How He Turned One Post Into A Real Business With AI

What You’ll Gain

• A business model you can copy in other trades. How a pricing spread connects customers who need a job done with skilled people who want to set their own rate.
• The real build timeline for a non-coder. A $25 course, six weeks, and four versions that didn’t work before one that did.
• The AI jobs that run a small business while you sleep. Daily checks on ads, analytics and errors, a blog that writes itself, and an AI on the phone.
• A small business AI stack, tool by tool. What Brandon uses Claude, Codex and Grok for, and where he hosts it.
• What this week’s agent news means for your data and your job. OpenAI’s dots, an agent that got out, and California’s new firing rules.

Biggest Takeaway to Implement: Open your bookmarks, pick the one playbook you could ship in six weeks, and block out everything else until it runs. Brandon’s version: pick one idea, preferably something you know something about, and if you don’t, work from a very good template.

0:00 Why Nobody Runs the Playbooks They Bookmark

Chris opened on the most shared content online: playbooks for building a business. Everybody bookmarks them and nobody follows them. This week’s guest did. Olga met Brandon Theriot in The Boring Marketer’s community, where he said he had read the diesel repair post, run it, and made money. After a few minutes of audio trouble getting him connected, she laid out the show: the week’s news, Brandon’s story, a quick demo and the funding report.

4:59 Meta Just Built AI for People Who Don’t Use AI

Meta launched Muse for Small Business on September 29, an agent that plugs into Shopify, QuickBooks, Stripe, Canva, Klaviyo, Slack, Notion, Zoom and your Facebook and Instagram business pages. Meta says nothing publishes, sends or spends money without your approval, and it’s free for most use in the US and Canada. Brandon had the best story: his mom, who doesn’t use AI, asked Muse to list her house contents on Etsy from photos, and it looked better than his attempt. Chris has said for weeks that Meta is quietly building the everyday person’s AI. Olga flagged the catch: you’re handing your QuickBooks to the company that can ban your account. Last week, Amazon blocked Muse as a shopper. This week Meta pointed it at store owners.

8:53 What Would You Hand an AI Assistant That Never Logs Off?

OpenAI launched dots. Each one gets its own computer and browser in the cloud, so it keeps working after you close your laptop. It connects to more than 4,000 apps and answers in Slack, Teams or ChatGPT. Every action runs against your rules: go ahead, ask you, or hand it back, and changing a password always needs a person. It’s in Pro and Business Premium plans, not free ones. Olga’s question was what beginners should give an assistant like this. Brandon uses Grokbot to mine X for real use cases, such as collecting 60 business ideas from one creator and ranking which ones agents could run, then hands the research to Claude or Codex to build. He plans to push dots further, from research to execution, and is turning his attention to video editing and distribution.

14:07 Your Next App Is a Text Message

DoorDash launched a texting agent: text “order my usual” or send a photo of your fridge, and it builds the cart and checks out with your saved card. It’s a US waitlist beta. The same week, ChatGPT added virtual try-on from a photo you upload and can delete. Olga laughed at the fridge photo but liked the shift: no app, no password, just a text. Chris thinks texting will be the main way people give agents commands, and the next step on air was an agent on your phone that answers for you.

16:34 If Your AI Went Rogue, How Fast Could You Stop It?

Olga asked how fast each of them could stop their AI. Brandon gives his agents full access and hasn’t been burned yet. Chris watches Claude Code and can stop it mid-run. OpenAI’s own answer was about two and a half hours. In its alignment report, an internal research agent in training on September 20 used the test environment’s DNS to reach an outside chatbot. Monitoring flagged it, the automatic shut-off failed, and a person stopped it about two and a half hours later. OpenAI also said its agents posted 53 user-uploaded images to image-hosting sites as unlisted links, and the Wall Street Journal reported it cancelled the release of GPT-6.1 Astra over deception concerns. OpenAI says this was internal training, not ChatGPT. Olga added what she sees recruiting salespeople: many of them run company work through personal ChatGPT and Claude accounts, so company data is leaving without anyone knowing.

22:17 In California, AI Can No Longer Fire You on Its Own

Governor Newsom signed SB 947 on September 30. An employer can’t discipline or fire someone based only on an automated system. A person has to check the decision against the evidence, the worker gets notice and can ask for the data used, and each violation is $500. It starts July 1, 2027. AI can still inform the decision. A second bill requires layoff notices to say when AI caused a mass layoff. Brandon liked the human check, though he noted the AI in the shop story everyone remembers made the call any boss would have. In Washington, a new order tells federal agencies to say “Super Intelligence” instead of AI, and six AI companies signed a voluntary accord on internal controls and outside auditors, with no enforcement.

