Good morning, today's Factwire has Google's AI study coach, Goldman Sachs limiting AI coding tools, and an MIT claim that one AI fixed another.
Learn AI one step at a time. The news worth knowing, checked against the source, plus 3 things to try. Every Monday and Friday.
Google puts an AI study coach in Google Classroom
In one line: Google says Guided Learning is now in Google Classroom today, and it coaches students with questions instead of handing over the answer.
What happened:
A student starts a Guided Learning activity their teacher assigned, and it works one on one with the class material.
Google plans a Search pilot in some US schools later this month.
Over the next few months, Gemini and AI Search will be off for K-12 student accounts until a school turns them on.
Why it matters: If a student in your house uses Classroom, this AI is built to coach their thinking, not hand over answers.
Google, Nov 3, 2026.
Factwire's take: Coaching beats copying, as long as students keep using it that way.
Hit reply: yes or no, would your family use an AI study coach?
Musk said vs. the record
Starlink dish: Musk said all early Starlink customers get a free new dish. @elonmusk on X, Nov 4, 2026.
Confirmed, with a limit: Starlink's help page offers a free Standard kit to eligible early customers still using a first-generation kit, by email invitation, so not everyone.
Record: Starlink support page, page undated, read Nov 5, 2026.
Starlink Mobile: Musk said there will at last be one network with good signal everywhere. @elonmusk on X, Nov 4, 2026.
Pending: SpaceX picked Qualcomm chips for its 5G base stations, and its US rollout is planned to begin in 2027, so nobody has that signal yet.
Record: Qualcomm's release, Nov 4, 2026, and Starlink on X, Nov 4, 2026.
Grok Bot shopping: Musk said Grok Bot will research purchase ideas, bring back a selection, order them, track them and return what you don't want. @elonmusk on X, Nov 4, 2026.
Confirmed in part: the changelog lists shopping results, orders you approve and order tracking. Returns: No record yet of the Bot sending items back for you.
Record: Grok Bot changelog, Oct 27, 2026.
Goldman Sachs limits AI coding tools to its engineers
In one line: Bloomberg reports Goldman now limits Claude Code and GitHub to engineers, citing internal guidance updated in October.
What happened:
For months, Bloomberg says, Goldman let bankers use those tools for work. Now it limits them to engineers.
Bloomberg says Goldman reviewed productivity and risk concerns before limiting the tools.
Why it matters: If your job uses AI tools, check what your employer allows before you paste work into one.
Bloomberg, Nov 2, 2026.
MIT researchers say an AI model fixed a small model by editing its weights
In one line: MIT CSAIL researchers say, in a blog post, that GPT-6 Astra fixed a small model's looping by editing 0.016% of its weights.
What happened:
In their tests, that edit lifted the model's MATH-500 score from 7.9% to 48.1%.
The gain came from stopping the model from looping, not from teaching it new math skills.
Why it matters: It hints that AI could one day help check and fix other AI without retraining, but this is early work.
MIT CSAIL researchers, Nov 2, 2026.
Start here
Try this: turn any AI into a study coach
Open Gemini, Grok Bot or ChatGPT and paste in a problem you're stuck on.
Type: Don't give me the answer. Ask me one question at a time so I can work it out myself.
Answer each question in your own words, and ask for a hint only if you're stuck.
Want to start with the basics? Get the free Grok Bot guides.
Word of the week: weights. They're the numbers inside an AI model that hold what it learned.
xAI facts
Model or product: Grok text watermark, for seven models (grok-4.7, grok-4.6, grok-4.5, grok-4.3 and three grok-4.20 versions)
Price: Not stated
Docs: Grok watermarking docs
Who can use it: Applies to text from Grok models used through the developer API (the way apps plug in). Team admins can opt out per model, per the docs page
Date: Starts Nov 7, 2026, per the docs page (the page itself is undated)
Wires
Apple says any app developer can now connect to Siri AI in the iOS 27.3 betas. Apps could soon act when you ask in your own words, and customers get it in a few months. Apple Developer, Nov 3, 2026.
From Nov 7, text from seven Grok models used by developers (through an API, the way apps plug in) carries an invisible watermark. Team admins can opt out. xAI, Nov 7, 2026.
Anthropic says text from Claude Mythos 5, Opus 4.7 and Sonnet 4.6 carries an invisible watermark on listed platforms, except some Sonnet settings. Anthropic, Nov 3, 2026.
OpenAI says its dots helpers can now also use GPT-6.1 Terra and pick different models for different parts of a task. OpenAI, Nov 3, 2026.
Anthropic says message threads, a developer beta, cut the median wait for a first answer by about 35% on very long Claude Code requests. Anthropic, Nov 2, 2026.
One Grok Bot user says it phoned three shipping companies and came back with exact quotes. He says that saved him a bit over 40 minutes. @TimDOES on X, Nov 3, 2026.
Tencent says its new Hy4 model can read charts and operate a computer. Open model files are promised in the coming days, so they are not out yet. Tencent, Nov 4, 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 Monday. Have a great day and enjoy your weekend.