12 June 2026

AI learned your job. Now it files for IPO.

Skills that remember your habits, a 269-page AI bill with teeth, and the week AI stopped being anyone's hobby.

---

Hello,

Two posts this week — one practical fix, one rather large week in AI news.

---

How AI Skills Turn Workflows into Reusable Instructions https://daehnhardt.com/blog/2026/06/10/what-are-ai-skills/

If you have ever explained your preferred format to an AI assistant for the fourth time in a row, this one is for you.

Think of a skill like a recipe card pinned inside the kitchen cupboard — the instructions are already there, the next time you need them, no dictating from memory. A skill is a small folder with a SKILL.md and any templates or scripts the task needs, loaded automatically when relevant.

Cheekily, the most underrated line in any skill is the one that says “do not flatten the author’s voice.” Without it, the assistant will helpfully sand away exactly the things that made your writing yours.

The post covers where to find curated skills to start with, how to write your first one in about twenty minutes, and why the “Do not” section does more work than all the positive instructions combined.

Meanwhile, the week’s larger story was playing out at rather a different altitude.

---

Who Controls AI? Fable 5, OpenAI’s S-1, and a 269-Page Bill https://daehnhardt.com/blog/2026/06/12/who-controls-ai-fable-5-openai-s-s-1-and-a-269-page-bil/

This was a big week. A 269-page draft federal AI governance bill landed in Congress — with mandatory third-party audits, whistleblower protections, and proposed fines of up to a million dollars per day. That is the sort of number that tends to focus attention in ways voluntary commitments never quite manage.

On the exact same day the bill dropped, OpenAI confirmed its S-1 filing. Their announcement read: “We expect it to leak so we’re just announcing it” — candid, to say the least. Two of the three biggest frontier AI labs are now heading for public markets simultaneously.

Anthropic released Claude Fable 5 on 9 June: an 11-point jump on SWE-bench Pro over Opus 4.8, large enough to actually change what your agentic pipelines can do. Apple reduced the cloud-inference cost barrier for Foundation Models for many smaller iOS developers at WWDC. And Jeff Bezos brought a $41 billion physical-world AI company out of stealth — all in the same seven-day stretch.

The common thread is not benchmarks. It is the messy, overdue process of deciding who funds frontier AI, who audits it, and on what terms it reaches the rest of us.

Take care, and do feel free to hit reply if anything sparked a thought.

Elena

---

← All issues