Michael Dempsey

Who they are

Michael Dempsey is Managing Partner at Compound — one of CB Insights' first 15 employees turned deep-tech VC, with a blog at michaeldempsey.me spanning AI, autonomous vehicles, and the epistemics of being early versus wrong.

Person

Michael came up through CB Insights as one of its first 15 employees, building out the data and research function when the company was still a scrappy startup — that early immersion in tech intelligence likely seeded the pattern-recognition he now deploys as an investor. From there he moved into venture, eventually reaching Managing Partner at Compound, a firm with a distinct thesis around deep tech and frontier technologies. He writes prolifically at michaeldempsey.me — essays covering AI, autonomous vehicles, robotics, and the meta-dynamics of venture — and keeps a notes site at notes.michaeldempsey.me. His 2019 essay 'Narratives & Pseudosecrets' on being early versus being wrong is the clearest window into how he thinks: the problem isn't having a contrarian view, it's distinguishing a true secret from a narrative the market hasn't priced yet. A 2024 LinkedIn post on 'The Shifting Dynamics & Meta-Moats of AI' shows he's kept updating that framework. He's spoken at Inquisitive VC on Compound's strategy, at Infinite Frontiers on resolving mismatched time horizons, and on a 2025 podcast specifically on why deep-tech startups mess up value capture — that last topic is probably his sharpest public edge. The through-line is someone who turned early access to tech data into a durable investment perspective, and who keeps publishing the working notes.

How they likely show up

Early CB Insights employee (one of first 15, data and research builder) → likely values rigorous information infrastructure before forming a view, not gut-first investing.
Active public writer across blog, Medium, LinkedIn, and a separate notes site → comfortable thinking in public and probably expects counterparts to have read at least something before the meeting.
Speaks specifically about 'mismatched time horizons' in deep tech → likely has a strong internal clock around when a thesis is early vs. wrong, and will probe whether you share that patience.
2025 podcast episode on why deep-tech startups mess up value capture → actively stress-tests commercialization logic, not just technical feasibility.
Possibly — mixed tenure shape across roles → may move deliberately between building phases rather than optimizing for long institutional stays.

Conversation tips

→ Read 'Narratives & Pseudosecrets' before the meeting — he built a whole epistemics framework around being early vs. wrong, and referencing it specifically will signal you engaged with his actual thinking.
→ Come with a view on value capture, not just technical differentiation — his 2025 podcast topic makes clear he finds the commercialization gap more interesting than the technology gap.
→ Ask about the CB Insights years: what it was like building research infrastructure at that scale as an early employee, and how it shaped his sourcing or diligence process as a VC.
→ Don't pitch a thesis without also articulating the counter-narrative — his 'meta-moats of AI' framing shows he thinks in terms of market dynamics, not just feature sets.

Toolbox

Openers

Open on the 'Narratives & Pseudosecrets' essay — he published it in 2019 and it's still the clearest articulation of his investing epistemics (early vs. wrong vs. secret); referencing a specific line lands better than a general compliment.
Lead with his 2025 podcast on why deep-tech startups mess up value capture — it's his most recent named public moment and signals you're tracking his current thinking, not a stale LinkedIn bio.
Bring up the CB Insights origin — being one of the first 15 employees at a data-intelligence company is an unusual foundation for a deep-tech VC, and he's spoken about it in the Moth Fund profile; it's a genuine differentiator worth unpacking.

Discovery questions

In your 'Shifting Dynamics & Meta-Moats of AI' post you mapped how moats are changing — a year on, has anything shifted your view on where durable advantage actually accrues?
Your 2025 talk was specifically on why deep-tech startups mess up value capture — what's the most common failure mode you keep seeing, and is it founder error or structural?
The 'mismatched time horizons' framing from Infinite Frontiers is interesting — how do you actually manage that mismatch inside a fund structure, especially when LPs have their own clocks?

Avoid

Don't show up with a surface-level 'AI is transforming everything' pitch — he's written explicitly about the gap between tech narratives and genuine secrets, and vague enthusiasm will read as exactly the noise he filters out.

Make it yours