Keep the receipts

Hint-Logs

A hint-log is one short, structured entry that captures a meaningful AI interaction — and proves a human, not the machine, made the final call.

Hint-logs are the small-scale, individual version of what the AiGovOps Foundation’s Beacon tool does for enterprises — they’re the receipts side of “keep the receipts.”

They’re the “policy as code” idea, scaled down for individuals. A hint-log is how a teacher proves to a parent that AI didn’t grade their kid’s IEP — a human did. How a small-business owner shows an auditor they reviewed the AI’s draft before it went out. How a community leader proves a campaign claim came from a verified source, not an AI hallucination.

The five fields of a hint-log

Five fields. Thirty seconds. The discipline is in writing the catch and the human call honestly.

1

The hint

The prompt or question you actually asked the AI.

Example
“Draft a parent-friendly summary of Maya’s reading-goal progress for the IEP meeting, 6th-grade level.”

2

The answer

A short summary of what the AI gave you back.

Example
Produced a 3-paragraph summary with 4 proposed goals and a fluency target of 120 WPM.

3

The catch

What the AI got wrong, missed, or made up. Be specific.

Example
Invented a ‘120 WPM’ benchmark that isn’t in our assessment data, and omitted the accommodation Maya already has.

4

The human call

What YOU decided, edited, or did differently. This is the proof a person stayed in charge.

Example
Deleted the fabricated WPM figure, pulled the real number from the running record (98 WPM), and re-added the extended-time accommodation.

5

The receipt

A link to the source chat, the doc, the email — wherever the artifact lives.

Example
Link to the saved chat + the final IEP draft in the shared Drive folder, dated and initialed.

Three worked examples

🍎

A Teacher’s hint-log

The hint Summarize Maya’s reading running-records into 3 IEP goal options at a 6th-grade level.
The answer Gave 3 goals plus a fluency target of 120 WPM and a tidy parent-facing paragraph.
The catch The 120 WPM target was invented — not from our data — and it dropped her extended-time accommodation.
The human call Replaced with the real 98 WPM from the running record, restored the accommodation, rewrote one goal in plainer language.
The receipt Chat link + final IEP draft v3 in the SpEd Drive folder, initialed JT, 5/14.
🛒

A SMB owner’s hint-log

The hint Write a refund-policy section for our website based on these notes and WA consumer law.
The answer Produced a clean policy citing a ‘30-day statutory refund window.’
The catch Washington has no blanket 30-day statutory refund right — that was a hallucinated legal claim.
The human call Removed the false legal claim, reframed it as our voluntary 30-day policy, ran the final by our attorney.
The receipt Perplexity search link to the WA AG page + the attorney’s approval email, saved 6/2.
🏛️

A Community leader’s hint-log

The hint Draft talking points claiming our river-cleanup removed ‘X tons’ of trash this year.
The answer Filled in ‘12 tons’ and wrote a confident press paragraph around it.
The catch The 12-ton figure was AI-guessed — our logs showed 4.6 tons; the claim would have been false in public.
The human call Swapped in the verified 4.6 tons from our volunteer logs and linked the logbook in the press kit.
The receipt Link to the volunteer log spreadsheet + the AI chat, attached to the board minutes, 6/9.

A blank template — copy & paste

hint-log.md
## Hint-Log — [date] — [your initials]

**1. The hint:** What did you ask the AI? Paste or paraphrase the prompt.

**2. The answer:** What did the AI give back? One- or two-sentence summary.

**3. The catch:** What did it get wrong, miss, or make up?

**4. The human call:** What did YOU decide, edit, verify, or do differently?

**5. The receipt:** Link to the chat / doc / email where the artifact lives.
📄

Download the Hint-Log spec

The full spec, five fields, five worked examples, and the Beacon connection — copy-paste ready Markdown.

Get it

When you outgrow the notebook

From a personal hint-log to Beacon

A hint-log is something you keep by hand, for the AI moments that matter. When you outgrow a personal hint-log, Beacon does this automatically at the enterprise scale — capturing the hint, the answer, the catch, the human call, and the receipt across an entire organization, as policy enforced in code rather than goodwill.