Elevenlabs voice cloning has gotten really good now and Claude Opus 5.5 is great at motion graphics, so there's really no excuse not to mak...

It's super easy to find problems with almost any company support AI agent you test. Here's how to do it: https://www.saxifrage.xyz/post/ai-a...

They just casually have their own copy of the internet for evals https://twitter.com/OpenAI/status/2113718797505036665

With AI it takes basically no time to build anything I want to build, so my product roadmap is now organized by how unsure am I about whethe...

Context engineer your colleagues, not just your agents

Made this ascii factory as a visual explainer to teach people how skill optimization through evals works – will open source it soon so you c...

What is a "head of Evals?" (aka the job title I made up for myself). Evals stands for evaluations, specifically evaluating how well AI agen...

I made a simple little toy version of the inner-outer skill optimization loop, to explain to my team how evals connect to business results. ...

I should mention that this was all done with Claude Opus 5.5 because I ran out of Fable. You can just do things! https://twitter.com/hammer_...

In 2020 when I left the 50 person marketing agency I built, I spent a year researching memetics, the study of how ideas spread and evolve. ...

They're just going to tax the datacenters - they're long term capital intensive projects that can't easily move once they're there. Even if ...

Agency holding companies collect talent like Thanos: do everything they can to get one of each type then use them to execute the wrong strat...

1/5th of Anthropics revenue comes from 2 unnamed customers. I can reveal to you the two companies are Meta and @every https://twitter.com/An...

We tested 41 shades of blue with AI agents and they actually do like some colors more than others. https://x.com/hammer_mt/status/2112646804...

The whole six-page memo thing was performative anyway and needed to die. If employees are using AI to jump through the hoops you set up, you...

The unreasonable effectiveness of just looking at the data

Codex voice mode is completely unusable for interviews - it keeps making up its own questions!

don't be a meat proxy for your agent (blindly copying and pasting AI slop without reading it). make your agent a metal proxy for you (a digi...

The most valuable thing you can do right now is find a way to make an unverifiable domain verifiable by AI and let it rip. https://twitter.c...

Optimizing my claude writing skill on my corrections made the future results more correct but made the writing sound less like me. Careful w...