What gets more valuable after each release?
The five apps need to form a loop: creators make work, agents help operate it, receipts prove state, and the next release starts from stronger ground.
a16z speedrun / upcoming launches
Five release commitments for SR007.
This is the Speedrun answer in release-name form, backed by the build mass already in the repo: what compounds, where the creative wedge lives, what can be shown live, and how the team-shape risk becomes explicit instead of hidden. No screenshots until the proof is strong enough to inspect.
before SR007
Nine months of build receipts before five release bets.
The launch list is not starting from a blank page. The current mono stats window shows 2,976 non-merge commits, +3,970 Swift files, +250 Swift packages, +536 DocC bundles, and +6,938 Markdown files. A separate cold source sweep, excluding maintainer repos and generated/vendor/build output, counts 107,575 source files and 1,368,014 source lines.
- non-merge commits
- 2,976
- Swift files added
- +3,970
- Swift packages added
- +250
- cold source lines
- 1.37M
Evaluate the proof path.
The ask is not "believe the platform memo." The ask is whether five named apps can become partner-inspectable releases during the Speedrun window, and whether each release retires a specific investor question instead of adding another abstract claim.
Where is the games-adjacent system?
save-point and desktop-studio make creative state inspectable. The wedge is not "another devtool"; it is work you can pause, resume, remix, and ship.
What can be shown in the first few minutes?
vapor-wares has to prove the factory recursively: this app tracks what is vapor, what is real, and what evidence changed the state.
What risk stays visible?
Agent-assisted delivery is leverage, not a substitute for people. The page should make the human collaboration gap clear enough to work on.
C16 / to-build / recursion demo
vapor-wares
The first live proof move: a release-state app that tracks vapor versus shipped work, starting with itself and hello-world, so the factory claim becomes inspectable instead of rhetorical.
C17 / in-progress / creator canvas
desktop-studio
A creator canvas for desktop workflows. It turns prior desktop-utilities work into a visible studio surface where creative state can be arranged, resumed, and shipped.
C18 / to-build / games-adjacent
save-point
A checkpoint tool for creative state. This is the games-adjacent release: save the world, resume the world, prove what changed.
C19 / in-progress / creator-facing
clia
The creator-facing graduation of CLIA's face and sprite pipeline into a distributable menu-bar app, proving that agent work can cross into a polished human-facing surface.
C20 / in-progress / local AI infra
inference-control
The local AI control layer. It keeps model choice, credentials, and inference routing explicit enough that agent-assisted delivery can be operated instead of hand-waved.
The cohort outcome is falsifiable.
If these five apps become partner-inspectable releases, the factory claim gets stronger. If they do not, the claim weakens. Speedrun is the right environment because the open gaps are specific: live proof, creative wedge, human collaboration, design partners, GTM, and narrative compression.