AI Policy and Acceptable Use

Hi all,

There’s still no slowdown in AI. This week the Legible team and I are talking a lot about Jev (a new model promising faster, cheaper performance, particularly for decision-making use cases), the continued relevance of the AI manifesto, and how radically our jobs have changed compared to even six months ago.

Events

On the event front, this Thursday, September 24 at 11AM PT, we are hosting a working session on how to write an AI Policy and Acceptable Use document. Every company deploying AI needs one, and we see a lot of folks getting it wrong. We are going to cover what actually belongs in one, what is just liability text, and demo a tool we built for a client that turns acceptable use into something people can actually learn from. Register here.

What we are writing about

Two posts from this week:

I wrote about why AI manifestos are worth doing at the company and departmental level, not just as a PR gesture. The most advanced companies we work with have a clear written position on where they need to be in 12 to 18 months. The ones that haven’t done the work are burning budget in all directions. Long live the AI Manifesto.

I also wrote about something smaller and a little stranger. Working on proposals last week, I tried to remember how I used to do this before AI. Did I take good enough notes? Did I even send personalized decks? How much was I missing? The old way of working is starting to drift. I’m starting to forget how I used to work.

What we are reading

Tobi Lütke’s interview on The Knowledge Project. Tobi was one of the first CEOs to publish a clear AI manifesto for his company. Hearing how his thinking has sharpened since is worth the time.

Introducing Jev, TypeSafe’s new model. A new, developer-facing model that promises faster performance at a significantly lower price. Showing great promise for classifier and decision-making use cases.

Lawrence