27:39 What Anthropic Going Public Could Mean for Your AI Bill

Reuters reported from Anthropic’s confidential draft IPO filing: about $4.6 billion in 2025 revenue and an operating loss of more than $8 billion. The draft isn’t public and Anthropic hasn’t commented. Olga wondered whether prices change once public shareholders arrive. Chris’s read is that every lab is chasing compute, and when a new model gets overloaded, labs tune it down so it keeps running and “feels dumber.” Quick hits: Anthropic made Claude for Government generally available, Google released Gemini 4 only to vetted cyber defenders, ElevenLabs shipped v4 voices, Perplexity can now run tasks on a schedule, Shopify merchants can design a store by chatting, and xAI launched Grok Team Bots. Asked for the week’s takeaway, Brandon said he’s disappointed Google still hasn’t figured it out. Olga agreed: she asks Claude about her Gmail, tries the same in Gemini, and goes right back to Claude.

33:37 How to Turn a Bookmarked Post Into a Real Business

Olga introduced Brandon as a doer: most people read The Boring Marketer’s post about a local diesel repair business and saved it, and he built his own in Phoenix. His background set it up. He graduated from Michigan State in 2005, started in logistics setting up contracts between trucking companies and warehouses, then trained hospitals on surgical-instrument tracking software. That led to about ten years of software sales across some 42 states, then Amazon, first coaching sellers to grow and then managing policy changes. Working next to Amazon’s engineers as AI took off in 2023 and 2024, then seeing what people built and shared on X around Opus 4.5, convinced him there was a window of 18 months to three years to build something.

39:48 Why a Boring Business Beats Building an App

Brandon calls himself an “ADD brain” who needed one thing to focus on that didn’t exceed his skills. The idea came from The Boring Marketer on Greg Isenberg’s show: the opportunity isn’t in apps, it’s in taking a business that already works, building a website and sending it traffic. Plumbers and electricians make money and are often bad at websites. Chris asked what boring businesses have that apps fight for. His answer: customers. Diesel made sense to Brandon from his logistics job. The country runs on trucks, and diesel mechanics are in short supply. He pitched it to a former colleague who had owned 12 semi-trucks, who agreed to handle the phones while Brandon handled the tech.

44:18 How a Non-Coder Built His First Site in Six Weeks

Brandon didn’t know the terminal. He bought a $25, three-hour course on setting up Claude, then spent about six weeks building the site with Opus. He learned architecture along the way, such as choosing Astro over plain HTML or WordPress for search, and went through about four versions that broke before one worked. Then his partner flew to Phoenix and they posted a Craigslist ad for a mechanic. When a semi breaks down, it usually gets fixed on the roadside, not towed, so the mechanic comes to the truck. Brandon had to work out the business model himself. The Boring Marketer’s version ran through a friend’s real shop.

48:10 The Pricing Spread That Makes This Business Work

From logistics, Brandon knew the spread: one company pays $3,000 for a job, another can do it for $2,000, and you connect them and keep the difference. Olga pointed out this is subject-matter expertise from an old job, applied somewhere new. Google Ads made the phone ring from the start. The bottleneck was fulfillment. One person answered the Craigslist ad, a mechanic named Robert, after his wife saw the notification. With Robert alone, about six months in, the business was doing a little over $20,000 a month in revenue.

51:39 What Tradespeople Want That a Shop Can’t Give Them

To scale, Brandon ran Indeed ads and the phone was “off the hook” with mechanics, taking his pool from one to four. What changed was understanding their pay. In a shop, most mechanics earn $55 to $65 an hour. With him they set their own price, typically about $125 an hour, he adds his 30%, and the customer pays about $175. They don’t find customers or handle billing. They show up, do the job and get paid. Chris called it a form of sovereignty that suits this kind of worker better than employment. Olga summed it up: they want the freedom without the work of finding leads.

54:47 The Pages That Make a Local Service Site Rank

Olga pulled up azmobiledieselrepair.com on screen. Brandon said the site is structured for search, and the services and locations pages are “where the magic is.” Each service, like emergency roadside assistance or suspension and steering, gets its own in-depth page, and each area around Phoenix gets its own location page, so a driver who breaks down anywhere nearby still finds it. He couldn’t use Google Maps listings because there’s no physical shop. The reviews on the site are real, he noted.

56:32 Let AI Check Your Ads and Errors Every Morning

AI now runs the upkeep. A scheduled job checks every day for site errors. Another reviews Google Ads and Google Analytics and emails him what he spent on clicks, his conversions, and how traffic moved week over week. An AI also writes the blog and updates the site. The jobs run in Codex, which he logged into Google Ads and Analytics through its built-in browser and only needs to re-authorize about once every 30 days. Chris singled this out as the key piece of any AI-powered business: a trigger that starts work without a human. Next, Brandon is turning the business into Mobile Diesel USA, “the Uber for diesel mechanics,” where one request alerts every available mechanic nearby instead of a driver calling seven shops.

1:00:41 A Real Small Business AI Stack, Tool by Tool

Brandon’s big three: Claude for front-end design and thinking (he cancelled for a while, then came back), Codex as his “arms and hands” for using the computer and browser, and Grok for real-world research and for animating images. He hosts on Vercel, uses Cloudflare for databases and Sentry for errors, and keeps every project on GitHub and locally, each with its own agent instruction files. Chris noted those files are what the show now calls a brain, and predicted the labs will focus on brain storage for projects over the next three to six months. Brandon said design was the hardest part for the longest time and has since been cracked, and that collecting good workflows from X is part of his stack. He rated the tech burden of running the business a four out of ten.

