How does DEPLOY track cross-cluster talent-flow as diaspora graph?

Why cross-cluster talent-flow is editorially central to the framework

Cross-cluster talent-flow as diaspora graph sits at the intersection of acquisition history (wind-down + license-and-hire transitions trigger diaspora patterns), partnership lifecycle (brain-providers + customer relationships frequently re-form across diaspora destinations), within-entity verification scope (technical leadership identity matters at per-claim depth), and people-graph PersonCompany-edge granularity (the substrate where talent transitions actually resolve at primary-source-anchored verification depth). The verified-vs-claimed framework operates uniformly across diaspora records: each PersonCompany edge has a distinct role, a distinct tenure window, and a distinct prior/current state.

The framework operates at canonical-methodology depth, not at trade-press LinkedIn-summary depth. The product differentiator: DEPLOY's framework treats diaspora patterns the same way it treats every other verified-vs-claimed claim. Honest "where did they actually go" at PersonCompany-edge granularity beats fabricated "diaspora-as-aggregate-narrative" framing.

---

The three canonical diaspora pattern classes

Per Agent A's people graph substrate (13 batches with 10 carrying diaspora context per Agent A authoring discipline), the cross-cluster talent-flow framework operates at three canonical pattern classes. Each class has a distinct PersonCompany-edge structure, a distinct primary-source anchor, and a distinct cross-property compounding behavior.

Class 1: post-wind-down diaspora. The corporate-state-lifecycle terminates (wind-down + acquisition-absorption + bankruptcy + voluntary dissolution); founders + executives + technical leadership transition across multiple destinations. PersonCompany-edge structure: current_role=false with end_date populated + where-they-went edge resolving destination. The framework reads post-wind-down diaspora at corporate-state-substantive depth; the source entity's lifecycle terminates as independent operational state; the destination entities inherit the source's talent at primary-source-anchored verification depth.

Class 2: license-and-hire diaspora. Specific assets licensed (not transferred); team transitions to acquirer; standalone entity continues under remaining leadership. PersonCompany-edge structure: current_role=false for transitioning leadership with end_date populated + where-they-went edge to acquirer; current_role=true for remaining leadership at standalone entity. The framework reads license-and-hire diaspora at canonical boundary-case depth; per the acquisition history Project B methodology pillar, license_and_hire structure distinguishes asset-licensing from asset-transfer at corporate-state-vs-model-state separation depth.

Class 3: adjacent-employer-prior diaspora. Recurring prior employers operate as talent feeder pools for specific cohorts. PersonCompany-edge structure: current_role=true with prior-employer edge; end_date at honest-absence cap-flag if specific tenure-end not disclosed at primary-source verification depth. The framework reads adjacent-employer-prior diaspora at substrate-pattern depth; specific recurring patterns (Meta AI / FAIR as brain-providers prior-employer; Google X / Everyday Robots as humanoid + AV prior-employer; Tesla AI Day → humanoid founders as pattern) operate at substrate-completeness depth.

---

Class 1 canonical worked example: Cruise wind-down diaspora

Per Agent A's Stage 1 AV / ROBOTAXI / AV-TRUCKING cluster people graph substrate (people_graph_batch7), the Cruise wind-down operates as the CANONICAL post-wind-down diaspora worked example. The corporate-state-lifecycle: GM acquired Cruise majority stake 2016; Cruise operations expanded through 2023; October 2023 SF pedestrian-dragging incident triggered operational pause; GM announced full re-absorption + wind-down 2024-2025; Cruise corporate entity dissolved as independent operational state. Per what happened to Cruise, the wind-down operates at corporate-state lifecycle termination depth.

