mAIn Street reaches edition 300 as Nvidia reportedly buys Hugging Face, SK hynix breaks ground in Indiana, and Best Buy sees AI device upgrades.
 |
Friday, August 28, 2026 · Edition 300 |
|
300th edition
Three hundred editions of making AI news useful.
Thank you for reading, sharing and putting this work to use.
|
|
|
|
Top 5
What matters today
|
| 01 |
The deal would give Nvidia the open-source model and dataset hub used by developers across the industry. Reuters said the price would rank among Nvidia’s largest acquisitions, even though Hugging Face generates about $150 million in annualized revenue.
|
|
| 02 |
The West Lafayette site will package high-bandwidth memory and house an AI semiconductor research hub. Cleanroom operations are due in late 2028, bringing a critical part of AI-chip production closer to U.S. customers.
|
|
| 03 |
The retailer cited resilient demand for replacements and newer categories including AI glasses and 3D printers. Shares still fell as investors looked past a tariff refund that helped lift quarterly profit.
|
|
| 04 |
The program is live at 10 hospitals, with eight more scheduled in September. It uses Andor Health’s platform and clinical team to help patients complete follow-up work after discharge.
|
|
| 05 |
The expansion covers more than 100,000 educators and staff. OpenAI says it now works with more than 100 K–12 organizations across 30 states and will keep the teacher product free for verified U.S. educators through June 2028.
|
|
|
|
|
|
AI Workflows | Friday
I used ChatGPT Sites to turn my home reports into a maintenance plan.
My inspection and appraisal reports became a private tracker for priorities, timing, costs and completed work.
|
|
For mAIn Street’s 300th edition, I wanted to share an AI workflow I’m using in real life. I’m currently buying a home. That has given me two long documents full of facts I will need after closing: an inspection report and an appraisal report.
OpenAI showed this week how ChatGPT can turn dense information into an interactive visual. I tested that idea with my own reports. I uploaded both and asked ChatGPT Sites to build a private home-maintenance tracker.
It organized the work by urgency and timing, added estimated cost ranges, compared doing the job myself with hiring a professional, and gave me a place to mark each item complete. The property, reports, dashboard and actual figures will stay private. The useful part is the method, and that is what I’ll show you today.
The reports become a living work list
An inspection report follows the inspector through the house. An appraisal explains the property’s condition and value. Both contain useful information, though neither is written as a schedule for next month, next year or several years from now.
I asked ChatGPT to read both reports together, find overlapping concerns and preserve the source for every item. The Site organized the results into useful time periods: before closing or moving in, the first 90 days, the first year, the next two to five years, and items that simply need watching.
Each entry includes the reason it matters, suggested timing, a rough cost range, and a comparison between doing the work myself and hiring someone. It identifies the kind of professional who may be needed and provides space for quotes, notes, actual costs and completion dates.
Give Sites the whole job
Sites can use attached or linked documents to build a private app with saved checklist data. It is currently in public beta on eligible paid plans, and workspace settings may affect access.
Copy-ready Sites assignment
Use @Sites to build a private, mobile-friendly home-maintenance tracker from my attached inspection and appraisal reports.
Read both documents completely. Find every current defect, recommended repair, maintenance item, safety concern, age or useful-life clue, and issue requiring specialist review. Merge duplicate findings while keeping every source reference.
For each item, show the home system or area, plain-language concern, why it matters, source document and page, priority, suggested time window, DIY cost range, professional cost range, pricing source and date, tools or skill needed, safety limits, recommended trade, status, notes, quote amount, actual cost, completion date and next due date when the work repeats.
Organize the tracker into concerns to settle before closing or moving in, the first 90 days, the first year, the next two to five years, and a watchlist. Add filters for priority, timeframe, home area, DIY or professional, and status. Use saved storage so updates remain after I close the Site.
Remove the street address, owner names, parcel numbers, lender information and report-order numbers from the visible interface. Ask only for my general metro area if local pricing would improve the estimates. Label all costs as planning ranges until I receive a real quote.
Never invent a defect, deadline, price, code requirement or recommendation. Flag missing details and conflicts between reports. Show me a source-check table before finalizing the schedule. Test the filters, saved status and main workflow on mobile and desktop. Keep the project private, save a version without deploying it, and never publish, share, change access or contact a contractor without my approval.
Prove that you learned the skill
Check at least ten Site entries against the original pages. Confirm that the priority makes sense, the cost is labeled as an estimate and no private information appears. Change two statuses, close the Site and make sure those changes remain.
Then start a fresh build with a different pair of documents. A vehicle inspection and owner’s manual could become a service tracker. An equipment report and warranty could become a workplace maintenance plan. You have learned the skill when you can explain the structure, catch a source error and create a second useful tool with new rules.
|
For managers: Use the same method with facility audits, equipment inspections, warranties or safety logs. Keep the source documents private, require page references and have the person responsible for the property approve priorities, costs and action.
|
Try it with two documents you already own
Choose one real decision buried inside two dense documents. Build the private tracker, verify ten entries and make one improvement based on what the first version got wrong.
|
|
mAIn Street AI Workflows gives workers practical ways to use current AI tools on real jobs.
|
|
|
AI Tool Spotlight
Branch one AI conversation into a map of related questions.
|
|
Wondering Canvas turns an AI chat into a visual workspace. Start with one question, then branch any term, sentence or follow-up into its own thread while the original conversation stays visible.
