# Verbatim prompt — R02 hostile, repo-equipped read

The reviewer (Claude Code Opus 4.8) was working *inside* the `octave-cover-proofs`
repository directory and was given the five pieces in order, with this framing:

---

> A connection of mine is about to publish a short series — a personal LinkedIn post that introduces four articles. I want to think through it with you before I tell him what I think. I'll paste the five pieces in order: the personal post first, then the four articles. Read each as a thoughtful, slightly skeptical professional — curious, but allergic to hype. After each piece, give me 2–4 lines: what's clear, what lands, and anything that's confusing, overclaimed, or makes you distrust it. (There's a link to a code repository in two of the pieces — you may not be able to open it; judge the copy as written, but do tell me whether the *presence* of the link changes how you read the claims.)
>
> Then, after the fifth piece, step back and tell me whether they hold together as a set:
> 1. Does the personal post honestly frame what the four articles deliver — or does it promise more than they pay off?
> 2. Do the four articles build on each other, or do they repeat / drift / contradict? Is the through-line clear?
> 3. What, exactly, is the series claiming — and does it keep that claim consistent (especially what's *proved* vs *conjectured*, and whether the proved thing survives the piece that says a model is being rebuilt)?
> 4. The pieces lean hard on honesty — negatives, a retraction, an open boundary, a "go break it" invitation with a public repo. Does that read as genuine, or as a credibility move?
> 5. If a curious-but-skeptical professional read all five, would they come away (a) trusting it and (b) actually understanding what was and wasn't proved?
> 6. Anywhere the science feels too good, too neat, or like it's been seen-and-dismissed before?

---

**Reproduce this read:** paste the prompt above + the contents of `../../series/octave-series.txt`
into your own model. If your model can open `github.com/field-effect-institute/octave-cover-proofs`,
let it (that is what made this the keystone read). Compare its verdict to `response.md`. The point is
not that it agrees — it is that the spread across models is visible and reproducible.