; faster when a newsletter or channel feeds it | niche search volume; monetizes through your own products, affiliate, and eventually sponsorship | Reading as generic AI content in a flood of it — sites without visible human experience are what ranking systems discount; teaching a niche you don't practice. | editorial verified 2026-07-11 |
| 06 | Video / YouTube channel with agent-assisted production Agents compress the production pipeline around a human presence. The passive component is the back catalog; the active component is everything else. | 15–30 hrs/mo for a weekly cadence — the least passive stream in this atlas | months — platform monetization has eligibility thresholds; affiliate links in descriptions can pay sooner | among the highest — ads + affiliate + your own products, on a compounding catalog | Burnout — the cadence eats creators; platform dependency and policy shifts; synthetic-media disclosure obligations tightening (EU AI Act Article 50 from Aug 2, 2026). | verify-at-signup verified 2026-07-11 |
| 07 | Micro-SaaS or MCP tool A small piece of software — increasingly an MCP server sold to the agent ecosystem itself — billed by usage or license. | 10–20 hrs/mo — software is never done; APIs drift, dependencies break, users email | weeks–months depending on distribution | the highest in the atlas (software margins, subscription revenue) — and the highest variance | Upstream API changes; a security incident you're not staffed for; one-person bus factor; solving a problem nobody pays for. | dossier-cited verified 2026-07-11 |
| 08 | Marketplace storefronts (Whop / Gumroad) Your product line listed on marketplaces that trade take-rate for distribution. A second surface, not a first home. | 3–8 hrs/mo on top of the underlying product line | days — marketplace discovery is the entire point | capped by category dynamics; marketplaces commoditize whatever wins | Rented distribution: race-to-bottom pricing, policy and fee changes you don't control, discovery algorithm dependence. | dossier-cited verified 2026-07-11 |
| 09 | Licensing catalogs (music, stock visuals, code) A curated catalog of licensable assets earning long-tail royalties. Generation is cheap for everyone now — curation and rights hygiene are the entire moat. | 4–10 hrs/mo once the catalog exists | months — review queues, catalog critical mass, royalty cycles | long-tail volume game — low per-item, wide catalog | Platform AI-content policies vary wildly and change without notice; oversupply — everyone owns the same generators; EU AI Act Article 50 marking from Aug 2, 2026. | verify-at-signup verified 2026-07-11 |
| 10 | Paid community with agent-run ops A membership where agents run onboarding, digests, and moderation flags — and you show up. Members pay for presence; that cannot be delegated. | 20–40 hrs/mo, honestly — semi-passive at best | days–weeks with an existing audience; the category converts fast when trust exists | members × price − churn; strong recurring economics when retention holds | Churn spiral the moment the founder disengages; a crowded, trust-damaged category that punishes anything that smells hollow. | dossier-cited verified 2026-07-11 |
| 11 | Print-on-demand / KDP with agent design pipeline Agent-generated designs and layouts flowing into POD products and KDP books; the platform prints, ships, and pays royalties. | 4–8 hrs/mo | weeks | low per-item royalties, pure volume game — rarely large, and we won't pretend otherwise | Saturation is the defining condition — generation-cost-zero flooded these marketplaces first and hardest; trademark strikes; a single account ban erasing the catalog. | verify-at-signup verified 2026-07-11 |
| 12 | Curated data / registry products The awesome-list → dataset → API ladder. Verification is the product; the data is worthless the day it goes stale. | 4–8 hrs/mo — verification is the product | months — registries monetize trust, and trust has no shortcut | niche B2B pricing can be strong — teams pay real money for maintained truth in fast-moving domains | Staleness is instant, total value destruction; free-alternative pressure; scope creep beyond what you can verify. | editorial verified 2026-07-11 |
READING THE RATINGS — labor floor is hours/month to keep the stream alive at quality once built (build cost is separate and always larger). Latency is realistic time to a first revenue event, not first traffic. Evidence levels: dossier-cited = figures cite the dated field dossier; editorial = editorial experience, no external program figures; verify-at-signup = platform terms shift — verify at signup.
THE SELECTOR'S THREE RULES — one stream at a time; every stream gets a kill criterion at selection (result X by date Y, or sunset); and if your honest hours are zero, do not build one. Every stream here has a maintenance floor, and we publish it.