AI is the biggest opportunity ordinary people have had in a century. Here are the facts on what it can do for you, where to learn it free, and real people already using it, so you can decide for yourself.
Benefits of AI
A job provides income, purpose, and community. Losing one can turn a family’s life upside down. That concern is real. The goal, though, should be better lives — not keeping every existing task in human hands. That is the argument in my unpublished draft, “Protect People, Not Just Jobs,” written after a discussion I heard on Brighter with Herbert with Cern Basher. I am not repeating their argument. This is my own reflection: protect people through the transition, and judge this change by how people live.
Abundance is a possibility, not a promise. The benefits could concentrate in a few hands. Systems can make mistakes, privacy can be violated, and powerful tools can be misused. Those risks deserve safeguards. The question is whether we can face the risks and still make the benefits available to more people — including schools, small businesses, older adults, and people with disabilities.
Health
Some of this is already here, not a forecast. Each item below is documented, with sources, on this site:
- AI has been used to find a new antibiotic candidate, match cancer patients to treatments, and design a drug that later reached a human trial. See The Off-Ramp to Abundance (Health and medicine), with links to MIT News, Tempus, and a 2025 Nature Medicine paper.
- Researchers are using scans, blood markers, and speech to try to catch Alzheimer’s earlier. Source notes are on that same Off-Ramp page, including the National Institute on Aging.
- The 2024 Nobel Prize in Chemistry went partly to the creators of AlphaFold, an AI system that predicted the structure of hundreds of millions of proteins and is used in labs around the world. See the Off-Ramp page and Instances of real people on the off-ramp.
- There are already named family cases in which AI helped doctors or parents find a lead sooner — a delayed diagnosis, a rare-disease clue, a newborn matched to a drug already on the shelf. Those stories, with sources, are on the Instances page. They are markers on a long road, not a claim that the road is finished.
Time saved
- A Harvard randomized trial found that physics students using an AI tutor learned more, in less time, than in an active-learning class. It was a short study on two topics, which that write-up notes as a caveat. Details and the Scientific Reports link are on The Off-Ramp to Abundance and the Instances page.
- Other measured tutoring examples — including a WhatsApp math tutor in Ghana and a tool that coaches human tutors — are collected on the Instances page, with the study links.
- On the practical side, a capable assistant can take first drafts of notes, research, and paperwork off a person’s plate. The person still checks the work and still presses send. That is the “small-business sidekick” already described on AI for a Better World and AI Opportunities.
Cost of living and access
- Some of the learning tools that used to require a specialist are now inexpensive or free to try. Several companies already offer a free plan. See Several Free AI Plans Available.
- Cheaper services and faster research can lower costs over time, but that is not automatic. If only a few people control the tool, the savings may not reach ordinary households. Access — being able to use the tool, afford it, and know it exists — belongs in the discussion from the beginning.
Abundance
The off-ramp this site talks about is a choice: learn what is real, use the tool, and keep your judgment — or let fear be the whole conversation. Fear is not a plan. Neither is blind optimism. The longer preview, with each item labeled Today, Near, or Prediction, is on The Off-Ramp to Abundance. A view of people already building is on AI for a Better World. Energy and infrastructure context is on Data Centers and Power and Nuclear power.
Learn for Free
You do not need a computer science degree, and you are not limited to one company. Plans and limits change, so check each site for yourself. This list is ranked for a beginner: first a tool you can try today, then short free courses. It is a door, not a ranking of which lab is “best.”
A simple way to start, already on this site: pick any free assistant below and ask it, “Build me a beginner curriculum to teach me how to use you well. One short lesson a day. I will verify everything myself.” Then do verify. Fluent answers are not the same thing as sources. See How Can You Determine Truth?
1. Free assistants you can try today
These official sites loaded or are the same entry points already listed on this site. Free tiers have usage limits. Paid features are separate.
- ChatGPT (OpenAI) — free plan
- Claude (Anthropic) — free plan
- Gemini (Google) — free with a Google account
- Grok (xAI) — free access; also available via X
- Microsoft Copilot — free in Edge, Bing, and apps
- Perplexity — free searches with citations
- Meta AI — free where Meta offers it, with daily limits
The fuller roster, including tools that are paid or meant for developers, is on Several Free AI Plans Available and AI Assistants Available Now.
