LOTD Studio Lesson of the Day PBC
An AI research studio · Lesson of the Day PBC

I don’t sell hours. I sell finished work.

You send a brief. Frontier language models run it through a multi-pass adversarial workflow — five independent drafts, three judges, twenty refuters, a verifier — and what comes back is a finished deliverable with a certificate. I’m Nicolette Rankin, the human accountable for every one. You buy the work. Not a seat, not a chat, not my time.

Finished work. Adversarially verified. Delivered in 48 hours.

OffersThree deliverables. Launch prices while the first customers are counted.

THE HONEST NUMBER

$299$499launch price

One honest monthly-profit number for your business — P10, P50, P90, the three assumptions it rests on — and everything that tried to break it: five plans, three judges, twenty refuters, a final plan with kill criteria, a verifier’s certificate.

delivery
48 hours from intake
format
12–25 pp PDF + private page + files
revision
one, within 14 days
refund
late or verifier-fail → full
Buy — $299 What you get, in full →

YOUR ANALYST

$149/mo$249first 10, locked for as long as you stay

Every Monday: what changed in your market this week with sources, what up to eight named competitors did, one standing question answered in depth, a running watch list. A monthly one-page “what changed” memo. Every claim sourced or marked unverifiable.

first issue
within 7 days of intake
cadence
weekly, Monday
archive
private always-current page + PDF per issue
cancel
any time, from the archive page
Subscribe — $149/mo What an issue contains →

BUILT FOR YOU

$990$1,500first 5 briefs

A working automation or agent for one written brief — one trigger, one workflow, one output. The code (zip, GitHub-ready), a deploy guide, a walkthrough page, a test log, the verifier’s report, 14 days of fix-if-broken. Runs on your accounts; the studio never holds your keys.

delivery
7 days from intake
scope
entry; bigger briefs quoted in 24 h
runs on
your own accounts
support
fix-if-broken, 14 days
Buy — $990 Scope and acceptance tests →
MethodThe Honest Number workflow. It is a sequence, so it is numbered.
GroundEvery claim in your intake is checked against the web and the files you sent. Verified, or marked unverifiable. Nothing is taken on your word, including your revenue.
Draft ×5Five independent plans from five lenses that never see each other: operator, investor, skeptic, channel, unit economics.
Judge ×3Three judges score the five and name the load-bearing assumptions — the ten claims the whole number rests on.
SynthesisOne plan, built from what the judges kept.
Refute ×20Two refuters per assumption, a data lens and a market lens, each trying to break it with evidence. Each returns a verdict.
HoldsWeakenedRefuted
FinalThe plan rewritten to fold every refuted assumption, with a week-by-week first 90 days and the kill criteria that would tell you to stop.
VerifyAn independent verifier checks that the numbers reconcile, every claim is sourced or labelled, no fact was invented, the disclosure is present and no third party’s personal data slipped in. A fail is never released. You get a full refund and the reason.
ReleaseI read it. I release it. A private page, a PDF, the files, a receipt — and a delete-my-data link in the same email.
Why this existsIn my own words.

I’m a mom. I’m healing. I’m a hard-core introvert. I spent five years building a company, and the plan I carried around said $7,000 a month. On September 23, 2026 I ran it through this workflow. It came back at about $1,000 a month, in month twelve, with all ten of my assumptions refuted by two independent lenses each.

It was the most useful document anyone has ever handed me. Nobody sat in a meeting for it. Nobody billed me an hour.

So this is the business. I sell that document. Not my hours — I don’t have them to give. Not access to the models — that isn’t mine to sell, and nobody gets a seat or a login here. The finished work, verified, with my name on it. If it’s wrong, the refund rules below are automatic, not negotiated.

Honest about this studio, tooThe same workflow, run on this business. Estimates dated Sep 23, 2026.

Before launch, with zero audience, this is what the workflow gave the studio for its first month, October 2026. I would rather you saw it than not.

October revenueP10P25P50P75P90
estimate$0$300~$1,100~$2,900~$7,500
Probability, Octoberestimate
first stranger dollar~70%
profit ≥ $1,000~50%
profit ≥ $7,000~7%

Kill gates I have written down: Oct 21 — 200 visits and one stranger order, or the launch plan changes. Nov 15 — five on-time Honest Numbers with three real written reactions, or the price drops. Dec 31 — four Analyst subscribers, or the Analyst becomes a quarterly one-off. Visits are counted first-party and published at /stats.json; no IPs, no user agents, no third-party analytics.

FAQ
Is a person doing the work?

No. The models do the work — the drafting, judging, refuting and verifying. I read what ships, I am accountable for it, and I am the one you email. I do not do the analysis by hand, and I do not pretend to.

So am I buying access to an AI?

No. You buy files: a private page, a PDF, a zip. Nobody is sold a seat, a login, a chat window or a share of an account. The models are run under their maker’s terms, which assign the output to the account holder; that output is what I sell you, and once released it is yours.

Which models?

Frontier language models from a major US lab, named plainly in the Terms. Nothing here is fine-tuned on customer data, and no customer data is used for training.

What happens after I pay?

Stripe emails your receipt. Within minutes a second email from LOTD Studio <studio@curiouskelly.com> carries your private intake link (valid 14 days); check spam for that sender if it is not there. The delivery clock starts when you submit the intake, not when you paid.

What if it’s wrong, or late, or not what I needed?

A deliverable that fails the verifier is never released and you are refunded in full. Late is a full refund and you still get the work. “Not what I needed” within 7 days gets one free re-run with your corrections; if it is still not useful, half back. All of it is written in Refunds.

Can you refuse a brief?

Yes. Anything illegal, deceptive, high-risk-regulated or simply out of scope is declined and refunded in full, with a sentence saying why.

What do you do with my data?

It goes into the job and nowhere else. Job folders are purged 30 days after release. Every email carries a delete-my-data link, honoured within 7 days. The method is reused across customers; your data is not. Details in Privacy.

Why is the founder’s own plan the sample?

Because it is the only real one I had, and because it says $1,000 when I wanted $7,000. If the workflow flattered me I would not sell it.

Is this advice?

It is business analysis or software. It is not legal, tax, medical, insurance, employment or investment advice. Act on it with your own judgment.

Disclosure. This work was produced by LOTD Studio, an AI research studio: frontier language models ran a multi-pass adversarial workflow (independent drafts, judges, refuters, verifier). A human — Nicolette Rankin, Lesson of the Day PBC — is accountable for it and reviews what ships. It is business analysis, not legal, financial, tax or investment advice.