mAIn Street #294: Meta launches a new agentic app for Mac; What is Gemini Student Lab?; Alexa+ is free on FireTV-compatible devices



Meta turns its AI into a Mac work app, Google opens a student hub, and Alexa+ goes free on Fire TV.
mAIn Street Daily Newsletter. AI news for people who actually have jobs to do.
Thursday, August 20, 2026
On September 9, I’m hosting ChatGPT for Marketing: Using AI So It Doesn’t Look Like AI, a 90-minute workshop for marketers and business professionals who are tired of generic copy and look-alike visuals. We’ll work with writing, images, fonts and web pages, then use real examples to show how references, clear instructions, repeated edits and human judgment produce work that feels specific to you. See the workshop details and register.
Top 5
What matters today
01
Share a window and Meta AI can answer questions, suggest changes or create material from what it sees. It can also connect to Instagram, Facebook, Meta ad campaigns and Google Workspace to analyze performance, build documents and run recurring reports. Check the data terms first: information shared with Meta may be used for model training and ad targeting.
Source: Axios
02
Gemini’s new student hub can collect research in a study notebook, create practice material and add dates from a syllabus to Google Calendar. Deep Research can keep working after the phone is locked, while Lens can explain a photographed worksheet or coach a student through a mistake. Eligible U.S. students can also receive one year of Google AI Pro at no cost.
Source: The Verge
03
The conversational assistant can search across streaming services, recommend programs and control smart-home devices. Compatible Fire TV owners will receive it automatically, including people without Prime who previously faced a $19.99 monthly price.
Source: TechCrunch
04
Notetaker works with Zoom, Google Meet and Microsoft Teams, even when the meeting was scheduled somewhere else. Recaps are private until the user shares them, and notes can be revised first. Calendly says attendees receive recording notices and can stop the recording. Access is rolling out gradually to eligible users.
Source: TechCrunch
05
The company’s automated HIVE labs grow 20 kinds of human tissue, apply experimental treatments and track what happens. Vivodyne says this controlled, causal data can help AI systems predict how drugs may behave in people. The approach is promising and still has to prove itself at scale.
Source: TechCrunch
AI Workflows
Use AI to make sense of a home purchase or sale
A house deal comes with people, documents and fees most of us encounter only a few times. AI can help you keep the moving parts straight.
The national housing market is awkward for everyone. Existing-home sales fell 2.4% in June even as the median price reached $440,600. The average 30-year mortgage rate was 6.67% on August 13. Buyers have more listings to consider in many places, yet prices and monthly payments remain difficult. Sellers have equity, yet a smaller pool of qualified buyers can make pricing mistakes more expensive. Your local market may look very different from the national numbers.
AI is most useful here as an organizer, calculator, document translator and question builder. Give it the actual materials from your transaction, require page references, and make it separate facts from assumptions. The people handling the transaction still confirm the numbers, legal language and final decisions.
Start with a map of who does what
A listing agent generally represents the seller. A buyer’s agent generally represents the buyer. Dual agency, designated agency, transaction brokerage and other arrangements depend on state law. Your lender handles financing. The appraiser develops an independent value opinion for the lender. The inspector evaluates condition. Title, escrow or settlement professionals examine ownership records, move money and prepare the closing. Attorneys have a larger role in some states and transactions than others.
Agency relationships vary widely by state, so ask early who represents you, what duties that person owes you and where a conflict could arise.
Prompt: I am the [buyer/seller] in a home transaction in [state]. Using only the agreements and disclosures I provide, make a table showing each person or company, who hired them, whom they represent, their responsibilities, their fee, when they are paid and which decisions remain mine. Quote the exact contract language and page number for every conclusion. Put state-law questions in a separate list for my agent or attorney.
Put every fee in one place
Broker fees and commissions are negotiable and are not set by law. Many buyers working with an MLS participant will sign a written buyer agreement before touring a home. That document should state the services, compensation and length of the relationship. Sellers should find the same details in the listing agreement, along with the marketing plan and cancellation terms.
The full cost can also include lender charges, discount points, an appraisal, inspections, title or settlement services, recording fees, transfer taxes, prepaid taxes and insurance, HOA charges, repair credits, seller concessions and moving costs. Some amounts are estimates. Some depend on the final price or closing date. Some can be negotiated.
Prompt: Build a fee ledger from these documents. Use columns for the amount or formula, who charges it, who pays it, when it is due, whether it appears negotiable, whether it is refundable and the source page. Leave missing amounts blank. End with the questions I should ask before signing.
For buyers: collect Loan Estimates from more than one lender and ask AI to compare the same loan type, down payment and lock period. Focus on the total monthly payment, cash to close, origination charges, points, lender credits and five-year borrowing cost. The CFPB’s Loan Estimate guide explains each line.
For sellers: ask AI to produce low, middle and high net-proceeds scenarios using the mortgage payoff, brokerage compensation, buyer concessions, repair credits, taxes, title charges and other expected costs. Have your agent or settlement professional confirm the result.
Build a price case from evidence
The listing price is an asking price. A comparable market analysis is an agent’s pricing opinion. An appraisal is an independent value opinion, usually for the lender. The assessed value is used for property taxes. An automated estimate from a real estate site is one more data point. These numbers can disagree without any of them being fraudulent.
Ask your agent for recent closed sales from the MLS, current competing listings, pending sales when details are available, days on market, price changes, seller concessions and the local sale-to-list-price ratio. Add county tax and recorder records, permits, the seller disclosure, HOA documents, inspection findings, insurance quotes and official FEMA flood information. Public listing sites are convenient, though their asking prices and histories can be incomplete.
Prompt: Using only the property information, local market report and comparable sales I provide, compare the subject home with each comp. Separate verified facts, reasonable estimates and missing information. Model three price scenarios and explain the evidence for each. Include possible appraisal, repair, insurance and resale risks. Do not choose the final offer or listing price for me.
Ask your agent questions that force useful answers
Buyers should ask: Who do you represent? What services and compensation are in our agreement? How can I end the agreement? Which closed sales support this price? What contingencies and deadlines are customary here? What would create an appraisal gap? What do we know about taxes, insurance, flood risk, the HOA and major repairs? How much cash would I need under several offer structures?
Sellers should ask: Which services are included in the listing agreement? Which comps support the suggested price? How will the property be marketed? What compensation or concessions should I consider and why? What would I net at several sale prices? When will we review price and strategy? How will you compare financing strength, contingencies, concessions, timing and other terms when offers arrive?
