Remember spending an entire Saturday afternoon rewinding VHS tapes with a pencil? You’d grip that thing between your fingers and keep turning, keep turning, until the tape finally hit the beginning. Stupid, repetitive work that a machine could’ve done in seconds. Well, that’s basically what AI loops do—except instead of wasting your time, they save it. And I’m telling you, this concept changes everything about how you can actually use AI to build something without losing your mind in the process.

Why This Keeps Coming Up

Every time I talk to someone about using AI to scale their work, they hit the same wall. They prompt it once, get something halfway decent, then spend three hours editing, rewriting, fixing tone, checking facts. They’re doing the work. The AI is just the first draft machine. It’s like hiring someone who gets it right maybe 40 percent of the time, and you’re responsible for the other 60 percent. That’s not leverage—that’s just more work with extra steps.

But here’s what I kept noticing: the people actually building real income with AI weren’t fighting this problem. They were just… not doing it. They’d set something up once and walk away. I couldn’t figure out how until I realized they were using loops. They weren’t prompting AI once. They were telling it to keep going until it nailed a specific standard.

What I Actually Found Out

A loop in AI terms is simple. You give it three things: a goal (what you want), a standard (what “done” actually means), and permission to iterate. Then you let it rip. The AI doesn’t stop at draft one. It keeps revising, keeps checking itself against that standard you set, until it hits the bar. No human intervention required between loops. You define the finish line, and it runs toward it.

I tested this with blog content first because that’s what I had lying around. Instead of “write me a 1,000-word post about AI loops,” I set it up like this: write 1,000 words, make sure it hits conversational GenX tone, check that it has three real examples from my own experience, verify the facts, and don’t stop until it reads like something I’d actually publish without touching it. Then I set it loose. Three loops later, I had something that genuinely needed zero editing. That had never happened before.

The key part—and this matters—is that the loop has to have a measurable standard. “Make it good” doesn’t work. “Make it conversational and include three personal examples and pass a fact-check against these sources” does work. The AI can actually evaluate whether it’s hit that bar or not. That’s what lets it keep going without you.

The VHS tape metaphor actually holds up perfectly. You don’t keep rewinding and saying “go back more.” You set the counter to zero and push rewind. The machine handles it. Loops are the same way.

What You Can Do With This Today

If you’re sitting on any content that needs doing—emails, social posts, product descriptions, even rough presentations—you can loop it today. Pick the smallest thing first. Define what success looks like for that thing with as much specificity as you can manage. Then tell your AI to iterate until it hits that standard and report back when it’s done.

I’ve built this into my income work and it’s cut my editing time roughly in half. That’s not nothing for someone like me who’s supposed to be building three different things at once. I’m not saying the first draft comes back perfect—it doesn’t always. But it comes back *close enough* that I’m not rewriting from scratch. That’s the difference between this feeling like work and feeling like I’m actually moving the needle.

This is the kind of thing GenX actually gets. We grew up doing things the hard way. We know what real efficiency looks like because we’ve lived without it. When something genuinely saves you work, you notice it. Loops do that.

Give it a shot this week. Define one thing you want done better, set the standard, and loop it. You’ll get why I’m writing about this.

—Jim

Watch the Full Video

I covered all of this in a short video too — sometimes it’s easier to watch than read.

Watch on TikTok →

We used to wait a week to get our photos back — and half of them were blurry. Now I’m getting pro-level images in about ten seconds. Inside Claude Code. I’m not exaggerating. I’m also not paying money to some AI image startup. This is happening because someone built a bridge between Claude and Google’s image generation, and it’s free.

Why This Keeps Coming Up

Every time I talk about building income with AI, someone asks the same question: “Yeah, but what about visuals?” They’re right to ask. Whether you’re making thumbnails, product mockups, social media graphics, or client deliverables, you need images fast. Most of us spent the last year jumping between tabs — Claude in one window, Midjourney in another, paying monthly subscriptions like we’re running a design agency.

The friction was killing productivity. You’d get a good image prompt written in Claude, then copy it, paste it somewhere else, wait for rendering, download it, come back. By then you’d lost your train of thought. I started looking for something that didn’t require context switching.

What I Actually Found Out

The tool is called Nano Banana MCP. MCP stands for Model Context Protocol — basically a way to bolt integrations onto Claude without Claude having to rebuild everything from scratch. What Nano Banana does is plug Google Gemini’s image generation directly into Claude Code. No switching windows. No new logins. You stay in Claude, write your prompt, hit run, and you get images back in about ten seconds.

