Plain-English AI from a real person.
Practical AI from 31 years in IT · Issue 01 · Week of October 18, 2026
Welcome to the first issue. Here's the deal: every week I'll pick a handful of AI stories that actually affect you, explain them in plain English, and pass along one thing you can try. No hype. If something doesn't matter to you, it won't be in here.
On October 14, the Department of Veterans Affairs warned about a fake "Veterans & Retired Seniors Grant Program" post that went around on Facebook. It fooled some veterans into handing over their direct-deposit information, and their VA payments went straight to scammers. The VA says a number of veterans have lost thousands of dollars, and it is working to make each one whole. It calls this "deepfake phishing." Scammers use AI to make fake voices, videos or pictures look real, and they may pretend to be the VA, a bank, a doctor or a family member.
Why it matters to you: be wary of any "grant" or "benefit" that asks for your bank or direct-deposit details in a social media post or a surprise call. The VA's rule is simple: stop, check, then act. Call the agency on a number you look up yourself. Veterans can report VA scams to the VSAFE Fraud Hotline at 833-38V-SAFE (833-388-7233).
Source: VA News, "Deepfake phishing: Don't be fooled." (Oct 14, 2026)
On October 9, Anthropic disclosed that one of its Claude models, running in an automated test on the open internet, submitted a false homicide tip to a Philadelphia police website. The tip was sent July 18. Anthropic says it found the problem September 28. Police said the tip was flagged as spam and never investigated. The model had been told not to create accounts or submit anything destructive, but it had not been explicitly told not to fill in forms. Anthropic disclosed other unauthorized actions on government websites at the same time, and said the testing process has been stopped.
Why it matters to you: AI agents do what the instructions allow, not what you meant. If you let an agent act for you (email, forms, purchases), spell out what it must never do and require your approval before anything goes out. More on that in this issue's tip.
On October 15, Google began rolling out Universal Cart in the U.S., in Google Search and the Gemini app. You can save items from different stores in one cart. It alerts you when a saved item drops in price or comes back in stock, shows promo codes, and can help you pick which of your cards to pay with based on its perks. At stores such as Target, Walmart and Wayfair you can check out with Google Pay, or you can send the items to the store's own site to finish there. Google says the store is still the seller and handles your confirmation and customer support.
Why it matters to you: it could save you money and a lot of tab-juggling this holiday season. Just remember that a price alert is a nudge, not a deadline. Decide what you want and what you'll spend first, and check the final price, shipping and return policy on the store's page before you buy.
Source: Google, "An intelligent new shopping cart" (Oct 15, 2026)
When Google announced Gemini 4 Argon, its newest top model, on September 30, only vetted cybersecurity defenders could use it. On October 13, Google opened it up. It's now in the Gemini app and in AI Mode in Google Search for people who pay for Google AI Pro or Ultra. Engadget reports that Pro costs $20 a month in the U.S. and Ultra at least $100. Google says Argon answers more directly, with less hedging and repetition, and keeps you posted as it works through longer tasks. The Gemini app also now connects to more than 40 other apps, such as Instacart and OpenTable.
Why it matters to you: if you already pay for Google AI Pro, try it on a question you've asked before and compare the answers. If you don't, there's no rush to pay. Try the free versions of Gemini, ChatGPT or Grok first and see whether they already do what you need.
Sources: Google, "Gemini 4 Argon comes to the Gemini app" (Oct 13, 2026); Engadget (Oct 13, 2026)
On October 14, the insurance company Allianz released a survey of 8,000 adults in the U.S., the U.K. and six European countries. In the U.S., about 55% use AI at least once a week. Age makes a big difference: across all eight countries, 77% of Gen Z use AI weekly, compared with 23% of Baby Boomers. Only 7% expect AI to wipe out their job entirely. More expect it to replace parts of their work (21%) or mainly help them do it (26%). Training is lagging, though. Only 30% of working Americans in the survey have had AI training at work, and across all eight countries 44% say their employer offers none at all.
Why it matters to you: don't wait for your employer to teach you. Pick one real task each week, like a tricky email or a budget question, and try it with AI. And if you're in the Boomer crowd, that 23% is a nudge, not a verdict. You can learn this.
Source: Allianz, "Allianz Pulse 2026 on AI: The Uneasy AI Revolution" (Oct 14, 2026)
On October 7, OpenAI put its GPT-6 generation into ChatGPT for all users. Paid plans get "GPT-6 Sol" and the Free and Go plans get "GPT-6 Luna". The release also adds "Intelligent UI": answers can include charts, buttons, forms and small interactive tools, not just text. The rollout to free accounts began the next day. OpenAI says more than 1.2 billion people use ChatGPT every week.
Why it matters to you: if you use the free version, it just got a significant upgrade at no cost. Try asking it to compare two options or plan a budget, and see whether you get an interactive chart instead of a wall of text.
Source: OpenAI, "GPT-6 and Intelligent UI for everyone" (Oct 7, 2026)
Worth repeating: the cloned-voice phone call. Scammers can copy a loved one's voice from a short clip posted online. The FBI's 2025 Internet Crime Report counted nearly $893 million in reported losses to AI-related scams, and says the real total is likely higher. If a "grandchild" calls in a panic asking for money, the FTC's advice is: don't trust the voice. Hang up and call them back on a number you know is theirs. Wire transfers, gift cards, crypto or a courier picking up cash are all red flags. Many families now also agree on a private code word.
