Weekend AI Project to Company: a16z Speedrun Checklist
A five-question checklist helps solo builders decide whether a weekend AI project has shipped, earns, and can survive a 12-week in-person push.
The deadline is the first test
You have a weekend AI project that works, and now you must decide whether to make it a company or keep it a side project. a16z Speedrun's SR008 residency applications are open through November 1, 2026, and the program can put up to $1 million into each startup, starting with a $500,000 SAFE for 10% and adding another $500,000 if the startup raises again within 18 months.
The SR008 cohort runs from February 1 to April 16, 2027, in San Francisco, and the priority application window is October 12 through November 1, 2026.
If you would not apply by that date, the project is not ready for a company. If you would, use the deadline to stop polishing and prove the product to a stranger.
The checklist is the gate
Run the project through the five questions below. The first three separate a demo from a product; the last two separate a product from a company.
- Shipped v1. Can a stranger use the core workflow without you explaining the roadmap?
- Repeatable weekend scope. Can you ship a useful change in a weekend without breaking the main flow?
- Real usage or revenue. Do people open it, pay for it, or ask for it again?
- Capacity for a 12-week in-person push. Say you need that time. Can you protect it?
- Willingness to accept priced terms. Can you live with a price, a future round, and investor rights?
A stranger needs a path in, a path out, and a reason to return, which makes shipped v1 the hardest to fake. Repeatable weekend scope protects your time: if every change requires a rewrite, the project will eat the time it needs to survive. Real usage or revenue separates attention from demand; a screenshot is weak evidence, and a repeat request is a signal.
Capacity is about your body and bank account, because a concentrated push changes your schedule, sleep, and cash flow. Priced terms are what most builders skip until the term sheet arrives; they are the cost of turning a side project into a company, so test your comfort with them before you test the market.
The terms are the price of the test
a16z Speedrun seeks pro rata rights in later rounds but does not take a board seat. Solo founders can apply, but a16z usually favors multi-founder teams. If you are solo, treat that preference as a warning.
The a16z Speedrun SR007 directory listed 79 companies as of October 8, 2026. That number matters because it tells you the field is large, but it also tells you the bar is not out of reach. A solo builder can still pass the test if the product is narrow, the workflow is repeatable, and the founder can show usage without a co-founder.
The bar is evidence you can point to, plus a clear path to the next step.
AI leverage changes the scope test
The length of tasks AI systems can reliably finish alone has been doubling about every four months, after an earlier pattern of doubling about every seven months.
By May 2026, Claude had written over 80% of the code merged into Anthropic's codebase. In Q2 2026, a typical Anthropic engineer merged 8 times as much code per day as in 2024.
Those numbers change what a solo builder can ship in a weekend and what an investor should expect from a v1 and a clear next step.
The stats do not promise that your weekend project will scale; use them to test whether your scope is still too big for the time you have.
If the answers are yes, incorporate and apply. If they are no, keep building on weekends and revisit the checklist later; that is fine.