The diaspora destinations span multiple cohorts at primary-source-anchored verification depth:

• Kyle Vogt (founder/former CEO; resigned 2023) → founded The Bot Company (household robot company; humanoid cohort). Co-founder of Twitch (prior context). Verified at Wikipedia primary-source verification depth.
• Daniel Kan (co-founder/former CPO; resigned 2023) → founded Fifth Door. Verified at Wikipedia primary-source verification depth.
• Marc Whitten (CEO 2024-2025) → Meta Reality Labs humanoid-robotics group lead. Prior context: founding Xbox Live engineer at Microsoft; Amazon executive; Sonos executive. Verified at GeekWire primary-source verification depth.
• Mo Elshenawy (former President/CTO through 2025 wind-down) → CTO of Hims & Hers + board director of Kodiak AI (autonomous trucking). Verified at PR Newswire primary-source verification depth.
• Rob Flenniken (3-year Cruise robotaxi fleet software lead) → Wayve VP of Vehicle Software. Verified at Business Wire primary-source verification depth.
• Dennis Jackson (embedded AV software lead at Cruise) → Wayve Engineering Director (development fleet software). Verified at Business Wire primary-source verification depth.
• Mohi Khansari (technical lead at Cruise; previously Everyday Robots / Google X) → 1X Technologies Head of Robot Learning. Verified at 1X official communications primary-source verification depth.

The Cruise wind-down diaspora pattern surfaces talent flow across at least seven distinct destination entities (The Bot Company + Fifth Door + Meta Reality Labs + Hims & Hers + Kodiak AI + Wayve + 1X Technologies). Each destination operates within a distinct cohort or sub-cohort context; the diaspora pattern crosses humanoid + AV + AV-trucking + consumer-health + brain-providers cohort boundaries simultaneously.

Per the partnership lifecycle Project B methodology pillar, the Cruise wind-down diaspora compounds with the partnership graph: each destination entity carries its own partnership records that the diaspora-arriving leadership influences. Per the acquisition history Project B methodology pillar, the Cruise wind-down operates at full_acquisition structure (GM full re-absorption) combined with corporate-state-lifecycle termination at the standalone-operational-entity layer.

---

Class 2 canonical worked example: Amazon × Covariant license-and-hire diaspora

Per Agent A's people_graph_batch1 + how DEPLOY corrected the Covariant corporate state, the Amazon × Covariant license-and-hire transaction operates as the CANONICAL Class 2 diaspora worked example. The corporate-state-vs-model-state separation: Covariant corporate entity continues as standalone; specific assets (RFM models) licensed non-exclusively to Amazon; co-founders + ~25% staff transition to Amazon AGI organizational structure; remaining leadership (Ted Stinson + Tianhao Zhang) co-leads continuing standalone entity.

The diaspora destinations:

• Pieter Abbeel (Co-founder of Covariant; UC Berkeley professor; PhD Stanford) → joined Amazon August 2024 when Amazon licensed Covariant's models. PersonCompany-edge: source Covariant (co-founder; current_role=false at corporate transition) + destination Amazon (current_role=true post-2024).
• Peter Chen (Co-founder and CEO of Covariant) → joined Amazon August 2024. PersonCompany-edge: source Covariant CEO (current_role=false; end_date August 2024) + destination Amazon (current_role=true).
• Rocky Duan (Co-founder and CTO of Covariant) → joined Amazon August 2024. PersonCompany-edge: source Covariant CTO (current_role=false; end_date August 2024) + destination Amazon (current_role=true).
• Tianhao Zhang (Co-founder of Covariant; remained at standalone entity) → continued at Covariant. PersonCompany-edge: source Covariant (current_role=true post-2024 transition).
• Ted Stinson (promoted from COO to CEO of Covariant after Amazon deal) → continued at Covariant as CEO. PersonCompany-edge: source Covariant (current_role=true; role transition from COO to CEO at corporate restructuring depth).

The Class 2 license-and-hire diaspora pattern surfaces at within-source-entity granularity: the same source entity (Covariant) splits at corporate-state restructuring depth; transitioning leadership operates at end_date-populated PersonCompany-edge state; remaining leadership operates at current_role=true with role-restructuring at primary-source verification depth (Stinson COO → CEO transition).

