Human Language Lab · v0
Programming should feel more human.
A human-first, agent-native language experiment: say what the software should do clearly, keep the meaning strict, and compile it into ordinary inspectable code.
The first tiny program
app GarageCRM
data Customer:
name
email
status = "lead"
when Customer is created:
send welcome email
notify sales
when payment succeeds:
Customer.status = "client"
create project
Why this exists
Today we spend enormous human and machine effort navigating syntax, boilerplate, framework conventions, and sprawling codebases. The experiment asks whether a smaller, clearer language can make software easier for people to read and cheaper for coding agents to change.
The current prototype compiles .human files into plain Python. AI can help write it, but AI is not required to run it.
What we measure next
Give agents the same change in Human, Python, and TypeScript, then compare tokens, retries, mistakes, elapsed time, and final correctness. If Human really helps, the benchmark should show it.