Context

This episode, originally from the a16z AI Policy Brief, features a16z's Matt Perault interviewing general partner and Speedrun lead Andrew Chen about what "little tech" actually looks like at the earliest possible stage of a company's life. Speedrun is a16z's program for investing up to a million dollars in brand-new startups, often when founders are still working out of a kitchen table with just one or two co-founders. Chen describes the day-to-day mechanics of these earliest-stage teams, including how heavily they now rely on AI tools instead of hiring, and Perault brings the policy angle: how regulation accumulates on companies that have no lobbyists, no policy teams, and often no time to even notice the rules affecting them until they hit one. For a PM or founder, this is a rare, concrete look at what building actually looks like before a company has metrics, a team, or even formal structure.

The Big Idea

The earliest stage of company formation, two or three people working out of a kitchen table with AI doing much of the work a larger team used to require, is systematically underrepresented in policy conversations, because these founders are too consumed with basic survival to advocate for themselves, even though regulatory friction disproportionately affects them.

Chen argues that if a city or state wants startups, it has to actively choose to make itself conducive to them, since the founders who would otherwise make that case rarely have the time, resources, or awareness to do it themselves.

Key Insights

AI coding lets tiny teams skip hiring entirely

Chen describes a common Speedrun team structure: one "business" co-founder focused on customer interviews and deals, and one "technology and product" co-founder who is now typically using AI coding tools heavily. Rather than adding headcount cost or outsourcing development, that single technical co-founder can often build and validate the product alone. The explicit goal at this stage isn't to build a scalable engineering organization, it's to determine as cheaply and quickly as possible whether the underlying business idea works at all, before spending money on hiring.

Founders decide company identity later than you'd expect

Chen notes that many well-known companies changed their core idea entirely after founding, citing Slack's origin as a browser-based video game company as the standard example. Because of this, Speedrun's selection criteria focus almost entirely on the people rather than the specific pitch: unusual life experience, a fast-growing open-source project, or a track record from an elite AI company, since the idea itself is treated as provisional and likely to change.

Regulatory paperwork is built for companies that don't exist yet

Chen's description of how a two-person, pre-incorporation team experiences regulation: nearly all compliance requirements (data provenance rules, privacy law, state-level AI legislation) assume a company already has legal and compliance staff to interpret and act on them. For a founder who hasn't even formally incorporated yet, that paperwork isn't a manageable cost of doing business, it's often the first real obstacle standing between them and finding out if their product works at all. Chen and Perault reference an a16z policy piece, "Greens from a Garage," making the case that the real regulatory burden on a founder is the cumulative weight of every law passed over years, not any single new bill considered in isolation.

Founders can't spend even a day advocating for themselves

Chen is direct about why little tech has no voice in policy debates: founders working 100-hour weeks on a company that might not exist as a business in three months literally cannot justify the time cost of flying to Sacramento or Washington to advocate for their interests, even when a specific regulation would materially affect their company. This creates a structural imbalance where policymakers hear disproportionately from larger, better-resourced incumbents (including companies actively being disrupted by these same startups), producing a skewed picture of what "the industry" actually wants.

Startups can choose where to plant roots, and increasingly do

Because a two-person team is maximally mobile, where to locate becomes one of the most consequential early decisions a startup makes, and Chen frames this choice as increasingly regulation-aware. He traces the historical migration of startup hubs (from the Peninsula to San Francisco proper, and now the emergence of New York, London, and other hubs) and notes indirect regulatory signals matter too: cost of living, availability of lab and warehouse space for deep-tech and robotics companies (citing El Segundo in Los Angeles and various Texas hubs), and proximity to investors, since raising real scaling capital still requires being near the people who provide it, even though building an early product and finding first customers can now happen from almost anywhere.

Failure gets recycled into future founders, not wasted

Chen describes what happens to the majority of Speedrun companies that don't succeed, using the well-established venture pattern where roughly half the companies fold outright, a few return modest capital, and only about one in ten produces the outsized return. His framework for what happens next: a16z frequently backs the same founder again on a better idea (he cites re-funding a founder the week of this recording), or the founder gets hired by another portfolio company for a few years before eventually spinning out to try again. He treats this recycling of experienced, currently-informed talent as a deliberate, expected part of how the ecosystem compounds value, not a failure of the model.

Mental Models & Frameworks

The composite regulatory burden, not the incremental bill

Perault describes a reframe his policy team developed: instead of asking "how risky or helpful is this one proposed law," ask what the full, cumulative regulatory picture looks like for a founder building from scratch today, counting every existing data provenance rule, privacy law, and AI-specific requirement stacked together. Individually, each law might seem reasonable; collectively, they can represent a wall of compliance work no two-person team can realistically absorb. This reframing (composite burden over incremental burden) is the practical tool Chen and Perault use to make the founder's actual experience legible to policymakers who otherwise only evaluate one bill at a time.