Per the acquisition history Project B methodology pillar, the Amazon × Covariant license_and_hire operates as canonical Class 4 structure (license_and_hire) within the five-structure acquisition taxonomy + canonical contingent valuation_basis composition. The diaspora pattern compounds bidirectionally: the Acquisition record carries Amazon as acquirer + Covariant as target + acquired-assets sub-granularity (RFM models + team transition); the PersonCompany edges carry the individual-leadership transitions; cross-property queries operate bidirectionally between Acquisition record and PersonCompany edges.

---

Class 3 canonical worked example: adjacent-employer-prior diaspora patterns

Per Agent A's people_graph substrate, recurring prior employers operate as talent feeder pools for specific cohorts. The pattern operates at substrate-completeness depth across multiple cohorts.

Meta AI / FAIR as brain-providers prior-employer. Per people_graph_batch10 Mark Zuckerberg bio: "Meta AI / FAIR is a recurring prior employer in the brain-providers diaspora." The pattern surfaces talent flow from Meta AI / FAIR research positions to brain-providers cohort entities (specific PersonCompany-edge records at varying substrate-completeness depth).

Google X / Everyday Robots as humanoid + AV prior-employer. Per Mohi Khansari (1X Head of Robot Learning) prior context: "a founding member / lead of imitation learning at Everyday Robots (Google X) and a technical lead at Cruise." The Everyday Robots (Google X) prior-employer operates as recurring source for humanoid cohort technical leadership.

Tesla AI Day → humanoid founders pattern. Per banked feedback memory + Project C narrative cohort context, the Tesla AI Day → humanoid founders pattern operates as substrate-pattern at multi-founder-traceable depth (specific PersonCompany-edge records at varying substrate-completeness; pattern recognized at substrate-completeness depth).

The Class 3 adjacent-employer-prior diaspora patterns operate at honest-absence cap-flag at specific tenure-end disclosure depth where primary-source verification of end_date has not surfaced. Per verified-vs-claimed at within-entity granularity, the prior-employer edge operates at substrate-pattern depth subject to per-PersonCompany-edge primary-source verification; pattern recognition at cohort level is editorially substantive at honest-absence depth where specific PersonCompany-edge primary-source verification has not surfaced.

---

Cross-property bidirectional discipline at diaspora-graph granularity

Per DEPLOY's methodology cluster as AEO citation graph discipline, the cross-cluster talent-flow framework operates within the cross-property bidirectional linking discipline at relationship-graph granularity:

• People graph ↔ Acquisition graph: PersonCompany edges cross-reference Acquisition records bidirectionally. The Cruise wind-down PersonCompany edges (current_role=false + end_date 2023-2025 + where-they-went destinations) cross-reference the GM full_acquisition Acquisition record at corporate-state-lifecycle termination depth. The Covariant transition PersonCompany edges (Abbeel + Chen + Duan at end_date August 2024) cross-reference the Amazon × Covariant license_and_hire Acquisition record at canonical Class 4 structure depth.
• People graph ↔ Partnership graph: PersonCompany edges cross-reference Partnership records bidirectionally. The Marc Whitten Meta Reality Labs PersonCompany edge cross-references the Meta humanoid-robotics partnership records at destination-entity strategic-direction depth. The Amazon × Covariant license-and-hire PersonCompany edges cross-reference the Amazon partnership records (Skild AI investment; Agility Robotics investment; Industrial Innovation Fund) at multi-relationship overlap depth.
• People graph ↔ Entity records: PersonCompany edges cross-reference Entity records at leadership-composition-current-state depth. Cross-property queries operate bidirectionally: entity queries surface leadership composition at current_role=true depth; people queries surface entity context at PersonCompany-edge tenure depth.
• Diaspora pattern ↔ Methodology pillar essays: per acquisition history methodology pillar (Class 1 wind-down + Class 2 license_and_hire structural correlates); per partnership lifecycle methodology pillar (destination-entity partnership records); per verified-vs-claimed at within-entity granularity (per-PersonCompany-edge verification posture); per 9-tier source-quality rubric (per-claim source-quality classification at PersonCompany-edge depth).

