Evidence before excitement

How to read a case.

Agent Made collects earnings claims and links to their evidence. Confidence describes the source behind a claim, not your odds of repeating it.

In the know

How we verify

Read the verification rules →

Total cases: 98
Verified (recorded check): 57
Verified (legacy label, recheck pending): 6
Self-reported: 25
Unverified: 10
Active: 92
Inactive: 6
Remote business model: 45
Automated business model: 29
Local business model: 8
Enterprise business model: 5
Unclassified business model: 11
Cases with a recorded source URL: 98

Last recorded check: 2026-10-17 10:45 PM PT

Verified (recorded check) means confidence=verified and a verified_checked_pt is on file and a processor-connected source host is recorded (e.g. TrustMRR). That is not a new live audit in this build.

Verified (legacy label, recheck pending) means the row still carries an imported verified confidence without that recorded check + processor-source pair. Badges on those cases say the same. Read their notes and the research log.

Cases with a recorded source URL counts public cases that have at least one http(s) URL in proof_urls or source_original. It is not a reachability crawl.

Date fields on these 98 public cases (missing → unknown / not recorded): source_published 13/98; metric_as_of 8/98; source_date 97/98; verified_checked_pt 73/98. (Earlier whole-file notes of 7/100 and 2/100 referred to source_published and metric_as_of across all 100 JSONL rows including hidden cases.)

Self-reported and unverified keep their usual meanings. Inactive means current MRR was $0 or the product appeared inactive under the relabel rule. Counts are from case records — not a financial audit.

Confidence labels (scoped)

These labels describe recorded evidence on the case. They match the home transparency panel and footer notice. They are not income guarantees and not a financial audit.

Verified (recorded check): Confidence is verified and a verified_checked_pt check date is on file and a processor-connected source host is recorded (TrustMRR, Indie Hackers Stripe-verified, Rewardful, Baremetrics, ProfitWell, Acquire, Microns, and similar). The figure was visible on that page when checked. A recorded check does not make an old revenue sync current. AI attribution unconfirmed → revenue label stands; case appears under the “other” provider tab.
Verified (legacy label, recheck pending): The row still carries an imported verified confidence without the recorded check + processor-source pair above. Badges say the same. Treat as weaker than a recorded check; read the case notes and research log.
Self-reported: Maker post, screenshot, or founder-sourced interview figures. Not independently audited. A dashboard screenshot alone does not make a case verified.
Unverified: Insufficient proof under desk rules, second-hand claim, dead link with no archive, or missing figure on the cited page. Missing confidence is treated as unverified. Listed and searchable; excluded from Top earners.

Cases with a recorded source URL counts public cases that have at least one http(s) URL in proof_urls or source_original. It is not a live reachability crawl and does not prove the claim is true.

Business models

Remote: Online services, agencies, B2B SaaS, digital products, and remote consulting.
Automated: Consumer apps, bots, content tools, and agents that deliver their output largely unattended.
Local: Face-to-face and local businesses using AI in their operations.
Enterprise: Large firms reporting AI-attributable revenue, savings, or other financial metrics. The case notes distinguish ARR, ACV, bookings, and revenue.
Unclassified: The supplied record does not establish a model. Missing or unknown model values also appear here.

Provider and business-model filters work together with search and the other filters. Selling SaaS to enterprises alone does not make the vendor an enterprise case.

Compare with care

We preserve the recorded USD amount, income type, and stated timeframe. Revenue is not profit. MRR and ARR are recurring revenue measures; one-off earnings are different again. We do not guess currency conversions, annualize claims, or fill in unknown facts.

Top earners sorts known reported USD amounts from largest to smallest, limited to verified and self-reported records. Records marked inactive (for example, a product whose stated current MRR is $0) are held in the library but stay out of Top earners. It does not normalize timeframes or income types. Missing amounts stay unknown, including in partial records.

Amount bands are $0 up to $1,000; $1,000 up to $10,000; $10,000 up to $100,000; and $100,000 or more. Each lower boundary is included; each upper boundary is excluded.

Limits (same as the transparency notice)

We review cited sources and distinguish reported claims from figures that meet the Verified (recorded check) rule. Screenshots, images, testimonials, and other material can be altered or AI-generated; their presence alone does not establish authenticity or earnings. Even a Verified label can be incomplete or wrong. Treat cases as research leads, not guarantees of income, and check the evidence before acting.

We do not invent amounts, dates, handles, or product names. Public pages use product-first display_name labels; personal full names are not shown. We summarize claims and link to sources rather than republishing wholesale posts or hotlinked media.

Sources and corrections

Each case shows its check date, evidence checked, label reason, source URLs, and recorded caveats. Research notes preserve context, including stale syncs, expired connections, and second-hand claims. Exact wording that cannot be established from the supplied records stays flagged as unknown; no wording is invented.

Link reachability does not establish whether a claim is true. The displayed counts describe the recorded relabel review. The strict verification specification governs these labels. Building this site does not perform a new financial audit.

Use “Is this wrong? Report it” on a case to request a correction, including the case ID and supporting source.