2. Claude Academy — free courses from Anthropic
Claude Academy describes itself as free courses, tutorials, and use cases from Anthropic’s education team. A practical first course is Claude 101 (about 2.5 hours). The course page says you need a Claude account on any plan, including the free plan, and no prior experience. You can browse without signing in; a free account saves progress and can earn a completion badge. This site already points to Claude Academy from the AI Opportunities and free-plans pages.
3. Elements of AI — University of Helsinki and MinnaLearn
Elements of AI is a free, self-paced introduction for non-experts. No programming or advanced math is required for Introduction to AI. A follow-on course, Building AI, is also free but recommends some basic Python. The University of Helsinki lists the current Introduction MOOC as free of charge, running 1 September 2026 through 31 August 2027. You do not need a university account to study. Optional ECTS credit, where offered, is a separate registration step.
4. Microsoft Learn — free beginner modules
Microsoft’s training library is free to read. A Microsoft account is needed if you want to save progress or record a passing score.
- Get started with AI — basic terms, writing prompts, summarizing a document, and creating an image. No prerequisites. Beginner level.
- Work smarter with AI — a short beginner path on using AI for workplace tasks and using it responsibly. It assumes some familiarity with everyday Microsoft apps such as Word, Excel, Outlook, or PowerPoint.
A wider menu is on the Microsoft AI learning hub.
5. Khan Academy — AI for Education (free course)
Khan Academy’s Help Center, updated February 19, 2026, describes a free course called AI for Education. Unit 1 covers what generative AI is, how it works, and ethical questions. Later units are aimed more at teachers (classroom practice and lesson plans). The course page itself was behind a bot check when we fetched it on November 17, 2026; the official Help Center article is here. Note: Khanmigo, Khan Academy’s AI tutor for families, is a paid subscription for parents and learners. Teacher tools are free only for eligible educators. The regular Khan Academy lesson library remains free.
6. Google — a short introduction, with a check
Introduction to Generative AI on Google Skills is a 45-minute introductory course (what generative AI is and how it differs from older machine-learning tools). A Google account is required to enroll. Google Skills also sells a paid Pro plan for hands-on labs. This short course has been offered at no cost; confirm the enroll button still says so before you start, because access terms can change.
Do not confuse that short course with Google AI Essentials. As of November 17, 2026, Google’s own page says that course is not free in the U.S. and Canada: $49 a month after a 7-day trial. It is not on this list as a free offering.
Courses we checked and are not listing as free. Andrew Ng’s AI for Everyone (DeepLearning.AI, on Coursera) is a well-known beginner course with no coding required. DeepLearning.AI’s page lists it at $49 for 180 days of certificate eligibility, with financial aid available. Coursera’s current preview model often unlocks only the first module for free. We did not treat it as a free course. “Career Essentials in Generative AI” from Microsoft and LinkedIn sits on LinkedIn Learning; we could not confirm that it is free without a LinkedIn Learning subscription, an employer, a school, or a library card, so it is not listed here either.
People Using It for Business
The useful part of these stories is the work a person actually handed to an AI helper — research, email, drafts, coding, customer replies — while keeping approval in human hands. Headline income figures are the weak part. They are labeled as the person’s own claim, even when a newspaper repeated them. Several of these people already had an audience, a tech job, or investor money. That does not erase the example. It should be said plainly so no one is sold a miracle.
Billy Howell — The Arlington Bagel
Billy Howell publishes The Arlington Bagel, a weekly newsletter about events and local businesses in Arlington, Virginia. Issues were still going out as of November 5, 2026. In an August 21, 2026, interview with Greg Isenberg, he said he used to have an intern (his cousin) research, assemble, edit, and watch the inbox, and that a team of AI assistants now does that work.
What they actually did
- He keeps a separate assistant account for each project so two businesses do not share one memory.
- He started with a “chief of staff” assistant, connected it to the tools he already used (Notion, Slack, and Gmail), and asked it to audit the newsletter and recommend the first three helpers. The audit named a research assistant, a sales assistant, and a newsletter-platform specialist.
- His four-week pattern: week one, build the team; week two, just run it, without adding new helpers; week three, add or drop helpers based on real gaps (for example, an inbox assistant); week four, put the work on a routine so it happens without being asked each time.
- Each assistant sends the chief of staff a short report: what shipped, what is stuck, and what needs a human. Howell reviews and unblocks.
