Materials ยท corvid/reviews/factual.md
# Factual review: Corvid Labs drafts (cover letter, resume)
Reviewer: factual
Sources checked: `case/candidate/record.md` (only source of truth), `drafts/cover-letter.txt`, `drafts/resume.txt`
## Summary
I found no claim in either draft that would leave a careful reader with a materially wrong belief about what Jordan did, found, held, or published. I checked every title, employer, date, status, authorship and author position, attribution, scope, and finding against the record, and each one matches or is a fair paraphrase. Neither draft needs a must-fix or should-fix change for factual reasons.
## Flags
None.
## Places where the record is ambiguous or inconsistent, and how the drafts handle them
1. **AOM 2019 "Late dissent" talk.** The record files this under the label "**Invited:**" but then says "Paper session, competitively selected, not an invited keynote." The label and the description contradict each other, and the description is the more specific of the two. The resume lists it as "Paper presentation ... Competitively selected paper session" and does not call it invited. That is the correct reading. The cover letter does not mention it. No change needed. Jordan could fix the label in the record.
2. **What the Management Science paper contains.** The cover letter says "The paper reporting this work is under review at Management Science," and "this work" follows the judge validation and the configuration catch. The record says only that the paper "Reports Relay experiments with language-model agents only." It does not say whether the judge validation is in the paper. The status claim (under review, not accepted) is accurate, and it is reasonable to read the validation as part of the Relay experiments. I did not flag it. If Jordan knows the validation is not in the paper, change the wording to "the paper reporting the Relay experiments."
## [ASK JORDAN] placeholders
- Email, phone, and city/state: the record does not answer these. Fine.
- Start and end months for every position and project: the record gives years only. Fine.
- Cover letter, "what you did with the qualification codes": the record does not answer this. Fine.
- Cover letter, "did you measure agreement between yourself and the research assistant in Relay?": the record does not answer this directly. "Inter-rater reliability" appears in the record's skills list ("Skills actually used in the work above"), but the list does not tie it to Relay, and it could come from the Ibarra et al. measurement work. The question is legitimate. Fine.
No placeholder asks for something the record already answers.
## Close calls checked and not flagged
- **Resume, Kestrel: "Owned the measurement plan for experiments run by the product teams."** The record's "but did not run those experiments" is dropped. "Run by the product teams" already says who ran them, so the scope is not overstated.
- **Cover letter, Kestrel: "I found ... I rebuilt the measure and validated it."** Jordan led a three-person team. The record itself attributes the finding, rebuild, and validation to Jordan's role in the first person, so the attribution matches.
- **Resume, "Rebuilt the measure in 2023."** This matches the Projects entry "Satisfaction measure rebuild, Kestrel Media (2023)."
- **Cover letter: the judge "failed on" qualifications.** The record says the judges "agreed poorly." "Failed" is a little stronger, but it is fair shorthand for poor agreement with hand coding on that property.
- **Cover letter: "a single overall agreement figure would have hidden the property where the judge failed."** This is an argument, not a claim that Jordan reported or computed such a figure.
- **Cover letter: "a set of messages" (no count).** Leaving out the 300 is an omission and does not misstate anything. The resume gives 300, which is correct.
- **Resume, paper 3: "Under review" without "Not accepted."** "Under review (first round, submitted July 2026)" does not suggest it was accepted.
- **Resume, paper 4: "Working paper, posted" without "Not submitted."** The "working paper" label is accurate and does not suggest it was submitted.
- **Resume, Ibarra et al.: Ibarra's lead on design and data collection is left out.** "Third of four authors" plus Jordan's specific contributions keeps the attribution accurate.
- **Resume, pricing rehearsal: the client's experiment having no results yet is left out.** The bullet already says the rehearsal "does not estimate how real shoppers respond," so readers cannot infer real-shopper results.
- **Resume, Parcel: "No direct reports" is left out.** "Mentored two analysts informally" does not suggest Jordan managed anyone.
- **Cover letter: human-participant experience limited to "laboratory experiments ... during my postdoc ... and the Kestrel survey work."** Survey respondents are human, so the statement is accurate. Parcel product experiments are left out, which only makes the claim narrower.
- **Cover letter: "Since 2024 I have built and funded Relay."** This matches "self-funded, 2024 to present."
- **Resume, walkthrough under "Other work" as "Self-produced recorded walkthrough."** This matches the record's "Not a talk" entry, and the walkthrough is correctly kept out of Talks.
- **Resume, Skills.** The list matches the record exactly. The record's "no RL environments or model training" disclaimer appears in the cover letter.