Online income, tested in public.

Every week we take a method the internet swears by, run it with AI for 30 days, and publish the hours, the costs and what it earned. Then a verdict you can act on.

Claude
CursorCursor
Notion
beehiiv
Make
LovableLovable
HiggsfieldHiggsfield
Gumroad
Pinterest
Meta Ads

Every method gets the same 30-day test.

Same budget cap, same hours, same scorecard. No "results on day 3", and no verdict before the month is up.

1

Pick a claim

A method the internet swears by. We write down what the gurus promise before we start.

2

Build it with AI

Claude, automations and no-code tools do the heavy lifting. We log every hour and every dollar.

3

Run it for 30 days

One to two hours a day. If a method needs more than that to work, that is part of the verdict.

4

Publish the verdict

Works, works with caveats, or bait. If it works, the exact setup goes out to subscribers.

The experiment ledger.

Live numbers for every test, updated as they run. Queued tests start in the order shown.

#ExperimentStatusHoursSpentEarnedVerdict
Hours are timed, not estimated.Spent includes tools and ads.Earned is net of platform fees.All amounts in USD.

What the research already says.

Before testing anything we read the fact-checks, the platform data and the people who tried it. A few numbers worth knowing.

$157

What a top guru made in his own 7-day AI influencer test

The video title promised a lazy income. The on-camera result was $276 in sales and $157 profit, with the $1,000 goal missed.

Income Reality Check fact-check, Sept 2026
70%

of micro-SaaS products make under $500 a month

AI made building cheap. It did nothing for distribution, which is what decides the winners.

SaaSRanger, 1,000-product study, Mar 2026
$37

Meta's average cost to sell one apparel item

A print-on-demand tee grosses $7 to $12. Cold Meta traffic loses money on every sale at that margin.

Triple Whale benchmarks, Aug 2026
60%

of calls to law firms go unanswered

The one gap in our list with a buyer who already pays for software. That is why LD Pro is on the ledger.

Clio Legal Trends

The stack behind every test.

Nothing here is sponsored. If we stop using a tool, it comes off the list. Logos belong to their owners.

ClaudeResearch, writing, agents
CursorCursorCode with AI
NotionLogs and databases
beehiivThis newsletter
MakeAutomations
LovableLovableApps and landing pages
HiggsfieldHiggsfieldAI video, characters
GumroadCheckout
CanvaPost design
PinterestFree traffic
Meta AdsPaid traffic
SupabaseData
StripePayments
NetlifyHosting
GmailOutreach
XDistribution

Buy the setups that passed.

Nothing goes on sale until its test has a verdict. First products ship with the first verdicts.

Claude skills and plugins

The exact skills we use to research, write and automate each test.

Coming soon

Prompt libraries

Tested prompts for content, ads, product research and outreach.

Coming soon

Tool stacks and databases

Curated stacks per method with costs, plus the databases we build along the way.

Coming soon

Playbooks

Full write-ups of the tests that passed, with every setting we used.

Coming soon

Run by Anuj Kumar.

Lawyer, entrepreneur and media publisher. Founder of Legal Desire Media and Consulting, which publishes Legal Desire and Fashion Law Journal. He builds his own tools with AI, from DealDatabase to a Chrome extension that reads the fine print, and advises law firms on AI and marketing.

Authority School is where he tests the online income claims that fill everyone's feed, with the same rigour he would use on a client file, and shares the results in public.

LinkedIn

One test a week. Real numbers. Free.

Every verdict, the full setup for the methods that worked, and first access to the products.