- For the Thursday newsletter, a research assistant checks a source list twice a week and fills cards with restaurant and event items. A separate step writes the short blurbs in a fixed format. The newsletter platform sends the issue on Thursday.
- A sales assistant watches the inbox, prices ad slots, and builds a pitch sheet. Howell said it caught a sponsorship inquiry he had missed. He still described himself as the person who sends outbound messages and gives approval.
Revenue and reach: the site and the interview say “6,000+” readers. That is his own claim; we did not see a dashboard or an audit. He described a local bagel-shop sponsor found by the sales assistant. We did not find a public dollar figure for newsletter revenue, and we are not inventing one. Howell also tests tools for The Rundown, so this is a real local newsletter, but not a person with no tech background.
Sources: The Arlington Bagel · Startup Ideas Podcast interview
Troy Johnston — StackEasy
Troy Johnston, of Orlando, Florida, was described in the July 29, 2026, Wall Street Journal as using AI to code an app that helps people get the most out of credit-card benefits, and as running the business alone. His site identifies the product as StackEasy: due dates, 0% APR windows, utilization, and rewards, at $25 a month after a 14-day trial.
What they actually did
- The Journal’s account is short but specific: he used AI to write the code for the app, then ran the company with no employees.
- His own site says he built it out of managing 30-plus cards himself, and that he writes the editorial and corrects mistakes in public.
- The site also says he previously sold three companies (Foxbrim, Seller.Tools, and Neon Pulse). This is an experienced operator, not a first-time beginner.
Revenue: the Journal reported that the company “makes around $3,000 a month in profit, with no employees, and is continuing to grow.” That is a founder figure reported by the Journal — his own claim, not an audit. His public about page does not repeat the profit number. The modest size is the useful part: a working one-person product, not a “$100K a month” headline.
Sources: Wall Street Journal (often paywalled) · Hindustan Times republication of that article · StackEasy about page
Alex Finn — one-person businesses run with AI assistants
Alex Finn, an AI explainer and product builder, told The Calum Johnson Show (episode dated October 12, 2026) that he is “a total one man business,” with “zero employees,” and that an AI assistant is “basically the operating system behind most of what I do.”
What they actually did
- He gave an assistant named Cindy the job of watching and answering business email. He did not turn her loose on day one. He started with research, then drafts.
- Early drafts “didn’t sound like me,” so he fed the assistant example emails he had actually sent, told it to keep those in memory, and had it rewrite until the tone sounded human enough to use in a negotiation. Only after he trusted the drafts did he let it take actions.
- His starting advice: pick two or three parts of your life or work that eat time, and automate those first. He also described “reverse prompting”: tell the assistant about yourself and your goals, then ask it what should be automated next.
- In a separate project, he had an assistant help design and build Agent Use Cases, a public directory of things people are doing with AI helpers, plus a weekday newsletter. A second assistant maintains the site and reports traffic. He has said he still did the social posts that drove visitors himself.
Revenue and reach: on the same podcast, Finn said Cindy “has closed me five figures of deals over the last couple weeks” and “closed a $10,000 sponsorship deal.” Those are his own claims; we did not see a contract or a payment record. He also said his newsletter, Ship It Weekly, has “40,000 people” subscribed — his own claim. The show’s packaging used larger income figures than Finn’s on-mic wording; those host figures were not independently verified and are not repeated here as facts. A write-up in The Human Touch noted that Finn already has a large audience, “an advantage few people share.”
Sources: Calum Johnson Show episode / transcript · The Human Touch write-up · Agent Use Cases · Ship It Weekly
Claire Vo — ChatPRD
Claire Vo, of San Francisco, founded ChatPRD, a tool that helps people write product documents and specs. Before this she had been a chief product and technology officer at companies including LaunchDarkly, Color, and Optimizely. The July 29, 2026, Wall Street Journal is the main news source. ChatPRD is not a one-person company today: the Journal reported she hired an engineer after nine months, and as of mid-November 2026 the company’s about page listed a founder/CEO and a Head of Growth. The honest framing is: started alone with AI, still a very small team, with AI doing a large share of marketing, sales, and support.
What they actually did
- In late 2023, while still working as a tech executive, she used ChatGPT to help code an app for product documentation. She told the Journal, “I was copying and pasting from ChatGPT.”
