a16z Speedrun One-Page Investment Memo · May 2026

Level Up Factory

AI-native app factory that researches markets, builds with agents, automates ads and feedback, then uses real unit-economics data to decide what to scale, improve, or stop.

Executive Summary Problem Product Wedge Proof Fundraise Make Or Break Links
I. Executive Summary

The bottleneck moved up the stack.

AI agents made software much cheaper to build. The harder questions are now what to build, how to reach users, and whether the unit economics work.

Level Up Factory is a learning loop for that bottleneck. Every shipped or stopped product improves research, build infrastructure, creative generation, GTM automation, and founder judgment.

II. Customer Problem

Technical founders can ship prototypes fast. Many still overbuild before proving distribution.

Market choice

Which pain is urgent enough, reachable enough, and valuable enough?

Distribution

Can the product reach users through paid, organic, or relationship channels?

Timing

When should a team keep improving, scale spend, or stop the product?

III. Product

The factory combines selection, build, automated GTM, and operations.

Selection Engine

Ranks ideas by pain, buyer, market, build days, blockers, channel fit, likely CAC, and LTV/CAC.

Agent Build Harness

Moves selected ideas into reusable AI execution across workspace agents, VM assistants, app AI, and Paperclip tasks.

Marketing Loop

Gemini generates ads. Meta and Google campaigns are published, analyzed, and improved through the automated feedback loop.

DoneCo.app

Internal alpha AI project management dashboard for human-agent coordination, reporting, blockers, metrics, and decisions.

IV. Business Thesis And Wedge

Subscription products that pass selection and marketing tests.

The first user wedge is DoItFor.Life: a managed OpenClaw runtime for personal AI agents. It creates a preconfigured OpenClaw environment on a dedicated Google Cloud VM, with Gemini and messaging ready in about three minutes.

The operating wedge is DoneCo.app: the coordination layer for AI-native small teams running humans and agents together.

V. Current Proof

Real products, real loops, and reusable infrastructure.

LevelUpBasket.com

Flagship consumer proof with 500K users and AI Coach growth.

DoItFor.Life

Full loop tested: research, build, onboarding, automated ad generation/upload, paid traffic, usage analysis, support learning, and iteration.

DeckReview.app

Productized Selection Engine, useful enough for S16 to validate as a deck-review workflow.

DoneCo.app

Internal alpha operating dashboard for coordinating founder work and AI agents.

BookScout

Consumer app proof for vision, enrichment, pricing, and shareable AI interaction.

Vision Studio

AI creative pipeline proof for product and ad-generation workflows.

VI. Fundraise And Use Of Funds

Raising $500K strategic pre-seed / SAFE around an $8M valuation cap.

Development cost is low because the factory builds with agents and reusable infrastructure. Most capital goes into market validation, creative testing, infrastructure, support, and scaling winners.

The 12-month plan is three carefully researched products per month. Throughput can increase as the harness improves.

VII. Make Or Break

The main risks are not whether agents can write code. The risks are selection, distribution, and focus.

Selection Quality

Make: the Selection Engine improves with every product tested.

Break: outcomes are too noisy to improve future selection.

Current view: the engine already filters weak ideas before build time; the next step is more post-launch data.

Demand Signals

Make: paid tests and product usage reveal useful demand signals quickly.

Break: CAC is too high or channels do not scale.

Current view: paid tests are not the only GTM path, but they expose demand, objections, and unit-economics risk quickly.

First Wedge

Make: DoItFor.Life or DoneCo.app becomes the first clear wedge.

Break: the factory stays too broad and investors cannot understand the entry point.

Current view: DoItFor.Life is the clearest user-facing wedge. DoneCo.app is the clearest operating layer for AI-native small teams.