Using ChatGPT 5.6 for knowledge work
AI Weekly Round-Up #42: Claude can now send Outlook messages and the impact of Slack "tag"
Hi friends,
Who’s excited for this World Cup final?1
But first, I’ve got some personal news to share.
This won’t impact anything AI-related, but make sure you’re subscribed to my other newsletter RadReads for updates.
And if you’re new to the newsletter, every week I curate the top articles, podcasts, videos and tutorials in AI (so that you don’t have to).
Let’s jump in:
If you only read 3 things this week
How GPT-5.6 Changes Knowledge Work — This model inside the new ChatGPT for Work product is fast, smart and (relatively) cheap. I’m using it a lot more than Opus/Fable.
5 Trends That Defined AI Engineering at AI Engineer World’s Fair 2026 — AI engineers are the canary in the coal mine for how knowledge work will evolve.
Demis Hassabis’s Framework for Frontier AI and the Dawning of a New Age — Google DeepMind founder proposes a FINRA-style regulatory framework.
The easiest way to build AI agents for knowledge work
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.
From the desk of Khe Hy
Should you switch from Cowork to ChatGPT for Work — A framework for doing agentic knowledge work.
A Fable 5 vibe check ⤵️
What the frontier labs are cooking up
Fable 5 gets officially added to Max plans — The flip-flopping on pricing seems to have abated.
Is the new Kimi model a threat to frontier labs — The new open source model performs extremely well on benchmarks. Should the labs be afraid? Or are these companies “cheating” via distillation?
Claude Can Now Send Your Email. Should It? — The Microsoft 365 Connector quietly got more powerful this week.
AI and the future of work
Why I Think AGI Just Happened — Slack “tag” is an entirely new form factor for interacting with LLMs. It’s not getting a lot of press, but I think this is a BIG deal.
Why OpenAI and Anthropic Won’t Win the Enterprise App Layer — Glean’s founder on cost, permissions and other enterprise adoption challenges. (They compete with the foundation labs, despite relying on their APIs.)
Citadel’s Ken Griffin on AI, U.S.–China Tensions, and Data Centers — How Citadel’s quants are leveraging AI in their investment process.
How Groq Went From Near Death to a $20B NVIDIA Deal — Groq founder (Jonathan Ross) on getting acquired by NVIDIA and how chips run inference.
For our AI builders
Why Every Engineer Will Need a $100,000 Token Budget — An inside look at an AI-native organization serving Fortune 100 clients.
How Sierra AI-Pilled Its Company: Lessons Learned — The same company describes how they built their AI agents (for non-coders).
The correct way to build and manage AI agents — Devin/Cognition describe their push for cloud-based agents (versus the local ones we’ve gotten used to).
The easiest way to build AI agents for knowledge work
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 opened 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.
I have watched more football this month than in my entire life. I’m totally hooked and in the process of figuring out how to keep following in Europe. Current ideas are Aston Villa, Dortmund or Atlético Madrid — not interested in the Yankees-style teams. What do you think?