Prompt: I am interviewing a [buyer’s/listing] agent. Based on my priorities, budget, timing and the agreement provided, create 12 specific questions. For each question, explain what a complete answer should contain and which document or local data should support it.
Turn offers into decision briefs
A buyer’s brief can show the proposed price, financing, earnest money, contingencies, deadlines, requested seller costs, estimated monthly payment, cash to close, appraisal-gap exposure and repair reserve. AI can model several structures so you can see what changes when the price, rate, credit or down payment moves.
A seller’s comparison can place every offer in one table: price, financing, proof of funds or preapproval, contingencies, appraisal terms, requested concessions, earnest money, closing date, possession, estimated proceeds and the main risk to closing. The highest price may carry weaker terms or a lower likely net.
Prompt: Turn these offers into a side-by-side decision table. Calculate estimated proceeds or cash required using only the numbers supplied. Rank the major financial and closing risks, show every assumption, and create a question list for my agent. Preserve the original contract wording and do not draft replacement legal language.
Use AI for the last paperwork check
Buyers generally receive a Closing Disclosure at least three business days before closing. Ask AI to compare it line by line with the most recent Loan Estimate. Look for changes to the rate, points, credits, loan costs, taxes, insurance, prepaids, monthly payment and cash to close. The CFPB’s Closing Disclosure guide can help you understand the form. Sellers can compare the settlement statement with the most recent net sheet and purchase agreement.
Prompt: Compare these two versions line by line. List every changed number, term and deadline with the page where it appears. Separate changes that improve my position, worsen it or need an explanation. Create a short list for my lender, agent or settlement professional.
Keep the guardrails simple
Remove Social Security numbers, bank and account numbers, tax returns, signatures and wire instructions before uploading documents. Verify any wiring instruction by calling a known number for the title or settlement company. A last-minute email changing the destination deserves a phone call.
Use AI to expose questions and organize evidence. Use your agent, lender, inspector, appraiser, settlement professional and attorney for the work their licenses and responsibilities cover. Contracts, disclosure duties, agency relationships, attorney involvement, taxes, commissions and closing customs vary by state and sometimes by locality.
AI Tool Spotlight
Turn a plain-English instruction into a browser task that runs across tabs.
Composite is a Mac app that carries out browser work from a written instruction. It can open pages, click, type and move information between tabs while you continue using the browser. The company says the beta is free and does not require a credit card.
Where it could help a business
Research: visit a defined set of pages and collect the same details from each one.
Operations: move routine information between browser-based systems when a direct integration is unavailable.
Monitoring: check a short list of dashboards or pages and organize the results for review.
Try this first
Start with a low-risk task that has three to five clear steps and a result you can verify quickly. Watch the first few runs closely. Because the app can click and type in logged-in browser sessions, review its permissions and keep sensitive data, purchases, account changes and other hard-to-reverse actions out of the test.
Source: Composite
Headlines
And that was only the Top 5
Here are 25 more stories, grouped by topic so you can jump to the parts you care about.
Work, business and markets
Listening sessions will inform a state commission that is expected to propose labor-data and workforce policies by year-end. Women hold 84% of New York’s administrative, clerical and customer-service jobs.
The inquiry covers market structure, manipulation, customer protection and perpetual futures tied to computing capacity.
The app can see a shared window, take dictation and work with Instagram, Facebook, advertising campaigns and Google Workspace. Users should check how connected data may be used for training and ad targeting.
The assistant can produce transcripts, action items and email drafts. Calendly says participants are notified when it joins a meeting.
Microsoft 365 activity traces showed 21.2% more productivity-app actions and 7.1% more communication among heavy users. The authors say the pattern may also point to a shift toward more individual, document-centered work.
Health and education
Its HIVE system grows, treats and monitors 20 kinds of tissue. The goal is to give models controlled evidence about cause and effect.
A district committee chose the guided tutor after testing it against attempts to get direct answers or push the system off course.
Study notebooks now support graphs and images, Deep Research can continue in the background, and Lens can explain photographed worksheets. Eligible U.S. students can receive Google AI Pro free for one year.
A McGraw Hill survey of 1,300 educators across 19 countries found most teachers use AI and 60% believe their students do too.
EAB found 57% use AI in the college search and 34% have discovered schools through it. The findings raise the stakes for accurate, accessible university information.
Products, people and trust
The assistant adds conversational search, recommendations and smart-home controls for Prime members and nonmembers.
A Pew measure cited in the analysis found 52% of Americans more concerned than excited about AI, up from 37% in 2021. Jobs, data centers, product quality and creative rights all feed the unease.
Google says its coaching features remain a work in progress. Privacy concerns, unclear value and limited wearable adoption still stand between demonstrations and daily use.
Researchers who had access to advanced cyber capabilities were told to verify their identities again. The episode shows how access controls can interrupt legitimate security work.
The financing brings its total to $455 million and will help develop smaller autonomous vehicles, including work connected to DoorDash.
Infrastructure and builders
The CS-4 combines three wafer-sized chips and uses about half as many components as its predecessor. Shipments are planned for the third quarter.
Marvell also granted Google a warrant that could become a stake worth up to $12.2 billion. The agreement covers technology used with Google’s in-house AI chips.
The reactor uses molten sodium to hold heat and release it as needed. The company expects to announce its first data-center project this year.
The startup says light travels about 50% faster through its air-filled cable, which could help connect large computing campuses with less delay.
Origin can work with existing GitHub repositories. Cursor says future features will let agents manage more of the development process.
Policy, culture and research
The electronic-music store says it will permit tools that support human creation. Its survey found users strongly prefer music with a human creator behind it.
Speculation around model names, dates and capabilities can spread far ahead of verified information. Treat screenshots and anonymous launch claims as leads to check, not facts to repeat.
The company expects to use the money for data centers, its AI platform and other corporate needs.
The experimental system ranked possible diagnoses for a later visit and retrieved evidence physicians judged more useful than comparison research systems. Its accuracy remains far from clinical deployment.
Local governments are taking different approaches to power demand, water use, noise and land planning. The patchwork shows how much AI infrastructure policy is being set close to home.
mAIn Street gives nontechnical readers useful AI news they can put to work.

