Confluence.VC

There are a handful of interesting ideas (knowledge graph of the VC ecosystem, algorithmic investor scoring, using talent-community mapping to find founder mafias) but they are all stated at a high level and never unpacked with enough depth to be actionable. The back half of the episode is personal anecdote and quickfire lifestyle questions that generate no B2B-relevant insight.

“it could be using models to predict when people maybe leaving their company to start something new, right? What are maybe specific attributes there? Or it could be buying what are the best talent communities”

“we inject a lot of data, venture capital transaction data, social data, public market data, other forms of data. And we use that to reconstruct the networks that managers are embedded within”

The data-driven fund-of-funds framing is the one potentially fresh angle, but it is described at a pitch-deck level rather than with any first-principles reasoning. The most philosophically interesting point - the paradox of skill and the role of luck - is explicitly credited to others and recycled wholesale.

“I don't think that's an original idea of mine by any means. It's something I think about and something I think is underappreciated”

“when everyone, like in investing, becomes so skilled, it shrinks the standard deviation of skill, and there is the band. and so that actually again magnifies the importance that luck plays”

Jake Kupperman has real operational and early-stage venture experience, and the Levelset-to-Procore exit gives him a credible practitioner anchor. However, he is a Principal at a newly-launched $100M fund of funds, making him a relatively junior voice on topics like LP value creation and predictive scoring that would benefit from a more senior, longer-tenured perspective.

“I joined a firm here in New York, where I'm based now, called Rebel Partner, Classic Seed Stage, Enterprise Tech, an investing firm, got my feet wet on the investing side”

“was one of the first employees at a company called Levelset, which is in the construction tech space. Amazing company. Actually exited a Procore for about $500 million”

The episode provides a handful of concrete numbers (a $100M fund, 17 managers backed, the ~$500M Procore exit, team headcount of 2 data scientists and 4 data engineers) but the core claims - that the scoring model is predictive, that data applications measurably help managers - are entirely unsupported by data, examples, or timelines.

“That's a $100 million vehicle. We've partnered with, I would say, 17 managers to date with a couple more sort of an active legal and closing”

“our back test show, historically speaking, that those scores are actually quite predictive of future performance. Obviously, we'll see how that plays out moving forward”

The host asks structurally sensible questions but never pushes back on a single claim, accepts every assertion with affirmations like 'yeah, totally agree' and 'love it,' and lets the guest deliver uninterrupted pitch-mode monologues. The quickfire section abandons any pretense of B2B relevance entirely.

“Yeah, totally agree. It's interesting. To follow up to that”

“Yeah, I love it. I love it. And I appreciate that.”