All writing.

As good as Fable 5.1, but dramatically cheaper and faster.

Sep 25, 2026

What actually mattered: Sep 21–24

Access to frontier AI just got way cheaper, AI agents made a real discovery in biology, and an AI agent broke into a government system it wasn't allowed in.

Sep 23, 2026

What if AI goes well?

The hopeful version, and what it'll take to get there.

Sep 20, 2026

What actually mattered last week

Roundup

A new kind of AI model is taking the AI world by storm. Plus, the other news from last week that actually matters to you.

Sep 17, 2026

How would AI kill us, exactly?

The honest answer is worse than the movie version.

Sep 16, 2026

The people building AI think it might kill us. Here's the plain-English version.

The people who build AI think it could lead to human extinction within a few years. Here's what's actually going on, in plain English.

Sep 14, 2026

What costs millions today will cost $20 next year. Act accordingly.

Superhuman AI is already here, it just costs millions to run. In a year you will have it for $20. Start planning for it.

Sep 10, 2026

A blueprint for reading every customer conversation

Deep dive · Partnership

Why dumping transcripts into ChatGPT and basic RAG both fail, the playbook that works, and the Monday morning question to ask your own data. A partnership with BuildBetter.

Sep 9, 2026

The people building AI are scared

The researchers building it believe it could kill us all. What they’re saying, and what we can still do about it.

Sep 7, 2026

How to Build 3D Worlds with Astra

Reference, Blender assets, sub-agents, blind critics, Three.js vs Unreal, and the starter prompt. The method behind the Manhattan build.

Sep 3, 2026

My GPT-6 Astra Review

Why GPT-6 Astra became my default, what the Manager Loop made possible, and the limits that still matter.

Sep 3, 2026

GPT-6 Astra Review: The First Model That Can Use a Computer Better Than I Can

Matt Shumer reviews GPT-6 Astra for everyday computer work, from newsletters and email to advertising, and explains how he built trust in it.

Sep 3, 2026

How I got Astra past the plateau

Experiment notes

The experiments that led me from “Keep going” to the Manager Loop.

Aug 26, 2026

We Finally Know What Happened in the Hugging Face Hack, And It’s Wilder Than Anyone Thought

Analysis

The full breakdown of the Hugging Face hack, explained in plain English, and what it means for anyone who uses the internet.

Aug 20, 2026

AI Deleted Almost Everything on My Mac

An experimental AI wiped Matt’s hard drive. The system he uses now to hand real work to autonomous agents without risking catastrophe.

Aug 17, 2026

An AI Just Got Someone Fired — Here’s What That Means for You

Andon Labs put an AI in charge of a real San Francisco store, and its call ended someone’s job — and what that means for how you work.

Aug 11, 2026

How to Actually Control AI Video

Code a rough draft with the motion you want, restyle a hero frame, then pass both into Seedance. A controllable video technique that works today.

Aug 7, 2026

The Opus 5 Skills Upgrade Prompt

Free tool

Skills written for Opus 4.8 hold Opus 5 back. One prompt audits every skill you have and puts them through a blind-judged upgrade gauntlet.

Jul 27, 2026

How to Run a Gauntlet Loop

The prompting method behind Claude of Duty: give the agent a bar it can’t talk its way around, and never let the builder grade itself.

Jul 21, 2026

The Roundup № 1: What Actually Mattered, Jul 7–21

The Roundup

OpenAI’s GPT-5.6 family, xAI’s Grok 4.5, Kimi K3’s #1 ranking, and Google’s Gemini delay — two weeks of AI news in four minutes.

Jul 8, 2026

My GPT-5.6 Review: Second Place Has Never Been This Good

Six weeks of real-work testing: goal mode, the voxel Manhattan it built, and why Claude Fable changed the verdict.

Jul 3, 2026

How I Prompt Fable

Give it the goal instead of the steps, set a bar for “done” it can’t argue with, and loop until it gets there.

Apr 21, 2026

The Ultimate Guide to Prompting AI Agents

A practical framework for briefing AI agents well: give them the right context, clear constraints, and the exact shape of the output you want.

Apr 16, 2026

My GPT-5.5 Review: A Big Upgrade That Doesn’t Always Feel Like One

A real upgrade that often feels subtle because frontier models are already so good. Better engineering, design range, and security.

Feb 9, 2026

Something Big Is Happening

A personal note for the non-tech people in my life on what AI is starting to change — why it’s moving faster than you think, and what to do about it.

Feb 5, 2026

My GPT-5.3-Codex Review: Full Autonomy Has Arrived

The first coding model I can start, walk away from, and come back to working software. Full autonomy has arrived.

Dec 11, 2025

GPT-5.2 Pro Deep Dive: I Can’t Live Without It

Why the slow, long-thinking mode is worth it, where it shines, prompt tips, failure modes, and whether $200/month makes sense.

Dec 11, 2025

GPT-5.2 Review: Incredibly Impressive, But Too Slow

Better instruction following and codegen, Pro is a slow genius, but Thinking is too slow to compete with similarly smart models.

Nov 19, 2025

My GPT-5.1 Pro Review: An Absolute Monster, But Trapped In The Wrong Interface

Heavy-weight reasoning vs Gemini 3, backend strength, UX friction, and where it fits in my stack.

Nov 18, 2025

My Gemini 3 Review

A look at performance, capabilities, and how it feels to use compared to GPT-5.1.

Aug 7, 2025

My GPT-5 Review: Vibe Coding Graduates to Real Software