"Meet Dave Blundin: Who I Am and What I Stand for" — Peter Diamandis (Moonshots)

Why this is in the vault

Blundin is a recurring Moonshots co-host and a founder-turned-VC whose "remove all friction for early-stage founders" thesis maps directly to what RDCO's agent-COO architecture is building toward; his AI governance "GPU logging" startup idea is a signal worth tracking.

Episode summary

Bonus "Meet the Mates" profile of Dave Blundin, managing director of Link Ventures — a Cambridge-based founding-stage VC fund. The episode is a walking tour of Link Studio at Kendall Square, mixing personal backstory (NJ → Iran → London → MIT) with Blundin's investment thesis, teaching philosophy, and a live demo of "The Forge," a voice-driven AI room Link built. The back half covers Blundin's history with the Moonshots podcast and closes with two sharp takes: AI is a civilization-level change, not an industrial one, and if he were starting over he'd build a company that logs every GPU's token-in/token-out for accountability.

Key arguments / segments

• [00:00–01:30] Tour of Link Studio: Presents it as "highest talent density place on the planet" — 100k annual visitors, blue-bike accessible, food/payroll/compute provided to portfolio companies.

• [00:01:30–03:30] Background and founding identity: Born NJ, childhood in Iran (father took hardship pay), then London, Connecticut high school, MIT. Self-identifies as a founder first, investor second. MCed 200 consecutive monthly all-hands at Bestark as the company scaled.

• [00:03:30–05:30] Teaching at MIT + people-over-classes thesis: Technology changes faster than any curriculum can track. Real learning comes from peers — the co-founders and lifelong friends you meet at school, not the lectures. Best unanswered student question: "Can I get someone else to do my work?" — which Blundin says is exactly the right entrepreneurial instinct.

• [00:05:30–07:00] Link Ventures core thesis: Founders are getting younger and reaching billion-dollar valuations faster. They need more infrastructure support, not less. Link provides payroll, accounting, office, housing, food, and (soon) compute so founders can move without friction.

• [00:07:00–09:00] Differentiation from A16Z/Sequoia and deal criteria: Link writes $1–10M checks at founding stage vs. $100M+ growth checks. The "three best friend rule" — ideal team is three people, all technical, all trustworthy, who deeply trust each other. Only failure mode is disband; if they love each other they rarely will.

• [00:09:00–13:30] "The Forge" AI room demo: Wall-to-wall monitors, fully voice-driven. Launched May 4th. Dave narrates a video request to the AI ("Ember") — Star Trek theme, Chris Pine era, outer space, phaser fire — and notes the bottleneck is GPU availability. Estimates instant generation within 3 months once hardware is in-house.

  

• [00:13:30–15:30] Moonshots origin story: Peter and Salim already had the show; Blundin guested, it went well, he joined. Alex Gross was added as a guest and the audience loved him instantly. Peter and Dave go back 30+ years to MIT; met Salim through XPRIZE; met Alex by bumping into him at Kendall Square.

• [00:15:30–17:30] Co-host dynamics and AI takes: Describes Moonshots co-hosts as ingredients in a chocolate chip cookie. Elon Musk most surprising guest — "just a regular guy." AI exponential self-improvement is not industrial revolution scale — it's the difference between prehistoric humans and humanity 2.0.

• [00:17:30–18:37] GPU logging startup idea + founder urgency: If starting over, Blundin would build a company that captures every GPU's token logs for AI accountability/governance — governments can argue over access, but at least the record exists. Closing message: "Time is short. Do it 10 times faster. The window is the best tailwind I've ever seen."

Notable claims

• Link writes $1–10M checks; A16Z/Sequoia write $100M+ — purely stage-differentiated, not competing
• "Three best friend rule": 3 co-founders, all technical, all trusting — disbanding is the only failure mode they observe
• Blitzy (Link portfolio): 80 people today, growing to 300 by end of 2026
• Michael Saylor (MicroStrategy) was Blundin's MIT roommate
• AI impact is "civilization-level" — prehistoric humans → current humanity → humanity 2.0 (2026)
• Instant AI video generation (current bottleneck = GPU hardware, not software) — Blundin's estimate: within 3 months with in-house hardware
• GPU logging startup: every token in / token out captured centrally for AI safety governance

Guests

Dave Blundin — Managing Director and founder of Link Ventures (founding-stage VC, Cambridge MA); co-founder of Vestigo Ventures LLC; serial entrepreneur (3x operating company founder post-MIT); teaches entrepreneurship at MIT; XPRIZE Innovation Board member; Moonshots podcast co-host. MIT AI Lab background — early neural network research on the Connection Machine supercomputer. Website: db2.ai.

Mapping against Ray Data Co

Incubator-as-infrastructure model aligns with agent-COO philosophy. Link Ventures removes friction for early-stage founders (payroll, accounting, housing, compute) so founders can focus purely on building. This is exactly the posture RDCO's agent-COO architecture targets — absorb overhead, surface only judgment calls. Worth internalizing as a framing metaphor when pitching the COO-agent value prop to clients.

Three-best-friend rule as a solo-founder flag. Blundin's hardest filter is "where's the third?" — if your vision isn't compelling enough to recruit a third technical co-founder, he won't fund you. Ray is intentionally solo + AI-augmented; this is a useful tension to hold when positioning RDCO's model against traditional VC assumptions.

GPU logging startup idea = AI governance signal. Blundin would build a company capturing every token in/out at GPU level for accountability. This is a real whitespace — no dominant player yet. Tangentially relevant if RDCO ever enters AI governance/compliance consulting; worth flagging if phData opens a governance practice.

Urgency frame matches RDCO's current posture. "Run everywhere you go — the window is the best tailwind I've ever seen" is consistent with the L4→L5 acceleration and the phData bet timing. Useful reinforcement to pull when Ray needs urgency framing in internal or client contexts.

Related

• [[2026-07-01-moonshots-sonnet-5-drops-fable-5-ep268]] — same channel, same week, Moonshots episode on Claude Sonnet 5 / Fable 5 drops
• [[backfill-discovery-moonshots-2026-04-12]] — channel backfill discovery index; Blundin appears as recurring co-host across earlier episodes
• [[2022-moonshots-ep1-elon-musk-save-world]] — Blundin cites Elon Musk as most surprising Moonshots guest ("just a regular guy")