Interview: Kyle Bliss Dustin — 2026-04-28
Key Themes
Data desert at the core of the problem. Kyle’s DoD-funded Cornell project independently arrived at the same diagnosis: Bloomberg/Factset supply chain data is structurally inadequate (~300k firms, avg 2.5 suppliers/customer), forcing a pivot to ML-based prediction from 10-K filings and press releases. This independently validates our core problem statement.
Legal teams are the primary blocker, not operators. COOs and Chief Security Officers at major semis firms (e.g., Micron’s CSO) are enthusiastic about the problem and acutely aware of their tier 1-2 visibility gap. Legal teams — citing 7,000+ bespoke supplier NDAs — consistently killed data-sharing conversations. This is the structural friction that killed Kyle’s data acquisition efforts.
Go-to-market divergence: DoD-first vs. industry-first. Kyle’s team went government-first due to funding origins. We are going industry-first, using compliance requirements as a wedge to get legal teams to say yes. Kyle validated that government demand is real but noted that compliance currently only requires tier 2 visibility for DoD contracts — meaning the compliance wedge may be more potent on the commercial side.
The $150k Deloitte audit is the incumbent benchmark. Current DoD process: hire Deloitte for $150k, 3-month manual audit, no reuse across engagements. Kyle sees this as an obvious compression opportunity (<24 hours with a digital solution). This is a concrete incumbent price point and workflow to displace.
Competitive landscape is real but not dominant. Sayari, Govini, and Interos ($1B DoD contract) are all in the space. None appear to have cracked the data acquisition problem. The missing piece Kyle identifies is a ‘golden dataset’ — validated ground truth supply chain relationships to generate precision/recall curves.
Notable Quotes
- “Nobody could really tell you beyond tier one. Maybe a couple people on tier two, like, who was in their supply chain.”
- “Shit data in is like shit product out.”
- “Legal teams consistently blocked data sharing — 7,000+ bespoke NDAs with suppliers created insurmountable barriers.”
- “[DoD] hire Deloitte for $150k, 3-month manual audits. No reuse of previous audits, even for same companies.”
Surprises
- Scale of the Cornell effort: $3M DoD pilot, team of up to 25, 200-page deliverable — and still didn’t crack data acquisition. Sobering signal on how hard this is.
- Funding freeze mid-project due to Ivy League grant politics (not just DOGE-era cuts) — a real operational risk for government-funded approaches.
- Kyle explicitly frames our industry-first approach as superior to his DoD-first path in hindsight, particularly the compliance wedge as a legal-team unlock.
- Cornell stopped entity resolution at firm level (TSMC as one node, not TSMC Arizona vs. Taiwan) — a known limitation that creates a real product differentiation opportunity.
Open Questions
- Can Cornell’s 200-page report and 20-slide summary be shared? This could save significant research cycles.
- What does the Georgetown semiconductor industry overview paper contain — and does it overlap with our existing desk research?
- Which smaller private semiconductor companies might be willing to exchange supply chain data for free analysis — and does Kyle have warm intros?
- How does Interos’s $1B DoD contract affect the government sales motion — are they entrenched or is there room?
- What is the current status of the DoD pilot program renewal in the FY26 NDAA cycle?