The setup was easier than I expected. I went to the Nano Banana repository, grabbed the integration, connected my Google API key (takes maybe three minutes if you’ve never done it before), and it just worked. The real win came next: they built a free prompt template library built in. You’re not staring at a blank canvas trying to remember the syntax for “cinematic lighting” or “professional product photography.” Pick a template, swap in your specific details, run it.

I tested this with our Rewired channel. I needed four thumbnail options fast. Picked the “social media thumbnail” template, dropped in “Rewired channel thumbnail” as context, hit the run button in Claude Code. Four polished options came back. Not generic. Not “obviously AI.” Actual thumbnails I could use or hand off to someone else to refine.

Here’s what sold me: it’s repeatable. Claude Code is built for step-by-step work that you can do over and over. So now I can build templates. Run them. Tweak them. Run them again. Improve them. This is how actual work gets done — not one-off image requests, but a system you own and can scale.

What You Can Do With This Today

If you’re doing anything client-facing that needs visuals, this changes the game. Freelancers can knock out mockups in minutes instead of hours. Course creators can generate course graphics without waiting on Fiverr. Etsy sellers can test product designs before manufacturing. People building their own info products aren’t blocked on design anymore.

You could also use this for internal work that nobody sees. Testing ideas visually. Building mood boards for projects. Creating reference images for your actual designer (if you have one) to work from. The cost is zero because it’s riding on Google’s infrastructure.

I’ve started treating this like a utility now. Same way I use Claude for writing or thinking through problems, I’m using it for visuals. The speed changes what you attempt. When something takes two hours, you’re careful about when you do it. When it takes ten seconds, you just do it.

The bigger thing I’m noticing is that all these integrations are quietly stacking up. Claude with image generation. Claude with web access. Claude with code execution. We’re not waiting for the AI company to build the perfect all-in-one tool anymore. We’re assembling our own from pieces that actually work. That’s worth paying attention to if you’re trying to build something real.

Watch the Full Video

I covered all of this in a short video too — sometimes it’s easier to watch than read.

Watch on TikTok →

My Walkman ate a cassette mid-Side B back in ’92, and I jury-rigged it with a pencil and pure stubbornness. The song got finished. That same fix-it instinct—the GenX refusal to throw something out when you can hack it back to life—is exactly what I just applied to building a qualified lead generation site for under five hundred bucks. I’m making two grand a month routing pre-screened prospects to local service providers, and the whole thing runs on AI doing the heavy lifting.

Why This Keeps Coming Up

I’ve been talking about AI income streams for a while now, but what people actually ask about is the gap between “cool idea” and “money in the bank.” Lead generation is where that gap closes fast, because businesses will literally pay for qualified prospects. The trick is making the qualification happen without you sitting on a Zoom call with every tire-kicker who fills out a form. That’s where AI comes in—it does the vetting. You just set up the plumbing.

The micro-niche angle is the real move here. I went narrow: Social Security benefits consultants in specific cities. Not “financial advice.” Not “legal services.” Social Security. I picked that because I knew three providers in different states who were spending money trying to find people who actually needed their help. They weren’t getting qualified leads. They were getting noise. That’s a problem I could solve.

What I Actually Found Out

I built this using Claude on the free tier—seriously, free tier—to generate location pages for a dozen cities. Each page is slightly different, talking about benefits in that specific area, mentioning local landmarks, making it feel native instead of templated. Then I built one smart form using Airtable as the backend. The form asks three qualifying questions: rough monthly income, timeline for needing help, and zip code. Claude did the copywriting. It took me maybe six hours total, spread over two days.

Zapier runs free, and it’s the nervous system here. When someone submits the form, Zapier checks the answers against my qualification criteria—I need income under a certain threshold and someone who’s asking because they’re either retiring soon or already retired. If they qualify, Zapier automatically books them into a calendar slot with the service provider I’m working with, sends them a confirmation email, and logs everything in a spreadsheet I review once a week. Cold calling solved. Hand-qualifying solved. Administrative overhead solved.

The service provider pays me two hundred bucks a month for qualified leads. I’m averaging ten leads a month, so one of them converting to a client covers my entire cost structure plus profit. My actual costs: domain name, forty bucks a year. Airtable, free tier. Zapier, free. Claude, free. The only paid tool is Webflow at fourteen a month because I wanted the site to not look like a DIY disaster, and honestly, I could’ve built it on a free template if I’d wanted to.