Read All mAIn Street Back Issues Here


mAIn Street

Check out the resources I offer below and sign up for my new newsletter!

Read more from mAIn Street
Calculator, charts and planning materials arranged on a desk

AI is moving deeper into medical care, professional work, personal finance and the infrastructure debate. Wednesday, August 26, 2026 On September 9, I’m hosting ChatGPT for Marketing: Using AI So It Doesn’t Look Like AI, a 90-minute workshop for marketers and business professionals who are tired of generic copy and look-alike visuals. We’ll work with writing, images, fonts and web pages, then use real examples to show how references, clear instructions, repeated edits and human judgment...

Colleagues discussing work around a laptop during a meeting

AI pay packages, school rollout caution and supervised clinical agents lead Tuesday’s AI news. Tuesday, August 25, 2026 On September 9, I’m hosting ChatGPT for Marketing: Using AI So It Doesn’t Look Like AI, a 90-minute workshop for marketers and business professionals who are tired of generic copy and look-alike visuals. We’ll work with writing, images, fonts and web pages, then use real examples to show how references, clear instructions, repeated edits and human judgment produce work that...

A professional mapping a workflow with sticky notes

Nvidia server price hikes, Kaiser triage concerns and OpenAI’s safety-law reversal lead Monday’s AI news. Monday, August 24, 2026 On September 9, I’m hosting ChatGPT for Marketing: Using AI So It Doesn’t Look Like AI, a 90-minute workshop for marketers and business professionals who are tired of generic copy and look-alike visuals. We’ll work with writing, images, fonts and web pages, then use real examples to show how references, clear instructions, repeated edits and human judgment produce...