Show HN: Compile English specs into 22 MB neural functions that run locally

1 sources1 storiesFirst seen 4/15/2026Score16Mixed Progress

Single SourceContradictory Claims

Bigness

Coverage

Recency

Engagement

Velocity

Confidence

Clipability

Polarization

Claims

Contradictions

Breakthrough

Sentiment Mix

Positive0%

Neutral100%

Negative0%

Geography

Europe

Expert Signals

yuntian

author • 1 mention

Hacker News

source • 1 mention

AI-Generated Claims

Generated from linked receipts; click sources for full context.

Show HN: Compile English specs into 22 MB neural functions that run locally.

Supported by 1 story

We built ProgramAsWeights (PAW) — https://programasweights.comYou describe a function in English — like "classify if this message is urgent" — and PAW compiles it into a tiny neural program (22 MB) that runs locally like a normal Python function.

Supported by 1 story

No API keys, no internet after compilation, deterministic output.It's for tasks that are easy to describe but hard to code with rules: urgency triage, JSON repair, log filtering, tool routing for agents.

Supported by 1 story

pip install programasweights import programasweights as paw f = paw.compile_and_load("Classify if this is urgent or not.") f("Need your signature by EOD") # "urgent" Compilation takes a few seconds on our server.

Supported by 1 story

Each program is a LoRA adapter + text instructions that adapt a fixed pretrained interpreter (Qwen3 0.6B).

Supported by 1 story

Claim Contradictions

negation mismatch

A: We built ProgramAsWeights (PAW) — https://programasweights.comYou describe a function in English — like "classify if this message is urgent" — and PAW compiles it into a tiny neural program (22 MB) that runs locally like a normal Python function.

B: pip install programasweights import programasweights as paw f = paw.compile_and_load("Classify if this is urgent or not.") f("Need your signature by EOD") # "urgent" Compilation takes a few seconds on our server.