Latent Space

Latent Space

AI & Agents
By
swyx
swyx started it in 2022 for engineers who build on models, and the homepage lists over 194,000 subscribers.
SUBSCRIBERS
194,000+
HOW OFTEN
Not stated
ACCESS
Free
WRITER
swyx, with Richard MacManus
Last updated
August 5, 2026

Preview

Latent Space

What it is.

swyx started Latent Space in 2022, writing for the job he calls the AI Engineer. The about page says it crossed 200k subs and 10m viewers across all channels. The homepage lists over 194,000 subscribers.

Most editions are hour-long interviews with people inside labs and infrastructure companies, published with show notes. Written pieces cover engineering patterns, including one on ontologies as guardrails for probabilistic agents.

The community runs an LLM Paper Club with weekly Zoom meetings, and guests also appear at the AI Engineer Summit.

What you get.

  • Long podcast interviews with engineers from frontier labs
  • Written analysis of engineering patterns, like ontologies for agent guardrails
  • Conference reporting, including trend recaps from AI Engineer World's Fair
  • Show notes and timestamps alongside each episode
  • An open door to the Discord and the weekly LLM Paper Club
OUR PICKS

Editions we'd start with.

5 Trends That Defined AI Engineering at World's Fair 2026 ↗

Names five patterns from the 2026 conference instead of listing announcements.

Ontologies Are So Back: Why AI Agents Are Reviving the Semantic Web ↗

On wrapping probabilistic agents in logical guardrails, worth reading if your agent wanders off task.

Subscribe at the source.

Goes straight to swyx.
Subscribe

Questions about Latent Space

How long are the episodes?
Who is it for?

Questions about AI & Agents

What does an AI engineer do that an ML engineer doesn't?
Why do agent builders reach for ontologies?
Strengths
  • The homepage states over 194,000 subscribers, and the about page cites 200k subs and 10m viewers
  • swyx started it in 2022 and writes for engineers building on top of models
  • Free, and it comes with a Discord plus a weekly LLM Paper Club
Limitations
  • Episodes run over an hour, so budget a proper listen
  • Deeply technical, with little on GTM or pricing
  • How often it lands isn't stated, and posting clusters around conferences
  • The guest list leans toward labs and infrastructure vendors, so vendor framing turns up
Skip this if
  • Skip it if nobody on your team builds on models directly.
Ideal for
  • AI engineers
  • Technical founders
  • AI product leads
  • Platform teams
  • Stage: Startup • Scaleup • Enterprise

Reading is the cheap part.

TripleDart runs organic, paid, and GTM engineering for 300+ B2B tech companies. Bring us the number you have to hit.