It is useful when research keeps opening new paths: learning a topic, comparing choices or planning a complicated project. The canvas keeps each detour connected, and answers can include interactive diagrams instead of one long scroll.
|
|
|
Headlines
And that was only the Top 5
|
Money, chips & infrastructure |
|
The deal would give Nvidia the open-source model and dataset hub used by developers across the industry. Reuters said the price would rank among Nvidia’s largest acquisitions, even though Hugging Face generates about $150 million in annualized revenue.
|
|
The West Lafayette site will package high-bandwidth memory and house an AI semiconductor research hub. Cleanroom operations are due in late 2028.
|
|
Nvidia projected 70% revenue growth next fiscal year, well above the 44% expected by Wall Street. Its shares rose as much as 6.8%, while Intel, Micron, Broadcom and AI-cloud providers also gained.
|
|
S&P Global Ratings expects Alphabet, Amazon, Microsoft, Meta, Oracle and SpaceX to generate negative free operating cash flow in 2026 and 2027 as debt, leases and other financing support the buildout.
|
|
The retailer cited resilient demand for replacements and newer categories including AI glasses and 3D printers. Shares still fell as investors looked past a tariff refund that helped lift quarterly profit.
|
|
|
Work & business operations |
|
Insurers are deciding how an agent’s unauthorized actions fit existing coverage and who bears the risk when software operates within its assigned permissions. Most are adapting current cyber policies instead of creating a separate category.
|
|
Salesforce in Claude can prepare meetings, review deal health and update pipelines while routing actions through company rules. Select customers have the pilot now, with an open beta expected in September.
|
|
Unanimous AI says its (Co)agents follow text, voice or video conversations and surface relevant information while the discussion is happening. Teams will need clear rules for what the agent can hear, retrieve and interrupt.
|
|
Zoph announced the move Thursday. The departure gives Google DeepMind an experienced model researcher and removes a founding executive from one of the field’s closely watched startups.
|
|
GAPVelocity reviewed 1,105 organizations that asked about replacing old applications over the past year. The finding shows how much AI-assisted software work still begins with decades-old systems.
|
|
|
Health, schools & public service |
|
The program is live at 10 hospitals, with eight more scheduled in September. It uses Andor Health’s platform and clinical team to help patients complete follow-up work after discharge.
|
|
The expansion covers more than 100,000 educators and staff. A 16-state privacy agreement gives participating districts a shared framework for reviewing the product against student-data requirements.
|
|
The system aggregates, normalizes and reconciles records from a patient’s care journey for clinical workflows, monitoring and analytics. The value depends on preserving the source trail and surfacing conflicts for review.
|
|
Elevate Clinic, which operates in six Texas cities, will use Neurophet software to analyze MRI scans for traumatic brain injury, Alzheimer’s disease and other neurological conditions.
|
|
The firm estimates an 11-percentage-point margin drop by 2035 for a median system as demographics and costs raise pressure. Its report expects more personalized care supported by clinical AI and new capital.
|
|
|
Trust, security & policy |
|
Public Citizen found that crypto, online betting and AI companies together supplied more than half of the total. AI ranked third among corporate-donor industries after crypto and betting.
|
|
Politico reported that the administration is weighing the move. Another levy would add a new cost question for data centers, electronics makers and chip buyers already tracking supply constraints and domestic-manufacturing incentives.
|
|
The Synthetic Insider Threat Matrix covers unauthorized data access, autonomous exfiltration, privilege misuse and shadow AI. It gives security, legal and HR teams a shared vocabulary for investigating agent behavior.
|
|
The Hunter Agent looks for broken access control, business-logic problems and authentication or session failures. It is generally available in SonarQube Cloud and feeds confirmed findings into the existing review workflow.
|
|
More than 100 participants from utilities, government and the military worked through the same simulated attack. The exercise paired a live cyber range with executive decision-making and included testing for human and AI defenders.
|
|
|
Around the world |
|
The company says 35% of classified ChatGPT messages from Brazilian individual accounts were work-related in June, compared with 30% globally. The local operation will focus on developers, businesses and public-interest partners.
|
|
Huawei already offers compound-screening tools on its Ascend and Kunpeng chips and is exploring more work from drug manufacturing through clinical use. Its current projects are mainly with domestic pharmaceutical companies.
|
|
The 2026 lineup uses LG’s third-generation α11 processor to adjust picture and sound and tailor content. LG Shield is meant to protect the personal data behind those features.
|
|
Geotab’s interns created dashboards, inventory tools and automated workflows while learning to review data and keep people involved. The company says it has hosted more than 1,000 interns over five years.
|
|
In a randomized Bocconi University experiment, ChatGPT improved the quality and coherence of student work. Causal-reasoning training increased the variety and originality of ideas, and students who received both showed both gains.
|
|
|
|
mAIn Street gives nontechnical readers useful AI news they can put to work.
mAIn Street #300 · Friday, August 28, 2026
|
|
|
|
|
