a16z Speedrun

a16z now rejects small ideas, where it once rejected overly ambitious ones. Anish Acharya says that three years ago overly ambitious companies were passed on and a $100 million seed made no sense. Now ideas that are too small are what a16z avoids.

His reasoning is that AI removes the constraints that forced founders into narrow wedges. He cited Atoms as attempting something that would have felt insurmountable 5 to 10 years ago. The stance echoes Andreessen's first fund pitch: it would go to the moon or leave a moon-size crater. "An idea that's too small is not something that we want to engage with."

Lenny's Podcast▶ 1:05:25

Plan the company as if models were near-infinitely smart and cheap. Acharya tells founders to ask how the company would be reorganized if models were infinitely intelligent and astonishingly cheap. In the near term, winners are the firms that adopt the technology most ambitiously and rethink how they are organized.

He expects many industries to keep their current competitive balance because they are not bound by intelligence. His example is Pizza Hut, Domino's, Papa John's and Round Table: all would get a data center of PhDs, and he doubts any one ends up with 99% of the market. He uses electricity as the comparison: it took 40 years to move from swapping out coal to redesigning factories.

Lenny's Podcast

Run business functions as agent loops that ship and measure work automatically. Acharya defines an agent as a model in a loop with tools, memory and skill files. A growth team's loop generates and measures every variant, ships the stat-sig winner and starts the next experiment.

In a coding bug loop, a report comes in, a repro and fix are made and the fix is reviewed. Low-risk fixes ship, high-risk ones wait for a human, and the loop can email the customer within about 5 minutes. When Kavak's agents get stuck they call a human, then learn from the trace so they need not call again. Kavak's six-week "Jedi Academy" ends with every employee, mechanics included, shipping an agent in production.

Lenny's Podcast

Humans stay essential because loops stall at a local maximum. Agents hill-climb to a plateau and cannot choose a new direction, so people must pick the next hill. Work that is easy to verify turns into loops, and hard-to-verify work like sales, support, strategy and exceptions stays human.

Acharya suggests asking what you know that the model does not. He reports a Google executive who laid no one off and now gets through two years of roadmap in three months; the hardest problem became knowing what to add to the roadmap. Lenny adds that OpenAI's go-to-market team uses Codex more than its engineers do.

Lenny's Podcast

Moats come from shipping the many small ideas competitors cannot see. Cursor was criticized for having no moat, then captured reasoning traces and trained its own models. Acharya says big ideas rest on a dozen invisible small ideas that copycats never see.

He sees the biggest consumer opening in loops that make people happier, since people spend time more than they save it and entertainment and social are the largest products. He frames the gap as product design, not model capability. His stance rests on anecdotes and a few indicators, such as record job postings for radiologists and programmers, rather than data on who gains or loses.

Lenny's Podcast

a16z backs big early bets quickly after a public launch. It led Wabi's $20 million pre-seed round a month after Wabi's public beta launched in October 2025. Separately, Dmitri started a new company through Speedrun after selling YourMove AI, aiming to grow 10x faster with more resources.

Subversive

Pitch where the market is going, backed by revenue growth rather than usage. Sunflower, which lists a16z Speedrun among its investors, failed with a first deck built on the founder's background and usage numbers. The deck that raised after YC described the market's direction, positioned Sunflower in AI, showed revenue rising and included letters of intent from pilots.

Sunflower has raised about $6.5 million in total, with Flybridge leading the seed. "You can't raise funds based off of a playbook that came out of 2025."

Superwall▶ 42:20

Show a distribution channel before pitching; retention alone did not win a16z Speedrun. Flame's founder had strong retention, but a call with a16z Speedrun went nowhere and tier-one investors kept asking how the product was distributed. With no real answer there was no second meeting or term sheet, so the team monetized to survive.

"They knew I was bluffing."

Superwall▶ 19:20

a16z backs big early bets quickly after a public launch. It led Wabi's $20 million pre-seed round a month after Wabi's public beta launched in October 2025. Separately, Dmitri started a new company through Speedrun after selling YourMove AI, aiming to grow 10x faster with more resources.

Superwall