← ALL EPISODES

BEYOND THE REPORT · A CONVERSATION CARRYING THE RETHINK THESIS FORWARD

Venture reimagined by infrastructure

For decades, venture capital's edge came from being in the room. If yesterday's advantage was network and knowledge arbitrage, today's may be something different: data arbitrage.

AUGUST 202549 MINGUEST · ALBERT AZOUT

NETWORKS ARE ASYMMETRIC LEVEL VC
We want to refine these tools, have more of a data flywheel, and really have the most sophisticated data set in the market.
Albert Azout · on what is next for Level
PRESS PLAY TO LOAD THE VIDEO. NOTHING IS REQUESTED FROM YOUTUBE UNTIL YOU DO.
Under $100MTarget fund size for the emerging VCs they back, pre-seed and seed, with an average around $40M
80 / 20Roughly 80% fund investing and LP or GP led secondaries today, 20% co-investing, which they expect to grow
$9MThe size of a position they will take in a $30M to $40M fund, taking a large LP commitment and a lot of conviction
70%+Share of capital going to the multi-stage firms, the smart beta side of the bifurcating market
ABOUT THE EPISODE

For decades, venture capital's edge came from being in the room: the right dinners, the right syndicates, the right backchannels. But are those edges eroding? Companies stay private longer, DPI is suppressed, and capital pools are more crowded than ever. Selection risk is up, liquidity is down, and "alpha" doesn't look like it used to.

If yesterday's advantage was network and knowledge arbitrage, today's may be something different: data arbitrage.

The idea is simple but radical. What if you could reconstruct the invisible graphs behind venture, who co-invests with whom, where talent migrates, which circles spot signal first? What if benchmarks weren't generic Cambridge tables, but dynamic peer sets tailored to each segment? What if diligence cycles compressed from weeks to days, powered by proprietary models fused with LLMs?

At that point, a fund-of-funds is no longer just a fee layer. It starts to look more like an operating system: allocator, co-investor, and analytics engine in one. A platform that doesn't just access networks, but maps them before anyone else walks in the room.

That raises a deeper question: when networks and judgment can be modeled, does gut feel still rule VC, or are we watching the first serious attempt to systematize private markets and bring them closer to hedge funds?

That's the experiment being run at Level VC, founded by Albert Azout. Their approach shows how technology, AI, and data can reshape the way venture funds and portfolio companies operate. This is venture reimagined, not by instinct alone, but by infrastructure. The episode includes a rare inside view of their system and how they are mixing tech, AI and data to build a next-generation platform.

IN HIS WORDS
CHAPTERS

Click a timestamp to jump the player to that point.

MENTIONED

FIRM & PEOPLE

FIRM

Level VC

Started in 2021. Three parts: a fund of funds backing emerging pre-seed and seed VCs, a co-investment arm focused on Series B and sometimes pre-B, and a technology team of mostly data scientists and engineers. Now investing out of its second fund, started in January.

GUEST

Albert Azout

Twice exited founder; started his career at Lehman as a software engineer and technology analyst; was a GP at Cota Capital, then a venture partner there, and started Level in 2021.

HOST

Rohit Yadav

Investments thought leader, Big Book of VC.

THE SYSTEM

METHOD

Complexity theory and knowledge graphs

Reconstructing patterns of relationships between entities rather than looking at a GP in a vacuum: co-investor networks and talent networks, on the view that networks are asymmetric and where you sit affects performance.

MODEL

Graph neural network manager score

Outputs a prediction of a manager's potential top decile performance hit rate, fed by transactional data, the founders a GP backs, and the quality and experience of that talent.

BENCHMARKS

Their own, not Cambridge or PitchBook

The industry is benchmarked and sub-segmented in house, so a manager is compared against peers inside their own segment.

TOOLING

Network search for GPs

High definition search across the market and a team's own network for specific skill sets, expertise or experience, used for diligence, sourcing and portfolio support; also CRM enrichment.

PLATFORM

Remote MCP service

One of the first: a model context protocol letting foundation models interact with external services, with sub-resources the team uses to get up to speed inside the context of their own data.

THE ARGUMENT

FRAMEWORK

Smart beta and smart alpha

Multi-stage firms deploying large amounts into highly legible, momentum-driven opportunities on one side; on the other, GPs in overlooked and less legible areas, or with the structural ability to get into deals.

FRAMEWORK

Three sources of alpha

Network advantages that are non-redundant and economically viable, knowledge advantages that are specialised and non-redundant, and structural advantages driven by fund size and ownership.

CONCEPT

Data arbitrage

His view that as the market institutionalises, every piece of alpha gets eked out, and informational or data advantages sit alongside network and knowledge advantages.

PUBLIC MARKETS

60 to 70% passive

His figure for passive holders of equities, and his point that many of the strongest active public investors are crossover investors.

RELATED EPISODES