Mobility as leverage in choosing where to build

Chen's implicit framework for founders: because an early-stage team has almost no fixed infrastructure, it holds real leverage over which jurisdiction gets its economic activity, future hiring, and tax base. He argues this leverage is a underused lever for cities and states that want to attract startups, and one that's easy for policymakers to overlook since the actual founders exercising that choice rarely show up in person to explain their reasoning.

Trade-offs & Nuance

Speedrun's investor voice is not a substitute for founder voice

Perault flags a specific limitation directly in the conversation: when a16z shows up to advocate on behalf of little tech, policymakers often say they'd rather hear from the actual startups, not the investor backing them. But the startups themselves cannot make that trip, given their time constraints. This leaves a real, structurally hard-to-close gap: the people most affected by a regulation are the least able to explain its effects in person, and no proxy fully closes that gap, even a well-intentioned one.

The Bay Area's advantage is not guaranteed to last

Chen is careful to frame Silicon Valley's ongoing dominance as contingent, not permanent. He notes nearly 50 percent of venture-backed founders are first-generation immigrants who chose to relocate there, implying the ecosystem draws talent by choice rather than by birthright, and that choice could shift. He specifically names California's proposed wealth tax as a risk to the investor and family-office base that keeps the ecosystem well-capitalized, arguing that if that capital relocates, some of what makes the Bay Area uniquely fertile for startups could erode.

Practical Application

Design your earliest product cycle around one or two people, not a team

If validating a new business idea, Chen's Speedrun pattern suggests deliberately resisting the urge to hire before you know whether the idea works. Use AI coding tools to let a single technical co-founder build and iterate, and dedicate the other early co-founder entirely to customer conversations and deal-making, rather than adding headcount cost before there is real signal the business is viable.

Audit your product's exposure to jurisdiction-specific rules before scaling

Before committing to a location or expanding into a new state or country, map out the composite regulatory burden you'd face there. Rather than treating each proposed law as an isolated risk, consider the full cumulative stack of existing rules (data provenance, privacy, AI-specific legislation) a company operating in that jurisdiction has already had to absorb, since that composite total often matters more than any single piece of pending legislation.

Treat a failed first attempt as a career input, not an end point

If you have worked closely with a founder or team whose company didn't work out, treat that as useful signal rather than a disqualifier. Chen's account of a16z re-funding the same founder on a better idea, or portfolio companies hiring founders from failed startups for a few years before they start again, suggests that experience compounds even through failure, and that judging a founder solely on their most recent outcome misses the broader trajectory.

Questions to Consider

• If your company or product depends on a favorable regulatory environment, have you mapped the full composite burden of existing rules you'd face, not just the next single proposed law, the way Matt Perault's team did in their "Greens from a Garage" analysis?
• Where in your own hiring plan could an AI tool let one person do what used to require a team, the way Andrew Chen describes a single technical co-founder handling product development without hiring?
• If your team or company has no realistic capacity to advocate for itself in a policy process, who is currently speaking on your behalf, and does their perspective actually match yours?
• Are you evaluating a founder or team's potential based only on their most recent outcome, or are you accounting for the full arc of experience they've built even through a prior failure, the way Andrew Chen describes a16z re-backing founders?

Bottom Line

Andrew Chen's account of Speedrun makes clear that the earliest stage of startup formation, often two people and an AI coding tool working from a kitchen table, is both where a disproportionate amount of future value gets created and where founders have the least capacity to protect themselves from regulatory friction. Anyone building or evaluating early-stage products should recognize that AI is compressing team size at this stage, and that the composite weight of accumulated regulation, not any single rule, is what actually determines where and how these smallest teams can operate.

Case Studies Mentioned

Slack's origin as a video game company

Chen cites Slack, a product now used daily by millions, as having originally started as a browser-based video game company before pivoting entirely. He uses this as the standard cautionary example against evaluating an early-stage startup on the specifics of its initial idea, since founders frequently change direction substantially after their company already exists.

People to Follow

Andrew Chen

General partner at Andreessen Horowitz and lead of the firm's Speedrun program, which invests up to a million dollars in very early-stage startups, often when founders have not yet incorporated a company. Chen also leads Tech Week, a multi-city program (San Francisco, Los Angeles, New York, and as of this year, Boston) built around convening founders, investors, and increasingly policymakers around startup ecosystems.

Notable Quotes

"This is truly little tech. The average team is two to three people. They're running their companies not in their office, not in the co-working space. They're running it at the kitchen table." (Andrew Chen)

"It's a choice whether or not each state or each city wants to have startups or not." (Andrew Chen)