LogoSignal
Issue 001 — March 2026Join waitlist
0.820.67−0.40.91SIGNAL · VOL. I

Signal

Conversations with the minds training the future.

A weekly newsletter that sits down with the researchers, engineers, and founders building the machine intelligence layer — capturing how they think, not just what they ship.

Senior ML EngineersPhD CandidatesTechnical PMs
Hold My CopyFree. No spam. Unsubscribe anytime.

Vol. 1 — 2026

Scroll to read
I. THE PROBLEM
The Problem

The field moves so fast that nobody stops to ask why.

Every week, a new architecture. Every month, a new benchmark. Every quarter, a new lab claiming AGI is eighteen months out. The discourse accelerates and nobody asks the only question that matters: why did you choose that loss function?

What made you leave Google Brain at the peak of your career? What does it feel like to train a model that surprises you? The answers live in the minds of the people doing the work — and they're not in the arxiv abstract.

"Speed is not a virtue in journalism. It is a symptom of fear."

— Signal Editorial Manifesto, 2026
Continue reading →
II. THE FORMAT
The Format

Long-form profiles. Not five-minute podcast clips.

01

Long-form profiles

4,000–8,000 words. Not a listicle. Not five bullet points. A genuine account of how a person thinks, built from 3–4 hours of recorded conversation.

02

Primary source, always

Every claim traces back to the person who made it. No paraphrase without consent. No 'sources familiar with the matter.' Just the researcher, in their own words.

03

One subject per issue

One mind. One hour of reading. No sidebar interviews, no roundups, no listicles. The format is the message: depth is the point.

Each issue ships on a Sunday morning. You read it like a broadsheet — slowly, with coffee, without notifications.

III. THE SUBJECTS
Forthcoming Profiles

Three minds. Three questions nobody else asked.

Issue 001

Dr. Yuna Sato

Research Scientist, Anthropic

"I didn't leave academia because I wanted more compute. I left because I wanted to be wrong faster."

Issue 002

Marcus Adeyemi

Co-founder, Cohere

"Every time I explain a transformer to a non-engineer, I understand it differently. That's not a metaphor."

Issue 003

Priya Venkataraman

PhD Candidate, MIT CSAIL

"The loss function is a hypothesis about what matters. Nobody talks about that. They talk about the number it outputs."

Founding subscribers receive all three interviews free. No paywall, ever.

IV. THE CONVICTION
What We Believe

The most important ideas in machine intelligence are not in the paper. They are in the conversation before it.

Signal believes that the researchers, engineers, and founders building machine intelligence deserve a publication that treats them as thinkers, not just producers of benchmarks. We are building that publication — slowly, deliberately, with the care that the subject demands.

Issue 001 ships in March. Early subscribers get every founding interview free.

No name. No preferences. No friction. One field. One click. You're on the list.

847 readers already on the list

Free. No spam. Unsubscribe anytime.