PRADEEP.md

The One Pizza Startup

Around December 2025, seasoned engineers noticed that frontier coding models now generated consistently high-quality code at a low cost. This has changed the dynamics of starting companies.

Om

Writing about someone who wrote for the sheer joy of it is not easy, but I will try.

Running 4B+ models on Apple's Neural Engine

TL;DR: 4B and 9B models on ANE, bypassing CoreML entirely 11.66 tok/s single-stream, 28.62 tok/s with four concurrent streams — GPU and CPU idle Same 16-core ANE on every M1 through M5, regardless of SKU Bottleneck is dispatch overhead, not compute or bandwidth Code: ane.cpp + field guide

My Coding Agent Watches Me Get Frustrated, Then Fixes Itself

pi-mem: surprisingly useful plain-md memory for the pi coding agent

camofox-browser: An OpenClaw Browser That Doesn't Get Blocked

safe-skill-search: Quality-Filtered Local Search for Agent Skills

In my previous post , I analyzed 4,784 AI agent skills across 5 registries. An agent needs to navigate this space quickly and safely.

Agent Skills: 59% Ship Scripts. 12% Are Empty.

TL;DR: 4,784 skills scraped from 5 registries Average quality score ~78, 31% score 90+ 28% are duplicates ~10% install packages via npm/pip No verification, no provenance Treat like early npm—read what you install Browse all skills →