Noesa
← Home

Tells

The giveaway that AI is bluffing. 677 of them, each one a complete catch — think first, then tap.

Two usersUsing AI, day by day · Day 1

Explain why an AI recommended a specific ingredient for a recipe

Without the course

“Why did you pick olive oil for this recipe?”

Used the AI's explanation as proof olive oil was best

After this course

“Show the recipe context and tool logs that led to olive oil”

Spotted the AI invented a reason not in the source

Never trust an AI's self-report; verify against actual inputs or logs

HighlightData safety with AI · Day 1

Pasting a report into a chat is like placing a document in an addressed envelope and handing it over at the door. The handoff is real even if the reply arrives before you have time to notice the envelope left.

TellThe app you didn't write · Day 0

If you asked an AI how large your codebase is, or where a particular feature lives, what would it likely get wrong?

HighlightCatch the AI: verify before you trust · Day 1

It's like the charming guest at a dinner party who has an answer for everything — the wine, the neighborhood, the recipe, the news. Every story lands in the same warm, assured voice. Three of the stories are invented. Nothing in the voice tells you which three.

Two usersBuilding an Accounting Ledger · Day 5

Post the customer's annual payment to the ledger

Without the course

“Record the customer's $12,000 payment as revenue for this year.”

Credits $12,000 to revenue immediately upon cash receipt.

After this course

“Record the customer's $12,000 payment, specifying if service is delivered or deferred.”

Credits $12,000 to Deferred Revenue until service is earned.

Match revenue to when service is delivered, not when cash arrives.

HighlightContracts, understood · Day 1

Reading clauses before mapping the deal is like checking every stitch in a coat before asking whether it is your size.

Two usersFinance for Integrators · Day 1

Assign each billing decision to the correct team

Without the course

“Map these billing decisions to teams based on common ownership.”

Used generic owners; shipped wrong approver for refunds

After this course

“Map billing decisions to teams, then list every owner to confirm with the team.”

Flagged refund approver mismatch before encoding

Treat AI’s ownership map as a hypothesis; verify every owner

Casebook· this really happened

The chatbot that invented a refund policy — and the airline had to honour it

A support bot described a discount that did not exist. A tribunal held the airline to its words.

British Columbia, Canada · 2022–2024 · 4 sources

Read the case →
HighlightMoney, understood · Day 1

A budget you don't look at is a leaky bucket. It's not that the water's stolen — it's that you never checked the holes.