Hi friends,
I hope you’re having a great weekend. We’re embarking on our world trip in 17 days (and created the family Instagram account).
If you’re new here, every Sunday I curate the top AI blogs, tweets, podcasts and tutorials (so that you don’t have to).
Let’s jump in.
If you only read 3 things
How Kepler built verifiable AI for financial services — LLMs are “probability machines” and struggle with tasks that require 100% precision.
Details from the OpenAI Hugging Face incident — This is wild. I’ve read summaries of this hack but the actual examples of how it went down are straight out of a sci-fi movie.
Building and structuring an AI-native company — AI is much more than 20% productivity upgrades once you build self-learning agents.
From the desk of Khe Hy
Here’s what the frontier labs are cooking
OpenAI launches Computer History — With your permission, ChatGPT tracks every click (and feeds it into future context).
GrokBot launches personal agents — This looks like an OpenClaw for everyday users. And the early acclaim is very positive.
100 use cases for GrokBot — Cursor’s CEO shows the power of an agent that connects with all of your tools.
How Claude marks AI-generated content — The EU is requiring watermarks for all AI-generated text. Yup, you read that right.
The future is for everyone — Zuck’s 6,500 words on decentralized and open source AI. He’s going to compete on price, which should benefit all consumers.
Anthropic in talks to acquire Decart — Boosts inference efficiency and builds world models.
Create personal leverage using Claude Cowork
Imagine sitting down Monday morning and your meetings are already prepped, your inbox is triaged, and last week’s notes are organized — all before you open your laptop. That’s Claude Cowork running on a schedule you set once. This self-paced course teaches you to build that system. Finish it in a weekend.
Use code FABLE-5 to save $100 until August 21st.
For our builders
What is a personal agent — A primer on why these highly coveted Personal Agents are nothing more than a few tool calls and some memory.
The 5 problems in building personal AI — Everyone wants their personal agent. But hardware, memory and cost are all real challenges for mass adoption.
AI adoption is a myth — You can train power users. But how do you design AI systems for the remaining 70%+ of an organization who just want these tools to work in the background.
The company brain has a permissions problem — Permissions are really complex in an agent-first world. Do we need a new model beyond “yes, you have access” or “no”?