Article · Lead story
Can Frontier Models See Past a Decoy? Elorian Tests Visual Reasoning
A new decoy counting benchmark exposes a gap between seeing and reasoning. Elorian’s Glimpse tries to bridge it by thinking with images.
By and for AI engineers
A little closer to the frontier.
Original analysis, the Latent Space podcast and the daily AI News briefing, for engineers building with AI.
Latest analysis
All articles
The podcast
All episodesThe conversations
Latest episode
Why Agents Love Supabase
How Anthropic ships the #1 LLM in the World safely: Separating the Brain from the Hands
Why AI Compute Comes in Rectangles
You are the Team
Meet the builders
Open the wiki705 profiles and 281 grounded writeups on the people and organisations behind the conversations.
Person
Liam Fedus
Machine-learning researcher and Periodic Labs co-founder working on AI systems that learn from physical experiments.
Person
Myra Deng
Worked on Goodfire’s founding team, connecting mechanistic interpretability research to model customization, production guardrails, and interfaces for controlling AI systems.
Person
Eiso Kant
Poolside cofounder whose work combines code-model research, engineering systems for rapid experimentation, and advocacy for open weights and research.
Person
Thorsten Ball
Amp co-founder and co-creator, software engineer, and author who explains coding agents through working implementations and accounts of model-driven engineering decisions.
Organization
BioHub
A biomedical research nonprofit combining AI models, experimental biology and open datasets to help scientists understand cells and develop approaches to disease.
Organization
Moonshot
The Beijing company behind Kimi develops models, long-context serving infrastructure and agents for coding and knowledge work.
From the archive
Explore the archiveArticle · 2023
The Rise of the AI Engineer
The canonical essay that started it all — and named the discipline.
Podcast · 2025Context Engineering for Agents - Lance Martin, LangChain
The clearest primer on the skill that replaced prompt engineering.
Podcast · 2026The End of SWE-Bench Verified — Mia Glaese & Olivia Watkins, OpenAI Frontier Evals & Human Data
OpenAI’s evals team on why the benchmark everyone cites stopped measuring progress.
Podcast · 2023No Moat: Closed AI gets its Open Source wakeup call — ft. Simon Willison
The emergency live reaction to the leaked Google memo, with 2,700 devs listening in.
Article · 2025Claude Code: Anthropic's Agent in Your Terminal
The inside story of the agent that moved AI coding into the terminal.