Peter
Day

The Coder Who Became a GP

Peter Day arrived at super{set} carrying something most venture investors don't: he can write the code himself. That's not a parlor trick. It's the whole point.

At super{set}, the San Francisco startup studio and venture fund that co-builds AI-native companies from inception, Peter serves as General Partner. He doesn't sit at a boardroom table and offer opinions. He joins founding teams in the earliest, messiest phase - when the product is still a whiteboard argument and the go-to-market is a guess - and builds alongside them.

The model at super{set} is deliberate and unusual: they don't back companies that already exist. They identify a market problem, find a technical co-founder, and build the company together. It's what they call "the OG of AI-native ventures." Peter arrived for Fund II - the $90 million raise in March 2024 - and brought with him a career arc that looks less like a straight line and more like someone who kept following the hardest problems.

His PhD from the University of Liverpool is in machine learning. He finished it long before "AI" became the word everyone appended to everything. Then, rather than staying in academia or joining a tech startup, he turned toward financial markets - specifically, quantitative finance. For twelve years he worked in increasingly complex corners of derivatives pricing: low-latency systems, cross-valuation adjustments, equity derivatives risk and pricing technology. By the time he left, he was Executive Director at UBS Investment Bank, overseeing equity derivatives technology globally.

Then he made the turn that matters. In 2016 he joined Quantcast - the audience measurement and programmatic advertising company - as Director of Engineering. Within two years he was VP of Engineering. A year after that, Chief Technology Officer. By the time he left in early 2025, he was Chief Product and Technology Officer, having guided the company through the launch of its flagship AI platform, featuring the proprietary machine learning engine Ara™.

The Quantcast years weren't just title accumulation. They were the laboratory where Peter learned what it actually takes to build real-time machine learning systems at global scale - systems that had to process millions of data points, update audience models on the fly, and deliver targeting intelligence to advertisers and publishers around the world without latency anyone could perceive. That's a different engineering problem than most people have ever had to solve.

It also taught him something uncomfortable: the technical side, even at that scale, was the easier part. "Getting your sales motion right is harder than building complicated software," he has said. It's the kind of sentence that sounds simple until you've burned your way through a product launch that missed revenue targets despite a technically excellent solution.

That experience became the organizing principle of his investment philosophy. At super{set}, Peter prioritizes product adoption metrics above almost everything else. Not pitch decks. Not projected revenue models. Actual usage, from real customers, showing real behavior. He is skeptical of founders who spend eighteen months building and zero months validating. He's seen too many teams build brilliantly and sell badly.

When evaluating founders, he's looking for a specific and slightly paradoxical combination: people who hold deep conviction that they are changing something important, while simultaneously staying genuinely paranoid about what could go wrong. He wants tenacity. He wants the ability to context-switch - from product to customer to hiring to fundraising to architecture - without losing the thread. And he wants founders who understand that their first instinct about go-to-market is probably wrong.

On AI specifically, Peter is more skeptical than the breathless coverage suggests. He has said he believes most current AI-as-chatbot integrations will be viewed, in retrospect, as largely ineffective. The interfaces are novel; the utility is often thin. Building software has become easier - almost anyone with GitHub Copilot and a weekend can ship something. But validating AI outputs, in production, with real users, in regulated industries, with actual stakes attached? That's gotten exponentially harder. The moat isn't in building the model. It's in knowing what it got wrong.

super{set}'s portfolio already includes exits that bear this thesis out. Habu, one of the studio's earlier bets on privacy-compliant data collaboration, sold to LiveRamp in early 2025 for $200 million. It wasn't a chatbot. It was infrastructure - unglamorous, essential, and difficult to replace.

That's the super{set} pattern Peter is building into. Not consumer plays. Not productivity wrappers. Enterprise AI with proprietary data moats, built for regulated industries: healthcare, finance, enterprise operations. The kind of companies that require both deep technical credibility and serious go-to-market muscle. Which is, incidentally, exactly the combination Peter spent twenty years accumulating before he started writing checks.

Outside the work, he's a family man - and one of those rare investors who blocks full calendar days with zero meetings, protecting the kind of focused time that serious technical work requires. He listens to "From Zero to IPO" for company-building perspective. He prefers physical conference stages to webinars. And his advice to engineers who've lost the plot: "Get outside, take a break, and don't forget to talk to your customers."

He's now appeared as a speaker at Agent Conference and written for VentureBeat and Ad Age - adding a public intellectual dimension to a career that has mostly played out in engineering rooms and product reviews. The thinking is sharp, the sentences are short, and the opinions are backed by years of having been wrong in expensive ways.

The phrase super{set} uses for what Peter does is "co-building." It's the right word. He's not a financial engineer who learned to like technology. He's a technologist who spent time in finance and came back to build the things he knows how to build - this time with a fund behind him and a studio full of founders who need exactly what he has.