AI that works. News that's ranked. No filler. https://nitter.gq/t.co/VrgZrPJPUZ
UK
Joined July 2009
- Tweets35.1K
- Following2.9K
- Followers14.9K
- Likes2.5K
Who looks after your AI tools over Christmas?
What's happening: the person who set up the automations is often the one taking the longest break, and the tools keep running while they're away.
Three things to do before December:
1. List every automation that sends, books or updates something on its own.
2. Write down who can pause each one.
3. Check the cover person has the access to do it, or knows who does.
Firm of one, AI governance edition. One owner, one approver, one person who explains it to the client when it goes wrong.
Same person. Write it down anyway, for the day you hire.
Usage is the easiest AI number to collect and the worst guide to who uses it well.
Glean's survey (vendor-run, self-reported) found 54% of heavy users sometimes hand in AI work they couldn't explain if asked. Light users: 24%.
So who's your team copying?
readanswerable.com/p/your-he…
Local politics is now part of your AI vendor's delivery risk. Underwrite it....
Protesters breach Microsoft data-centre site in Amsterdam
About 100 activists entered a planned Microsoft data-centre construction site in Amsterdam, with some climbing a crane. The Verge and Reuters link the action to a wider European campaign against AI infrastructure. The project proceeds despite Amsterdam's ban on new data centres until 2030 because it was planned earlier.
theverge.com/news/1028549/pr…
On one OpenAI comparison, an older model got slightly more answers right than a newer one but was wrong on 75% of questions, against 26%. It almost never said "I don't know".
A headline accuracy figure can hide a lot of guessing.
aiprompthackers.com/p/ai-mix…
Website or Meta enquiry lands → an AI step in Zapier works out the case type → drafts a first reply with qualifying questions → creates the record in Clio Grow → texts, emails and phones the lead → the intake team is alerted. A lawyer still reviews client-facing messages.
Rozas Law Firm, 34 staff, Louisiana. It says that saves more than 66 hours of admin a month. Vendor story.
Six steps, one named human check. Would your firm be happy with that ratio?
Spaces for Muse. Boards for agents. Still no owner for the decision. Tools multiply. Accountability doesn't...
Meta adds Spaces so Muse agents can build project workspaces from chat
Meta added Spaces to Muse: describe a goal in chat and Muse builds a dedicated tracker, planner or project board that can stay updated from connected apps. Zuckerberg said Muse agents already had about 10 million users in the US and Canada within two months of launch.
techcrunch.com/2026/11/06/me…
The AI tax bill has a new line item: time you never spent.
For paid subscribers, on what happens when AI savings meet hourly billing, and why the IRS and the AICPA read it differently.
readanswerable.com/p/ai-savi…
Does your prompt deserve to be a skill? Four checks:
1. You've pasted it more than a handful of times this month.
2. The job stays the same while the details change.
3. You want the answer in the same shape every time.
4. You keep making the same correction afterwards.
aiprompthackers.com/p/you-ke…
OpenAI only raised eleven billion at a mere one-point-two-five trillion. Tragic. Somewhere a founder is comforting themselves with a yacht the size of Wiltshire. Markets are so cruel...
OpenAI raises about $11bn so far at a $1.25tn valuation
Bloomberg-sourced reports say OpenAI has raised roughly $11 billion in a still-open round valuing the company at $1.25 trillion, well below an earlier aim of about $30 billion at $1.4 trillion after the midterm-driven AI selloff. The round could still grow toward about $20 billion in coming weeks.
theverge.com/ai-artificial-i…
Before your next one-to-ones, write a short paragraph for each role saying what good AI use looks like in that job. Then ask each person what they need to get there.
readanswerable.com/p/would-y…
Prompt that gets a draft: "Reply to this customer."
Prompt that gets a reply you can send: "Reply to this customer. They want a refund date. We can't give one before Friday. Don't apologise twice."
Washington is withdrawing its 1978 guide to fair hiring tests, the one the four-fifths rule comes from. The law underneath doesn't change.
Tempting: stop asking your hiring tool for a bias report. Worse: tidy away old hiring data. One employment law firm warns that can look like destroying evidence.
readanswerable.com/p/ai-regu…
When researchers asked models including GPT-4 how confident they were, the answers mostly sat between 80% and 100%, usually in multiples of five, right or wrong.
Ask where a claim came from instead. That you can check.
aiprompthackers.com/p/ai-mix…
Anthropic's Claude Duet is pitched at small business owners. It sends a morning brief, checks your connected apps every few hours and drafts on its own. Before it sends, pays, signs or deletes, it shows you exactly what it's about to do and waits. Moving money between accounts is off-limits even with approval.
Sensible design. The approval step is only as good as the person tapping Approve at 7am.
In Anthropic's trial with 52 mostly junior engineers, those who handed the whole task to AI finished fastest and averaged under 40% on a quiz about the code they'd just produced.
Those who only asked it conceptual questions scored 65% or more.
Guess which group tops the usage dashboard.
readanswerable.com/p/your-he…