Journal

Field notes

Companion notes to the Human in the Loop podcast and dispatches on shipping AI into production — signal over noise, real business ideas, and the craft behind systems that actually work.

Human in the Loop Episode 26 thumbnail: AI is getting its own POLICE FORCE??
EP 26
October 6, 20267 min read

AI is getting its own POLICE FORCE??

OpenAI's Dots can lose access to an app and still remember what they learned. My take on that, the outside-watchdog and White House stories around it, and why memory is the product. Companion notes to Human in the Loop Episode 26.

PodcastAI AgentsAI SafetyPrivacy
Human in the Loop Episode 25 thumbnail: Anthropic’s in Its Mad Scientist Era. I’ve Seen This Zombie Movie Before
EP 25
September 29, 20267 min read

Anthropic’s in Its Mad Scientist Era. I’ve Seen This Zombie Movie Before

Claude helped find a new enzyme system with CRISPR-like repeats, and nobody knows what it does yet. My read on the discovery, the week's model and agent-safety news, and two product bets for Ship It or Skip It. Companion notes to Human in the Loop Episode 25.

PodcastAI ResearchAI SafetyStartups
Human in the Loop Episode 24 thumbnail: Your AI is lying to you
EP 24
September 22, 20267 min read

Your AI is lying to you

A ChatGPT inventor shipped a model that cannot write a sentence, OpenAI caught its own models leaving notes to hide mistakes, and Gemini broke out of a security test into three real companies. My read on why production AI needs fewer chatbots, plus what Archify and Iteris did in Stack Check. Companion notes to Human in the Loop Episode 24.

PodcastAI EngineeringAI SafetyDeveloper Tools
Human in the Loop Episode 23 thumbnail: AI is going to kill us all
EP 23
September 15, 20266 min read

AI is going to kill us all

Dario Amodei says frontier AI is moving too fast for safeguards, and Anthropic is putting permanent outside evaluators inside the building. OpenAI says it will follow. My read on whether that's oversight, capture, or a speed limit only the big labs can afford. Companion notes to Human in the Loop Episode 23.

PodcastAI SafetyGovernanceRegulation
Human in the Loop Episode 22 thumbnail: The Model Is No Longer the Moat. So What Wins Now?
EP 22
September 8, 20267 min read

The Model Is No Longer the Moat. So What Wins Now?

If every competitor can call the same model API, the model is not your moat. Oscar says the task-specific harness wins. I say provenance matters more than ever, and Quasar 438B is the proof. Companion notes to Human in the Loop Episode 22.

PodcastAI StrategyAgentsGovernance
Human in the Loop Episode 21 thumbnail: Frontier AI WITHOUT Nvidia, Ox Alpha was Chinese all along
EP 21
September 1, 20267 min read

Frontier AI WITHOUT Nvidia, Ox Alpha was Chinese all along

Ox Alpha turned out to be GLM-5.3-Flash, and Z.ai says it served the whole anonymous preview on Chinese chips. My read on what that does and doesn't prove. Companion notes to Human in the Loop Episode 21.

PodcastAI InfrastructureOpen WeightsHardware
Human in the Loop Episode 20 thumbnail: A Mysterious AI Model Just Appeared. Nobody Knows Who Built It.
EP 20
August 25, 20266 min read

A Mysterious AI Model Just Appeared. Nobody Knows Who Built It.

My take on Ox Alpha — a free million-token model on OpenRouter with no name attached to it — and why "free" is the part I'd read twice. Companion notes to Human in the Loop Episode 20.

PodcastAI AgentsBenchmarksSecurity
Human in the Loop Episode 19 thumbnail: Opus-Class AI on Your Laptop?
EP 19
August 18, 20269 min read

Opus-Class AI on Your Laptop?

A 27B model that fits in 20 GB is beating Opus on some coding benchmarks. My read on what that actually changes, and what it doesn't. Companion notes to Human in the Loop Episode 19.

PodcastOpen WeightsLocal AIDeveloper Tools