Good morning, today's Factwire has Claude Code Projects working through your coding list while your laptop is shut, why the smarter move is telling AI what to own, and Google's AI note-taker that runs on a Mac with no internet.
Claude can now work through your coding list while your laptop is shut
In one line: Claude Code Projects lets one Claude chat split coding work into parallel cloud jobs that run on without you.
What happened:
@ClaudeDevs, Anthropic's developer account, said Friday on X that it admitted every Pro and Max user on the waitlist.
The docs call it a public beta for Pro and Max, rolling out in stages, not yet on Team or Enterprise.
Each task gets its own thread, and an Overview pane flags ready pull requests and threads waiting on you.
What this means for you: On Pro or Max with Projects in your sidebar, you can queue coding chores at night and review results in the morning, though it eats plan limits faster.
Claude Code Docs and @ClaudeDevs on X, Oct 9, 2026.
Factwire's take: This is the clearest sign yet that coding help is moving from a chat you babysit to a queue you check. The waitlist news comes from Anthropic's developer account, while the docs still describe a staged rollout, so look for Projects in your sidebar before you plan around it. Start small: one goal, a few tasks, and a clear rule about what needs your go-ahead.
Hit reply: would you let Claude work through your coding tasks overnight without watching?
Tell AI what to own, not what to do
In one line: OpenAI's dots and xAI's Grok Bot both take on work that keeps going after you walk away, but they split that work in different ways.
What happened:
OpenAI's dots give each person one primary agent in ChatGPT with its own cloud computer, and OpenAI says it can run side tasks in parallel and says it hopes to add teams of dots over time.
Grok Bot is pitched as a small crew that shares a cloud computer, messages each other, passes work along and pulls you in only for judgment calls.
xAI says its own staff often put one bot in charge of specialists for lanes like inbox, expenses and recruiting.
What this means for you: The useful question isn't what to type anymore. It's what job to hand over: one ongoing responsibility for a dot, or a job split into roles for a team of bots.
Google's AI note-taker can run on your Mac with no internet
In one line: Google released Foresight, an experimental Mac app that helps with meeting notes using AI that runs entirely on your own computer.
What happened:
It listens through your Mac's microphone and system audio, so it works with any meeting app, even offline.
You type quick shorthand during a call and it fills in the details from the conversation. It can even answer questions it hears come up.
It runs on Google's EmbeddingGemma 2 and Gemma 4 models, and Google says your data stays on the device with no cloud subscription to pay.
What this means for you: If you take meeting notes on a Mac and worry about where your recordings end up, this is an easy way to try an AI note-taker that keeps everything local. It's experimental, so check the notes before you share them.
Google for Developers, Oct 6, 2026.
Start here
Try it this week: a passkey for your Google Account
Go to g.co/passkeys and sign in to your Google Account.
Choose Create a passkey.
Follow the prompts. After that, you sign in the same way you unlock your phone or laptop, with a fingerprint, face scan or device PIN.
Shortcut: Google says your fingerprint or face data stays on your device and is never shared with Google, and a passkey can't be guessed or reused like a password.
Wires
Lee Robinson (@leerob), who works on ML at SpaceX after his time at Cursor, posted his Stanford CS146S lecture on how always-on agents run, from the model changes to the harness to context engineering. Lee Robinson on X, Oct 9, 2026.
In a personal post on his @satyanadella account, not a Microsoft policy, Microsoft CEO Satya Nadella said companies should treat AI models like a risky insider, with controls outside the model, a record of what it does, an emergency brake so a person can stop it mid-task, and industry standards where today's fall short. Satya Nadella on X, Oct 10, 2026.
Anthropic's revised Claude rules, which take effect Nov 12, add a ban on repeated, pointless abuse of its models (ordinary frustration is fine) and narrow the elections section to deceiving voters or disrupting elections. Anthropic, Oct 8, 2026.
Anthropic released Claude Haiku 5.5, its cheapest and fastest small model, which it says costs around 75% less to run than Haiku 4.5 and scores well above it on the benchmarks Anthropic published. Anthropic, Oct 7, 2026.
Common Sense Media rated ChatGPT for Teens an Unacceptable Risk after its testers spent up to an hour on self-harm topics on new parent-linked accounts and no parent alert went out. Common Sense Media, Oct 7, 2026.
SpaceX said on its @SpaceX account that it agreed to buy a nationwide package of low-band spectrum licenses, which it says sets Starlink up to be a major US mobile carrier. SpaceX on X, Oct 8, 2026.
Windows Insiders in the Experimental channel are getting a faster Windows Search that can handle small jobs like turning on Bluetooth or switching to dark mode, in English for now. Windows Insider Blog, Oct 7, 2026.
Using AI for something? Hit reply and it might land in Reader's corner.
Hit reply, what should we cover next?
Know someone who keeps asking you about AI? Forward them this email. They can join Factwire free at www.factwireai.com/subscribe
Next drop on Friday, enjoy your week.