The era of "build fast, ship lean" is ending—VCs are now writing checks before founders even know what they're building.
The Summary
- South Park Commons is expanding its fund to write bigger checks and hold investments longer, betting on founders in the "minus one to zero" phase—before ideas crystallize
- The shift reflects AI's push toward capital-intensive ventures that need longer runways than typical startup timelines
- Early bet on Cognition CEO Scott Wu (pre-Devin) exemplifies the strategy: back the builder, not just the pitch deck
The Signal
South Park Commons isn't just raising a bigger fund. They're codifying what the AI era demands: patient capital for problems that take years to solve and millions to validate. General Partner Aditya Agarwal calls it "minus one to zero"—backing founders before they've even locked in on the specific problem worth solving.
This isn't typical pre-seed posturing. It's a structural response to how AI companies actually get built now. Training runs cost six figures. Compute contracts span quarters. Technical moats require PhD-level teams working for 18 months before anything ships. The old model—raise a small round, iterate fast, show traction in 6 months—doesn't work when your MVP needs a research paper and a GPU cluster.
"The firm's new fund will allow it to write bigger checks and invest for longer."
The Cognition bet tells the story. South Park Commons backed Scott Wu before Devin existed, before autonomous coding agents were a category. They bet on the founder's technical depth and gave him room to figure out what to build. Devin became one of the defining agent companies of 2024. That doesn't happen with a 12-month runway and pressure to show growth metrics.
What's changing isn't just check size. It's timeline. Traditional venture works on 7-10 year fund cycles with pressure to show progress every 18-24 months. AI infrastructure plays need longer. The companies building foundational agent orchestration layers or training new model architectures aren't hitting revenue milestones on a typical SaaS schedule. They're doing R&D that looks like a PhD program crossed with a startup.
South Park Commons is adapting by:
- Writing larger initial checks to fund longer pre-product phases
- Holding investments through extended development cycles
- Betting on technical pedigree over immediate market fit
This creates a new founder profile. Not the classic "iterate your way to product-market fit" entrepreneur. The AI-era founder South Park Commons wants is someone who can hold a hard technical problem in their head for years, attract world-class co-founders to an unsolved question, and build infrastructure before markets fully exist. That's a different skill set than growth hacking or customer development.
The Implication
If you're a technical founder with a genuinely hard AI problem and a multi-year vision, this environment is built for you. But you need to be comfortable with ambiguity longer than previous startup generations. The flip side: if you're building an LLM wrapper or an agent that's "ChatGPT but for X," patient capital isn't coming your way. VCs expanding fund timelines are doing it for moonshots, not marginal improvements.
Watch for more funds announcing similar expansions. When capital shifts to match the actual development timelines of transformative technology instead of forcing technology into capital timelines, you get Web4 infrastructure instead of Web2 apps with API calls.