How we rank and label numbers
AgentPickList ranks 158 autonomous AI business ideas by how much people talk about them on X this week. Revenue is shown only when a source states it, with a label saying who stated it; every other money figure is our labeled estimate.
Ranking: X buzz
For each idea we run one X search. We count matching posts in the last 7 days and add up likes, reposts, replies and quotes on the top matching posts (up to 25). X buzz = 10×log10(1 + posts) + 10×log10(1 + engagement). Logarithms keep one viral post from drowning everything. Posts only count if they match the idea, mention AI or agents, and are not crypto spam. 93 of 158 ideas have X data today; ideas without it score 0 and say “No X data”. Raw numbers and fetch dates are on each idea page.
Daily movement
Each weekday we save the day’s order. ▲ and ▼ show places gained or lost since the previous snapshot, — means no change, and NEW marks ideas added in the last 7 days.
Revenue labels
- Verified (50 ideas): TrustMRR says it calculated the figure from payment-provider data. We name the provider and the sync date.
- Third-party reported (8): an interview, podcast or article reported it.
- Self-reported (16): the founder’s own claim, linked.
- Estimate (all other ideas): customers × our suggested price. It is math, not a fact or a promise.
The Monthly revenue filter uses reported monthly revenue where it exists and the estimate otherwise. Totals that are not monthly are never converted.
Agent setup level
Easy (46): only no-code tools. Intermediate (72): adds an app builder, database or more moving parts. Advanced (40): custom code, APIs or MCP, mobile apps, real-time voice, or sensitive work like tax, medical or hiring. Full rubric.
Which ideas we include
Only ideas AI agents can run end to end while the owner oversees: digital products, content, data and online services. We hide ideas that need heavy manual work, a licensed professional’s sign-off or physical delivery, and anything shady (fake engagement, scraping against terms, betting). Ideas come from real posts on X, Reddit threads, public founder stories and TrustMRR listings. Every idea links its sources; summaries are in our own words.
Costs
Monthly cost is the sum of the public list prices of the tools in each idea’s stack, from free tier to paid plan, with links to each pricing page and the date we checked them.
Open data
Download the current ranking as JSON or CSV (rank, idea, how it makes money, monthly cost, setup level, X buzz, revenue and its basis), or the full source records at /api/use-cases.json. AI assistants can read llms.txt and llms-full.txt. All files update with each build.
Corrections
If a number or source is wrong, tell us through the submit page and we will fix or remove it.
Questions
How is the AgentPickList ranking calculated?
By X buzz: 10×log10(1 + posts matching the idea in the last 7 days) + 10×log10(1 + likes, reposts, replies and quotes on the top matching posts). Ties fall back to engagement on the original source and recency.
What do the ▲, ▼ and NEW markers mean?
They compare today’s rank with the previous daily snapshot (America/Los_Angeles dates). NEW means the idea was not in the previous snapshot or was first seen in the last 7 days.
What does “verified revenue” mean?
The TrustMRR page for that startup says the figure is calculated from payment-provider data (for example Stripe). We name the provider and the sync date. We do not verify it ourselves.
Are earnings estimates real revenue?
No. An estimate is simple math (customers × our suggested price) and is always labeled “Estimate”. It is not a fact or a promise.
Last updated · How we rank and label numbers
The weekly top 5.
The five hottest AI business ideas on X each week, with the steps and costs. Free.