a16z Speedrun SR007 · May 2026

Level Up Factory

AI-native app factory: research ideas, build with agents, automate ads, test unit economics, then scale, improve, or stop.

934 ideas researched ~100 AI execution contexts 500K Level Up Basketball users 6+ products shipped since February 2026

The Thesis

Building software is no longer the hard part.

The hard part is choosing what to build and proving distribution. Level Up Factory researches markets, ranks ideas, builds with agents, generates and uploads ads, tests demand, and feeds results back into the next bet.

Selection: rank ideas by pain, buyer, market, speed to MVP, channel viability, CAC, LTV/CAC, and testability.

Build: move selected ideas into reusable AI execution across workspace agents, Paperclip agents, VM assistants, and app-embedded AI.

Market feedback: generate video, image, and carousel ads, publish campaigns, analyze performance, and improve the next product decision.

Proof Today

Level Up Basketball is the flagship proof. DoItFor.Life is the current wedge.

Level Up Basketball

Consumer distribution proof with 500K users and AI Coach-driven product improvement.

DoItFor.Life

Managed OpenClaw runtime: dedicated Google Cloud VM, OpenClaw, Gemini, and messaging ready in about three minutes.

DeckReview.app

Productized Selection Engine for startup and deck research, validated by S16 as useful.

DoneCo.app

Internal AI project management and human-agent coordination layer for running the factory.

Why Speedrun

The system is real, but still early enough to shape.

I want feedback from AI-native founders and operators on the Selection Engine, the first clear product wedge, and repeatable distribution. Speedrun is also a chance to connect with other AI-native founders and small-team operators who are solving similar problems around agents, product velocity, quality, and GTM.

Read the full memo See products See Selection Engine