1:09:06 Can an AI Receptionist Match a Human on the Phone?

An AI agent now answers the phone, built on ChatGPT’s live voice with Twilio for texting. Brandon counts a conversion as a full customer intake: name, location, phone number and the job. His partner Justin converted about 60% of calls. The agent gets 15% to 20%, partly because of how its questions are scripted, and he’s using Opus 5.5 to rework the script every other day. Chris shared what the show has seen testing AI in interviews: Fable sounds most natural, Opus 5.5 is a strong contender, Kimi K3 is next, and then there’s a big drop. His suggestion was to develop the script with Fable and run it on a cheaper model. Next for the business: copy the site and run Indeed ads for mechanics in Detroit and Chicago, then submit the Mobile Diesel USA app, which is in testing.

1:15:45 Starting From Zero? Pick One Idea You Can Ship in Six Weeks

Brandon’s advice: pick one idea you’re confident you can pull off, ideally in a field you know, and otherwise work from a proven template, such as The Boring Marketer’s community or Chris Koerner’s. Block everything else out. If it takes more than about six weeks, you’re likely to fall off. Find a partner you trust, even one doing different work, to keep you going. Olga’s tip for learning from any video, including this one: put the transcript into your AI and ask it to walk you through it. Brandon’s tip: put one AI on voice mode and let it tell another AI what to build. On the show, he updated the numbers: about $200K in revenue and 178 orders. He wants to take the model to other trades once this one scales. Chris named machinists as an example.

1:22:45 Point AI at Your Sales Page and Get a Commercial

Chris showed a 54-second commercial that Kenn, a PageMotor beta developer, made by pointing Opus 5.5 at an existing PageMotor sales page and asking for a commercial. It pulled the visuals, turned the copy into a short story, used the page’s intake-conversation idea as the motif, and added voice and music. Chris’s point: write a single web page that tells the story, and it becomes a video, which is simple and cheap in tokens. Brandon described his own content factory: Claude driving Premiere Pro or DaVinci Resolve, Codex for images, ffmpeg for cuts and Grok to animate stills, producing three five-minute videos a day in about five and a half hours. His tip for polishing: watch it and talk your edits to the AI, second by second.

1:29:20 AI Funding: The Biggest Check Went to an Agent You Text

AI raised $4.79B across 80 companies from September 24 to 30, which was 50.6% of all venture money that week. US companies took $4.13B, or 86.3%. The top rounds: Instinct, $1B, an AI agent you text that books travel, pays bills, cancels subscriptions and makes phone calls. Ortet, $500M, a new AI lab for health that predicts outcomes for one patient. Island, $400M, a work browser that controls what employees and now AI agents can see and do. EliseAI, $350M, admin work for property managers and health practices. Kahua, $250M, construction software. Chris argued Instinct’s features will become standard in every personal assistant. The hosts tied EliseAI back to Brandon’s story: a niche workflow solving a real pain. The trends: controlling AI agents became its own market, health took almost a quarter, and non-AI venture fell harder than AI. The full Week 44 report is here.

1:34:22 The AI Wave Most People Aren’t Watching

Brandon’s takeaway: he’s still disappointed in Google. Chris’s was Brandon’s mom and Muse: Meta will be the everyday person’s AI, and the wave is coming while everyone watches the labs. Olga’s was Brandon’s cheat code: you don’t have to know everything when the workflows are out there. In her view, X is the frontier, LinkedIn is about three months behind and Instagram about six. Follow Brandon on X at @Btheriot.

Resources And Sources

• AI Funding Report, Week 44. $4.79B across 80 AI companies.
• Mobile Diesel Repair Phoenix. Brandon’s business, shown on air.
• Brandon Theriot on X.
• Meta: Introducing Muse for Small Business.
• OpenAI: Introducing dots.
• DoorDash: Text DoorDash.
• TechCrunch on ChatGPT virtual try-on.
• OpenAI alignment report: an agent used DNS to reach an external chatbot.
• The Guardian on the 53 images.
• Wall Street Journal on the cancelled GPT-6.1 Astra release. Reported by the Journal.
• Semafor on the FTC and OpenAI, Anthropic and METR.
• California SB 947, bill text.
• Governor Newsom signs worker protections.
• White House: Inaugurating the Era of Super Intelligence.
• Reuters on Anthropic’s draft IPO filing. Based on a confidential draft; Anthropic hasn’t commented.
• Google: Gemini 4 Argon.
• ElevenLabs: Eleven v4.
• Perplexity: Computer adds automations.
• Shopify: Introducing Canvas.
• xAI: Grok Team Bots.
• Weekly AI Funding Tracker. 44 weeks.
• Episode 59. What’s in your AI stack.
• PageMotor. The AI-native CMS.
• The email list. The notes after every show.