The Kid Who Lived in a Car Is Finally Going to Japan

I was 14 when my mom scraped together money she didn't have.

Not for a vacation. Not for anything fun. For self-defense lessons — because I was getting bullied, badly, and she couldn't watch it happen anymore. We didn't have the cash. There were stretches where we didn't have a house either; I spent part of my childhood living in a car. She found the money anyway, the way moms somehow do.

I trained for years. And somewhere in those years, something changed that had nothing to do with throwing a punch. I got confident. I learned to keep showing up when it was hard and boring and I wasn't good yet. Persistence and perseverance — those two words sound like motivational-poster filler until you've actually had to live them. They're the two things I credit for almost everything good that's happened since.

Here's why I'm telling you this instead of keeping it to myself: if you're building something right now — an agency, a business, a skill you're still bad at — the thing that decides whether you make it isn't talent. It's whether you keep going after it stops being exciting. That's the whole secret. The people who win are just the ones who didn't quit on the boring middle part.

That bullied kid who lived in a car had one dream he never let go of: Japan. He's been carrying it for about 35 years.

In a couple of weeks, I'm finally going — with my son. The kid who couldn't afford the trip is taking his own kid to see the thing he dreamed about his whole life. I don't really have words for that one, so I'll just say it's the reason any of this work matters.

And while we're counting wins — we just crossed 100,000 subscribers on YouTube. I'm not going to pretend that's small to me. Every one of you who hit follow, watched a build, sent a message — thank you. The whole reason I make this stuff is to hand you the things nobody handed me when I was coming up broke and clueless in this business.

Speaking of which — I just gave away another 4-hour Build-A-Thon, free. That's 12 total hours of complete builds, from a blank screen, no gatekeeping, plus a stack of deals and freebies I dropped along the way. And I'm not taking any of it down. I'm making the builds evergreen — watch them anytime — and keeping the freebies open so you can grab them on your schedule, not mine. Whether you're building your own business or an agency to serve clients, it's all there waiting. (Both linked down in the deals.)

So that's me for the next couple of weeks. I'll be in Japan with my son, finally cashing in a 35-year-old dream, and I'll bring the photos and stories back here when I'm home.

Persistence got the bullied kid off the car floor and onto a plane to the one place he always wanted to go. Whatever you're in the middle of right now — keep going. The boring middle is where it gets decided.

See you in a couple weeks. Thank you for being here.

— Nuno

🚀Your Inbox Now Answers Itself — And Books the Call

You know the pile. The unanswered emails sitting in your inbox right now — the "do you offer X?" and "how much is Y?" and "can you help with Z?" — the ones that should take two minutes each but somehow eat your whole afternoon. The one a paying customer sent four days ago that you still haven't replied to.

HighLevel just killed that pile.

They dropped an AI email agent that lives inside your CRM, reads incoming replies, and answers them in your voice — booking appointments, qualifying leads, and running first-line customer service without you touching the keyboard. And before you roll your eyes: yeah, the email AI used to be broken. It read everything, made a mess of chains, wasn't usable. That version's gone. This one works.

Here's what's now on the table:

• A real agent inside your email, not a canned autoresponder. You build it under AI Agents → Conversational AI (emails count as conversations), flip it to email-only, and it picks up replies and responds like a trained team member — not a "your message is important to us" robot.

• It answers from a knowledge base you control. Feed it a doc with your services, pricing, links, and FAQs, and it talks like someone who actually works for you. (Nuno generated his entire knowledge base by asking Claude Co-Work — which already knows his business — to write it. Two minutes, one markdown file, done.)

• It closes, not just chats. In a live test, a reply asking "do you offer one-on-one coaching?" came back with the price ($497/hr), exactly how it works, and a booking link — in a single message. That's not customer service. That's a salesperson who never sleeps.

• Deliverability is baked in — if you set it right. Connect your sending domain (Mailgun, whatever you use) under email services, then keep every reply in plain text. No design editor, no image-stuffed templates. Plain text is what keeps these in the inbox instead of spam.

• Human handover when it matters. Anyone who asks for a manager — or hits a question the bot can't handle — gets routed straight to a real person. You're not handing your reputation to a bot with no exits.

Real talk on the one catch: this only fires on replies to emails that went out of HighLevel first. Someone has to already be in a thread with your system — it's not parking on a cold inbox catching every random message from any address on earth. So the move is to make HighLevel your sending engine for anything that might get a reply: confirmations, follow-ups, outreach. Once the thread starts there, the agent owns it.

The shift here isn't "email got easier." It's that the front line of your business — the first reply a customer gets, the one that decides whether they book or bounce — no longer needs a human in the seat. Set the knowledge base, set the guardrails, and the inbox runs without you. A whole job function, handled, for the cost of a prompt.

Here is the SOP: https://drive.google.com/file/d/1i7ordsw9ehRsGJSiA8JrHQMyObWh0DNg/view?usp=sharing

 🤖 The Claude Code Habit That Cuts Your Bill and Keeps It Sharp

Two things happen when you run Claude Code for an hour straight. Your token bill climbs. And the output gets dumber.

You've felt it — the answers were sharp at the start, then somewhere around minute 30 the thing starts forgetting what you told it, repeating itself, inventing files that don't exist. That's not the model being bad. That's the context window getting bloated, and you paying top-tier rates to drag all that weight around on every single response.