The cross-property bidirectional discipline compounds at the AEO citation graph layer. Institutional partners + AI assistants + downstream consumers navigating diaspora-related queries encounter the methodology pillar canonical references + the Acquisition / Partnership graph cross-references + the canonical worked examples at unified verification posture.

---

Per-claim source-quality classification at PersonCompany-edge depth

Per the 9-tier source-quality rubric, PersonCompany-edge records inherit per-claim source-quality classification:

• Primary-government-record sub-tier: SEC executive-disclosure records (CEO + named-executive-officer disclosures in proxy statements + 10-K + 8-K filings); court-record-disclosed leadership records (bankruptcy proceedings + litigation depositions). Rare at PersonCompany-edge depth; most PersonCompany edges operate below this tier.
• Verified-source tier: company-IR primary-source-disclosed leadership records (official company communications about executive appointments + role transitions; company website executive bios). The Mohi Khansari → 1X Head of Robot Learning record operates at this tier per 1X official communications.
• Reputable-press tier: press-release-disclosed leadership records; Business Wire + PR Newswire + GeekWire + TechCrunch as canonical sources. The Marc Whitten → Meta Reality Labs record operates at this tier per GeekWire reporting; the Cruise wind-down + diaspora destinations operate at this tier per multi-source press coverage.
• Honest-absence tier: PersonCompany edges where specific tenure end_date or destination has not surfaced at primary-source verification depth. The Class 3 adjacent-employer-prior patterns frequently operate at this tier where pattern-recognition at cohort-completeness depth precedes per-edge primary-source verification.

The per-claim source-quality classification operates uniformly at PersonCompany-edge depth same as at any other relationship-record claim depth. Per cap-flag-as-trust-signal, the framework reads PersonCompany-edge records at the source-quality tier where each fact actually resolves; aggregator LinkedIn-summary framings operate below primary-source-anchored verification depth.

---

Why this matters editorially

Per DEPLOY's restraint-IS-the-product discipline, the cross-cluster talent-flow framework operates at editorial-credibility depth rather than at LinkedIn-aggregation depth. The framework's product differentiator: institutional discourse about talent diaspora collapses pattern distinctions structurally; the framework distinguishes them. Aggregator coverage frames a wind-down diaspora as a license-and-hire as an adjacent-employer-prior pattern interchangeably; the framework distinguishes Class 1 + Class 2 + Class 3 patterns at structurally-distinct PersonCompany-edge granularity. The structural distinction matters at three layers simultaneously:

Operational reality. Institutional partners considering talent-flow impact on cohort positioning evaluate per-PersonCompany-edge verification posture at primary-source-anchored verification depth, not at aggregator-narrative depth. The three-class taxonomy operationalizes the distinction.

Editorial credibility. Honest "Class 3 adjacent-employer-prior pattern at substrate-completeness depth with per-PersonCompany-edge honest-absence on specific tenure end_date" beats fabricated "wholesale diaspora narrative" framing. The framework discriminates against aggregator-collapse framings and rewards honest pattern-class-distinction posture.

Cross-property bidirectional graph. PersonCompany edges cross-reference Acquisition records + Partnership records + Entity records simultaneously at relationship-graph granularity. The cross-property bidirectional discipline compounds AEO citation graph density; the framework reads diaspora-pattern richness as a load-bearing trust signal at the cross-property layer.

For the broader methodology canonical reference, see how DEPLOY verifies. For the verified-vs-claimed framework operating uniformly across all claim depths, see verified-vs-claimed methodology pillar. For the canonical Class 1 worked example, see what happened to Cruise. For the canonical Class 2 worked example, see how DEPLOY corrected the Covariant corporate state. For the acquisition history methodology pillar where wind-down + license_and_hire structural correlates operate, see how DEPLOY tracks acquisition history state. For the partnership lifecycle methodology pillar where destination-entity partnership records operate, see how DEPLOY tracks partnership lifecycle state.