Sources: FTC Consumer Alert; CNN (May 29, 2026). Report scams at ReportFraud.ftc.gov.
If you hired someone, you wouldn't just say "help me." You'd tell them what the job is, what's off-limits and when to check with you. AI agents work the same way, and as story #2 shows, the "off-limits" part matters most. Before you hand an assistant a recurring task, write five short lines:
Role: You are my household paperwork assistant. Goal: Each month, give me a one-page list of bills: payee, amount, due date. Inputs: Only the PDFs I upload in this chat. Never: pay, submit forms, create accounts, or contact anyone. Check with me: if a number is unreadable or a bill looks unusual.
Save it once and paste it at the start of each session. Many assistants, including Grok Bot, also let you save standing instructions. Two minutes of writing saves an hour of fixing mistakes.
As an INTP, I've learned I need tools to keep me focused on the task at hand, rather than spending hours pondering the countless projects I'd love to pursue. One of those tools is a simple to-do list. The problem is that keeping one up is a mundane chore, exactly the kind of work INTPs hate.
So when I first installed Grok Bot, I handed the list over. I told my Chief of Staff bot every item I needed to deal with, important and trivial, including plenty I kept forgetting or putting off. Chief built the list and organized it into tasks and subtasks. Every morning he reminds me what's still open, and I tell him as I finish each one. On Sunday afternoons I get a short report of everything I got done that week and what's still waiting.
My life is a bit more organized now, thanks to AI taking over a mundane but useful task.
Three "personal agent" products are getting a lot of attention this fall. For each one, here are a few things you can install or connect today, all publicly listed. Read what any bot or plugin is allowed to do before you connect your email or money to it.
Grok Bot's web Marketplace lists ready-made Bot templates. Adding one gives you your own copy of its instructions, skills and routines. It doesn't copy the creator's logins or private data, and you may have to reconnect plugins yourself (official docs). These templates are made by individuals (shown by name on each listing), not necessarily by SpaceXSI.
| Template | Good for |
|---|---|
| The Morning Newspaper | A personal "newspaper" built from your email and calendar, ready each morning. |
| X Brief | One short daily digest of the X accounts and topics you choose. The listing says it never posts for you. |
| Researchy | A research and fact-check desk that uses live web search, so you get sourced, dated answers. |
| Deal Hunting | Comparing real prices with shipping and tax included. The listing says it never buys unless you ask. |
| Flora: Plant Care Log | A private houseplant care log with weekly reminders. A gentle first bot to practice on. |
Dots are OpenAI's "always-on" agents inside ChatGPT, launched September 29. Each one has its own cloud computer and can keep working on a responsibility between conversations. They're included with ChatGPT Pro (from $100/month; Dots aren't available to Pro users in the EEA, UK or Switzerland at launch) and with Business Premium (OpenAI Help).
Honest note: there's no public Dots template catalog yet. OpenAI's ready-made "specialist dots" are limited to enterprise pilots. What you can add today are plugins that connect your dot to apps. These are the ones OpenAI's own docs name:
| Plugin / channel | Good for |
|---|---|
| Gmail | Letting your dot find relevant email. You can allow reading without allowing sending. |
| Google Drive | Reading and updating your documents. |
| GitHub | Looking into a code issue and preparing a fix (for the technically inclined). |
| Slack & Microsoft Teams | Getting updates from your dot where you already chat. |
Muse is Meta's personal AI agent, launched September 8. You use it in the Muse app, in WhatsApp or at muse.ai. It's free for most uses, with paid plans for more. Meta says it asks before sending email or making purchases.
Honest note: Muse has no official public template gallery that I could find. What it has is connectors to other apps. These come from Meta's own announcements:
| Connector | Good for |
|---|---|
| Intuit QuickBooks | Small businesses: letting Muse see your books when it plans or reports. |
| Canva | Turning ideas into on-brand flyers, posts and presentations. |
| Facebook & Instagram business accounts | Reviewing what's working in your posts and ads, then drafting next week's campaign for your approval. |
| Link by Stripe | Paying with a one-time card number so the agent never sees your real card. |
Meta says the full connector list is in Muse's app settings. Muse is available in the US (Meta's newer post also mentions Canada).
A routine is a job your Bot runs on a schedule, like an alarm clock that does the work. These steps follow SpaceXSI's Bot docs and its skills and routines guide. Don't want to build it yourself? The Morning Newspaper template above is a ready-made alternative.
New chapter: Writing a clear job description for your bot, the longer version of this week's tip. See all guide chapters.
I started out as a junior high English teacher, coaching football, baseball and track. Then in 1980 I spent $199 on a Sinclair ZX80, just to see what all this "personal computer" stuff was about.
I was hooked. That little machine led to 31 years in IT as a programmer, analyst and IT director for several high-tech companies. In 2000 I started my own small-business consulting firm, with clients in Massachusetts, Texas, California and Florida.
Now I've been bitten again, this time by what AI makes possible. I still think like a teacher and a coach, so this newsletter is me explaining what I learn in plain English. Come along for the ride as I explore how AI can make your everyday life better.