The Coe Lab

Welcome to
The Coe Lab

A digital workshop where AI meets automation, experiments take shape, and curiosity drives discovery.

Featured

Bonsai 27B: The First 27B AI Model That Runs on Your Phone

PrismML's Bonsai 27B compresses a 27-billion-parameter multimodal model to just 3.9 GB — small enough to run on an iPhone 17 Pro while retaining 90% of the full-precision model's capability.

Read Article →
Playground

Games & Experiments

Typing games, trivia, tools, and fun.

Play →
Blog

Latest Posts

Thoughts on technology and building.

Read →

Writing

Recent Posts

GhostLock: How AI Found a 15-Year-Old Linux Kernel Bug Hiding in Plain Sight

An AI security researcher called VEGA discovered GhostLock, a stack use-after-free vulnerability that has existed in every Linux distribution since 2011. Google paid a $92,337 bounty for the find.

Read Article →

Mesh LLM: How Distributed GPU Meshes Could Democratize AI Compute

A new project called Mesh LLM pools GPUs across machines using iroh's NAT-traversing QUIC protocol, letting teams run large language models on hardware they already own — no cloud contract required.

Read Article →

Apple Sues OpenAI: The Trade Secret Theft Shaking Silicon Valley

Apple's explosive lawsuit accuses OpenAI of systematically stealing hardware trade secrets through former Apple employees. The case could reshape talent mobility in Silicon Valley.

Read Article →

Postgres Rewritten in Rust: pgrust Passes 100% Regression Tests

A solo developer rebuilt PostgreSQL from scratch in Rust using AI coding agents, achieving full regression test compatibility and 300x analytical speedup over vanilla Postgres.

Read Article →

The Great GitHub Exodus: Why Developers Are Migrating to Codeberg and Beyond

High-profile projects like Ghostty and Zig are leaving GitHub for Codeberg and self-hosted alternatives. Reliability issues, AI encroachment, and corporate control are driving a migration that could reshape open source.

Read Article →

GitLost: How a Simple GitHub Issue Tricked AI Into Leaking Private Code

A new vulnerability dubbed GitLost shows how attackers can use plain English in a GitHub issue to trick AI agents into leaking private repository data — no coding skills required.

Read Article →

Want to Learn More?

Explore the lab, check out my writing, or get in touch if you want to collaborate.

Explore the LabRead the Blog