What You Can Do With This Today

Start by picking a niche you actually know something about—something specific enough that you can name three to five businesses that would pay for leads. Don’t go broad. Go the opposite direction. Think “solar panel installers in Austin” instead of “home improvement.” Think “divorce mediation counselors in suburban Denver” instead of “legal services.”

Build your site in a weekend. Write the core copy yourself or use Claude. Create five location pages using the same template with different city names. Set up one form that asks three qualifying questions—the questions that would make a service provider sit up and pay attention. Wire it to Airtable and then to Zapier. Set Zapier to email your provider and log the lead. Reach out to a few local businesses in that niche and tell them what you’re building. Offer the first month for free just to prove it works.

This isn’t going to make you wealthy, but two grand a month is two grand a month. And once it’s running, you literally check it once a week. The rest is AI doing what it’s weirdly good at: asking screening questions and moving data around.

Anyway, that’s what’s working for me right now. Hit me up if you actually build one—I’m genuinely curious what niches you all find.

Watch the Full Video

I covered all of this in a short video too — sometimes it’s easier to watch than read.

Watch on TikTok →

We used to fax resumes and redial a busy signal for an hour just to order something. Now AI builds you a website that earns money while you sleep. I’m not exaggerating. I picked a random niche, spent a weekend writing, and two weeks later I had my first affiliate commission hit my account. This isn’t some get-rich-quick thing, but it’s real income from something that actually works.

Why This Keeps Coming Up

Every GenX person I know is thinking about the same thing right now: we’re in our fifties and we need more than one income stream. The old pension-and-401k model doesn’t cut it anymore, and honestly, we’re too old to pretend we want to climb a corporate ladder. What we need is something that scales without us grinding ourselves into dust.

Affiliate sites used to require either serious SEO knowledge or hiring someone who did. You’d wait six months for traffic. I’ve been reading about this stuff for years and never pulled the trigger because the barrier felt too high. Then I realized AI had quietly made the whole thing stupid simple. Like, embarrassingly simple.

What I Actually Found Out

I picked ergonomic office chairs for remote workers because I actually use one and figured if I needed real information, other people did too. I spent maybe twenty minutes in Perplexity asking it to research the whole market—bestsellers, price ranges, what people complain about, the gaps in current reviews. I got back actual sources, not hallucinations, and I now understood the market better than I would have spending three hours on Reddit.

Then I dumped that into Claude and asked it to outline five comparison articles. Not generic garbage—real structured comparisons between specific models, talking about who should buy what. Claude did the scaffolding in about an hour. I went through each one and added my actual opinions based on my own chair, mentioned specific problems I’d solved, and basically made it sound like a real person wrote it instead of a robot.

The whole thing took a weekend. I’m not being cute about that. Friday night I started, Sunday evening I had five articles ready to publish. I set up a basic WordPress site, grabbed affiliate links from Amazon Associates, and posted the articles with internal links between them so they reinforced each other.

Two weeks later, I made my first commission. Thirty-seven dollars. That sounds stupid until you realize nobody was working for those thirty-seven dollars. It just happened because someone read my article, clicked my link, and bought a chair.

What You Can Do With This Today

The bottleneck for GenX building affiliate sites was always content creation. You either wrote it yourself (slow and painful) or hired someone (expensive and often terrible). AI closes that gap almost entirely. You’re not paying someone to pump out mediocre reviews. You’re paying yourself to have good taste and opinions, which you already have for free.

The real work is picking a niche you can actually talk about credibly. I didn’t pick something random and try to fake expertise. I picked something I use. That matters more than people think. Your site doesn’t need to be about what’s trendy or competitive. It needs to be about something where you’ve already spent money and have formed real opinions. That’s actually your advantage over pure content farms.

Then you research the market (AI does this now), outline the content (AI does this), write most of it (AI does this), and then you edit it to sound like you because you’re the actual human with a perspective. You publish it on WordPress. You add affiliate links. You don’t have to wait for organic traffic to explode because you’re building systems that compound slowly instead of betting everything on one viral post.

The income is quiet and slow-building, but it’s real, it’s passive once you stop actively maintaining it, and it takes about forty hours to set up something that could make a few hundred dollars a month within six months. That’s not a retirement plan, but it’s a decent secondary income for someone my age.

Look, we didn’t get GenX where we are by believing everything we read. But this one actually works. I’d rather tell you I figured it out so you don’t have to waste three months researching whether it’s worth doing.

That’s it for this week. Go build something.

Watch the Full Video

I covered all of this in a short video too — sometimes it’s easier to watch than read.

Watch on TikTok →