|
First issue, so I will keep this simple. I spend my mornings reading the AI news so you do not have to, and I write up what changed and what it means for your work. This week had a clear theme: the money got real, and the oversight showed up right behind it. | |||||||||
|
▮ TOP STORY 
Open weights are back, with a catchMeta released Muse Spark Open, an open-weight model anyone can download for free. Running it at full size is the catch: about a terabyte of memory, like two high-end Mac Studios linked together. The same day, the first laptops built to run models on your desk went on sale, and agents moved into cars and office processes with the permission limits stated out loud. | |||||||||
This week's briefs | |||||||||
| |||||||||
Concept of the week: the harnessYou will hear people talk about "the model" like it is the whole product. It is not. The model is the engine. The harness is everything built around it: what the AI is allowed to touch, what it has to ask permission for, what gets written down, and how a person stops it. This week was a harness week, from Nadella's emergency brake to Shopify asking your approval before an agent buys something.
| |||||||||
One thing to tryTake one small, real task this week and run it through an AI agent or assistant your organization has already approved, with approvals turned on. Let it draft the email, build the list, or summarize the document, and approve each step yourself. Then read the activity log afterward and see if it matches what you thought happened. That one habit, checking the record, is the whole game this year. | |||||||||
One question before you goWhere would you want an emergency brake on the AI tools in your work? Hit reply and tell me. I read every response, and the best ones shape next week's briefs. Jeremy (researched and drafted by my personal AI assistant and approved by me — full articles always at jeremymckellar.com) | |||||||||
