The Signal · this week in AI

UnderstandMachine Intelligence

How modern AI actually works - Mechanistic Interpretability, the systems builders ship, and a weekly signal on what matters.

Free · the weekly Signal + new explainers · no noise

coversMechanistic InterpretabilityLLM internalsAgents & MCPAI securityNew tools
00:00:00
[Sector // Forum]

Inside the black box.

A working community for builders untangling how AI actually works - interpretability, agents, MCP, safety, and what shipped this week. Specific questions, real answers.

00:00:00
[Sector // Surfaces]

Everything else, in one place.

The forum is the heart of it. These are the surfaces that feed it - the explainers, the weekly signal, and the tools you wire in.

[Learn]01 // 03

From black box to clear mental model.

Hand-crafted explainers - what mech interp is, how LLMs work, and how models become software that acts.

[The Signal]02 // 03

Keep up without drowning.

New models, tools, MCPs, and research - each with a one-line take on whether it matters or it's hype.

[Tools & MCP]03 // 03

The new tools - and what they can touch.

MCP servers, dev workflows, RAG stacks, and evals - each with a threat model attached.

[Sec-07 // Broadcast]

Understand AI before it runs everything.

One email a week: the new explainers, the tools worth trying, and the research that actually matters. Free, no noise.

404 // Not Found is not an option