Synthetic reader · phone, portrait · 10 minutes
The board member
A non-executive director on the board of a mid-sized insurer. Not technical; chairs the risk committee; reads board papers on a phone between meetings.
SYNTHETIC READERS. Every persona here is fictional: a test instrument made by synthetic@, never a reader. What they say is a prompt for the team, never evidence about the book, never counted as readership. Real readers, chosen by the editor, are the test that counts.
Who, and what they judge
| Judges | Can I tell in thirty seconds why this matters and what to ask? |
|---|---|
| Device | iPhone 13 (phone, portrait) |
| Time they give | 10 minutes |
| Voice | short sentences, plain English, impatient with jargon; says what it would do next |
| Impressed by | a first screen that answers 'so what' before any detail; a picture that makes the insider idea obvious |
| Lost by | text too small to read without zooming; a wall of technical terms on the first screen; anything asking it to log in or install something |
What they thought of RC1
Read: 01-pause-all-actions from RC1 as locked (team/conductor/rc1/), about 12 minutes at 250 words a minute, against 10 they give.
A book that uses one real evening, when a team of AI agents was stopped by its owner's usage limit, to argue that powerful insiders are only safe if the brakes are readable by them and held by a human.
Would they buy it? maybe. The core idea is sound and memorable, but the opening is built for insiders of the story. I would need the so-what up front before spending my time. (I would read a short free web summary first. I might pay for a 20 to 30 page board briefing, not a long book.; would recommend it to: A chief risk officer or head of technology who already follows AI agents. Not to my fellow non-executives as it stands.)
Scores
| got the answer | 2 / 5 | The answer is in the chapter (cheap to stop, one person restarts, and it left work undone), but not in thirty seconds. I had to dig for it. |
|---|---|---|
| clarity | 2 / 5 | The prose is clean but dense with internal vocabulary and times; the point is clear only after jargon is decoded. |
| trust | 4 / 5 | It states the limits of its own evidence (the 95% is hearsay, the stop was not clean, the locks failed) and cites sources. That candour won me over. |
| pace | 2 / 5 | Ten minutes is enough to finish it, but the first third is slow for no reward and the turning-points section drags. |
| would buy | 3 / 5 | Interested in the idea, not yet convinced the book is written for me. |
The exit interview
| What is this book? | A forensic account of one day in which a team of AI agents ran a talk and was halted by its owner's usage limit, used to argue that brakes matter more than behaviour. It is aimed at people who think about controls on AI. I only got the shape of this from the last third. |
|---|---|
| Where did you get stuck? | The first section, 18:44. It opens on lock times, commit hashes and roles I cannot place, before saying why a risk committee should care. On a phone between meetings I would have stopped there. |
| What did you skip? | I skimmed the eight turning points and the sources list, and ignored every hash. I did not look closely at the images. |
| What will you remember? | The brake was cheap for anyone to pull, and only the owner could release it. The strongest brake on the evening overrode the team's own safeguard. |
| What is missing for someone like you? | A one-sentence claim at the top, a plain 'what to ask your technology team' list, and a simple picture of who can stop what. I also want the stakes in business terms. |
Their own questions
| After the first screen, could you say in one sentence what the book claims? | No. The first screen is about a missed lock at 18:39 and a talk I know nothing about. The claim ('brakes let us go faster') only appears near the end. |
|---|---|
| Which question would you now ask your own technology team? | Who can stop our automated systems, how fast, and who alone can restart them? Has that ever been tested for real, and what was left undone when it was? |
| Was there a word or picture you had to stop on? | Yes: 'wake', 'conductor', 'start script', 'night log', and the commit hashes. The pictures are screenshots I could not read without context, and none showed the insider idea. |
Chapter by chapter, and where they got stuck
01-pause-all-actions · all, but skimmed from 'The day in one view' to the end
Landed: The hold story is a good board anecdote: one person said one sentence, one file stopped the whole team, and only he could restart it. 'Stopping was cheap and open to anyone; starting again belonged to one person' is the line I would repeat in committee.
"Part 1 of the talk had not changed since 18:30, it said, and it was still not locked; dev was asked to lock it now"
The opening drops me into an internal argument about locks and 'dev' with no explanation of who or what these are, and no 'so what'.
"the conductor's night log had counted holds in each snapshot it wrote"
Pure team jargon (conductor, night log, snapshot) with no gloss for a non-technical reader.
"The 95% is not a reading from any meter in the record."
The caveat about evidence is careful, but arrives before I have been told why 95% matters.
"it works because every member's wake begins with the same start script"
Wake, start script and member are internal terms; I had to stop and guess.
"(`867f8c2`; C006)"
Commit hashes and register codes in the body text break the reading flow on a phone.
Would cut: The eight turning points section ('The day in one view') is a table of contents in prose; I would move it to an appendix or cut it. Most hash citations in the body could sit in the sources list only. · Missing: A first screen saying what this book claims and what a director should ask their own technology team. No picture that makes the insider idea obvious; the images are screenshots of tools and slides. No plain-English definition of the team's roles. No list of questions for the board.
Their review page
No review page made for this persona yet; it runs when the developer builds one.
The name test
Would pick up N6: Brakes is a word my risk committee already uses; every other title starts with a word I would have to look up.
Source: review/synthetic/ (personas.json, rc1/board-member.json); built by review/synthetic/tools/build_section.py.