- She put the app online for $1 a month. The Journal said that within weeks people downloaded it thousands of times.
- She ran the company solo for nine months, then hired an engineer. The Journal said that nearly three years later the company had 100,000 users, and that AI handles marketing, sales, and customer support.
- In a July 15, 2026, interview she described several specialized assistants with separate jobs — including one with email and calendar access, and a sales assistant that works inbound leads and books meetings. That interview is her own account and promotional in tone.
Revenue and reach: the Journal reported the company “is on track to make seven figures in profit this year,” and cited 100,000 users. Those are founder figures reported by the Journal — her own claim, not an audit. Vo herself added a caveat worth keeping: “I think people over-index on how easy AI is and under-index on how much I did to get to this point.” In that same interview she said, in substance, that a person would have to start years earlier building an audience and decades earlier building expertise.
Sources: Wall Street Journal · Hindustan Times republication · ChatPRD about page · Solo Founders interview
Ben Broca — Polsia
Ben Broca, of Sausalito, California, was described by the Wall Street Journal (Te-Ping Chen, July 29, 2026) as running, alone, a company that offers AI tools to entrepreneurs. A later write-up of his own conference talk identifies the company as Polsia. This is a venture-backed company selling AI tools, not a shopkeeper who picked up a free assistant. The operating lesson — one person using AI for email, code, and support — is still real. The “ordinary person” framing is not.
What they actually did
- Per the Journal: AI tools answer his email, help write and debug code, field customer requests, sign up new subscribers, and issue refunds.
- Per a write-up of his talk: a user describes a business idea; one assistant makes a plan; other assistants handle jobs such as outreach, ads, support, operations, and engineering. Each runs a loop — look at the state of the business, pick a task, do the work, report back.
- Cost control is part of the story. The Journal said he was losing money on many customer accounts while paying for a paid AI model, and that he later switched to free open-source models. In the earlier talk he described bills that “didn’t really match how much I was charging customers,” and routing routine steps through cheaper models.
Revenue: the Journal reported that he had already added 10,000 paying customers and was “on track to bring in $10 million in revenue this year,” and that he said he had raised $30 million from investors. These are founder figures reported by the Journal — his own claim, not an outside audit of the books.
Sources: Wall Street Journal · Hindustan Times republication · Render blog, July 23, 2026 · Polsia
A caution from the same Journal article
The Journal also described Samir Ahmad, of Breinigsville, Pennsylvania. After almost two decades at Verizon he left to start a solo coaching and consulting business. He used AI to chart a business plan and help with marketing and called it “like my chief of staff, a second in command.” The business “petered out within months,” and he went back to a full-time corporate job. A tool is not the same thing as a working business.
People Using It in Their Personal Lives
The same habit works at a kitchen table as it does in a shop: hand one repeated job to an assistant, keep approval in your own hands, and prove it before you add the next job.
An electrician, 82, and an assistant named Nova
I am 82, and I still work as an electrician. I use an AI assistant named Nova to help run my day — scheduling, email, invoices, research, and staying on top of bills. I check the work. The assistant does not get the last word. Used that way, it takes the back-and-forth out of the desk work so I can spend more of my time on the jobs themselves. That is a practical example, not a marketing claim. A shorter version of this already sits on AI Opportunities.
Two public examples already documented on this site
These are family and medical stories, not business start-ups. They are already sourced on Instances of real people on the off-ramp. They are leads and time saved, not a promise that AI is a doctor.
- Joshua Resnikoff and his son Shiloh. Shiloh went years without an answer. Resnikoff then built a company that uses AI to flag other children sooner, on the order of weeks instead of years. That is not a treatment he found for Shiloh. Source, as listed on the Instances page: EOFire, “From 7 Years to 12 Weeks,” with Joshua Resnikoff. eofire.com/podcast/joshuaresnikoff
- Matthew Might, a computer scientist. An AI reader of the medical literature pointed to a compound he used for his son Bertrand, who had NGLY1 deficiency. A later use of a related tool pointed to a lead for a different rare condition, ADNP syndrome, that moved toward a clinical trial. Sources, as listed on the Instances page: Rare Disease Advisor and Medical Xpress, May 2020.
More everyday cases belong here as they check out. Education examples — tutoring tools used in classrooms and after-school programs — are also collected on the Instances page.