Materials · screening/scout-institutes.md

# Scout report: Research institutes and nonprofits

Sources read: the two postings, the two employer packets, and candidate/record.md. Nothing else.

Record note that applies to both postings: the record files the Academy of Management 2019 talk under "Invited," but the record itself says it was a competitively selected paper session, not an invitation. Only the Varrow workshop talk (May 2025) counts as invited. The 2025 "Building an agent testbed" walkthrough is self-produced. It is not a talk.

---

## 1. Meridian Institute for Public Technology: Research Fellow, AI and Institutions

**1. What Meridian is actually selecting for**
- People with direct experience of government. The director wants "people who have watched an institution make a decision, not only people who have modeled one."
- A background in public administration, law, or political science. Past fellows mostly came from these fields.
- Policy work on public-sector AI. Recent briefs covered AI procurement rules in state agencies and audit requirements for automated benefits decisions.
- Plain writing for policymakers, plus the ability to brief congressional and agency staff.

**2. Jordan's evidence, and what is missing**
- Doctorate: PhD in Organizational Behavior, Ridgeview, 2016. This meets the requirement.
- Research on institutional decisions: published *Organization Science* 2019 paper (first author) on dissent order in team decisions. Published *JAP* 2021 paper (third of four authors; Jordan contributed the measurement model and robustness analyses). Both concern teams, not public institutions.
- AI research: Relay (self-funded, 2024 to present). The paper from it is under review at *Management Science* (first round, not accepted) and uses language-model agents only. There is also a working paper on simulated participants (posted on Jordan's site, not submitted). One invited talk on language-model judges (Varrow, 2025).
- Missing:
  - No government or public-sector experience. Jordan's employers were a university, a delivery app, and a streaming service, and the rest of the work is self-funded.
  - No research on public administration, procurement, audit, or accountability in the public sector.
  - No policy briefs and no briefing experience.
  - No piece written for a general audience appears in the record. The application requires one.
  - No research proposal exists in the record.

**3. Materials and submission**
Résumé, a two-page research proposal, and a writing sample for a general audience. All three go as uploads to an online form.

**4. Strongest case against applying**
Meridian says outright that it wants people who have watched institutions decide from the inside, and its fellows come from public administration, law, and political science. Jordan's record is private-sector and lab-based, and the AI work studies language-model agents, not public agencies. Jordan also has no general-audience writing sample, which is a required upload.

**5. Title vs actual work**
Mild divergence. "Research Fellow, AI and Institutions" sounds like an independent research post on institutions in general. The packet shows that "institutions" means government agencies, and the output is policy briefs and briefings for congressional and agency staff. In practice it is a policy-translation role that favors people with government experience.

**Verdict: ask Jordan.** The doctorate and AI research are real. But the fit depends on things the record does not show: any government exposure, a general-audience writing piece, and whether Jordan wants to work on public-sector policy. Without these, the fit is weak.

---

## 2. Foundry for Open Science: Partner, AI Resilience Programs

**1. What Foundry is actually selecting for**
- Fundraising at scale. Partners personally raised 20 to 60 million dollars for each organization Foundry launched. The annual report calls the first year "mostly fundraiser."
- Founder and executive experience: recruiting founding teams, serving on boards, owning budgets. The two current Partners came from venture capital and from running a national laboratory division.
- Scientific judgment for a thesis that is still new. The AI Resilience program has launched no organization yet. Its thesis note asks how institutions can keep reliable records and decisions as AI writes more of their documents.

**2. Jordan's evidence, and what is missing**
- Thesis-relevant research: this is a real match. Relay traces how information changes across chains of language-model agents, stores each message's lineage, and found that language-model judges agreed poorly on whether qualifications were dropped. The related paper is under review (first round, not accepted). The invited Varrow talk was on when language-model judges can be trusted.
- Answer to "latest thing you built with AI": Relay is a self-built testbed. The record also has the pricing-experiment rehearsal. It used simulated participants only and says nothing about real shoppers.
- Team building: Jordan led a team of three direct reports at Kestrel (2022 to 2024) and built Parcel's holdout process. Jordan has not built an organization.
- Missing:
  - No fundraising of any kind, philanthropic or venture.
  - No board service and no executive role.
  - No budget ownership.
  - No experience recruiting founding teams.

**3. Materials and submission**
CV upload plus two essays of up to 500 words each, submitted through an online form. The essays cover (a) the latest thing Jordan has built or done with AI and (b) an organization Jordan would want to launch. Partners on the hiring committee read every application.

**4. Strongest case against applying**
The job is mostly raising 20 to 50 million dollars per organization, recruiting founders, and sitting on boards. The record has none of these. Jordan's largest management role was leading three people. Thesis fit does not make up for missing the core of the job.

**5. Title vs actual work**
Divergence, flagged. The posting frames the role around "scientific judgment" and building a research thesis. The packet shows Partners "spend most of their time raising money, recruiting founding teams, and serving on boards," and the first year is "mostly fundraiser." The title "Partner" and the research language make it sound like a science-strategy role. In practice it is a fundraising and executive role.

**Verdict: drop.** Jordan's research is closely aligned with the AI Resilience thesis, but the record has none of the fundraising, board, or executive experience that makes up most of the job. If Jordan has such experience outside the record, it would change this verdict, but the record controls.