There's no secret model the pros are using. There's a habit. It's session hygiene, and it's three moves:

• Right-size the model before you start — never mid-session. Type /model and pick the tier that fits the job. Quick questions, simple edits, thinking out loud? Drop to Haiku — fast and cheap. A genuinely complex build? Bring the heavy model. The one rule: don't switch mid-project. Swapping models mid-session reloads your entire context into the new model, and you pay for all of it again. Set the tier on the way in.

• Cut the session at clean stopping points. When a thread's been running long, don't just keep piling on. Tell Claude: "Good stopping point — save what we've done to memory and write me a short bullet-point handoff prompt for the next session." It writes the handoff. You grab it and move.

• Then /clear and paste the handoff to start fresh. New session, clean context, the work already saved to memory. The next answer comes back sharper and cheaper, because the model isn't re-reading a novel's worth of history just to respond to you. (There's also /compact, which summarizes in place — but a clean /clear plus a tight "here's what's next" handoff gives better quality. Captured memory, empty window.)

Same moves work in Codex — Nuno walks through both in the video. And a bonus trick he buried in there: when a technical concept is over your head (he hit this with a dense article on prompt caching), paste it into a fresh Claude chat and ask for the bullet-point, explain-it-like-I'm-five version. Understand the thing in two minutes instead of pretending you read it.

Here's the part most people miss: the gap between someone burning $400 a day on Claude Code and someone shipping the same work for a fraction of it isn't talent or access. It's whether they treat the context window like it's free. It isn't. Manage it like an operator and the same tools get cheaper and smarter at the same time.

 🧠 The Week the Government Pulled Anthropic's Best Model — and an Open One Crashed the Party

Real wild week. A nearly trillion-dollar AI lab had its top model yanked off the market by the US government — overnight, for every user on the planet — and three days later a free, open-weight model showed up benching near the very top. If you build anything on AI, both stories are really about the same question: how much can you trust any one model to still be there tomorrow?

Anthropic got its flagship pulled — first time that's ever happened to a commercial model.
On June 12, the US government forced Anthropic to suspend Mythos 5 and Fable 5 — its most advanced models — over export-control concerns about foreign access. Not throttled. Gone, for everyone, with basically no notice. Anthropic called the underlying issue a minor vulnerability; the government clearly disagreed; and there's a messy backstory about a jailbreak dispute and who flagged it that's still mostly speculation, so I'll leave the palace intrigue alone. Reports say access could return within days. But here's the part that matters whether you love Anthropic or not: a model thousands of people were building on disappeared on a Friday evening because of a decision they had zero say in. The takeaway for builders isn't who was right — it's that you can't treat any single model as guaranteed to be there tomorrow.

Meanwhile, Z.ai dropped GLM 5.2 — open-weight, near-frontier, basically free.
Same week, a Chinese lab shipped an open-weight model built for coding and agentic work: 1-million-token context, MIT license (download it, fine-tune it, do whatever you want). On coding benchmarks it's landing right behind the top Claude models and ahead of GPT-5.5 on some of them — at a fraction of frontier pricing, roughly a quarter of Opus-tier cost by the numbers shown. It's not magic (in a live test it took a few tries to build a simple game, and the slide deck it spat out was clean but not jaw-dropping). But "open, free to try, and nearly as good at code as a model that costs 4x" is a real shift. The open models aren't catching up someday. They're caught up now.

Put those two side by side and the lesson writes itself: the model you depend on can disappear, and the cheap open alternative is closing the gap fast. Optionality just stopped being a nice-to-have.

Rapid-fire — the drops that actually matter for your business:

• Google "Ask Ad Manager" — a chatbot inside Google Ads that reads your campaign data and tells you what to fix and where to optimize, the same way Ask Studio works for YouTube. If you run paid traffic for yourself or clients, that's free optimization sitting in the dashboard.

• OpenAI Codex "Record & Replay" — record yourself doing a repetitive task once (uploading a video, pulling from a sheet), and it turns that recording into a reusable skill it runs for you on command. SOPs that actually execute.

• Claude Design, on-canvas editing — edit designs directly on the canvas, pull in components from GitHub repos or design files, and it builds against your design system before you ever see it. Hops between Lovable, Canva, Vercel, and more.

• Perplexity "Brain" — an agent that reviews its own work overnight, learns from your corrections and its dead ends, and gets more efficient the more you use it. Their answer to the self-improving agent everyone's been hyping.

• Adobe's AI assistant lands in Premiere, Illustrator, and InDesign — prompt the edit instead of clicking through it.

• Midjourney Medical — yep, the image-gen company launched a medical arm with an ultrasound body-scanner and a "spa" concept (San Francisco, 2027). Bootstrapped, no investors, almost certainly a data-for-AI play. (Experts are already saying: stop calling it an MRI replacement — it isn't.)

• Pew survey — about half of US adults now use AI chatbots, up from a third in 2024, and ChatGPT still owns the category at 44%. Adoption went mainstream; the brand most of your customers mean when they say "AI" is still ChatGPT.

The pace isn't slowing — a top model pulled, an open model landing near-frontier, four agent tools shipped, all in one week. You're not going to keep up with every release, and you don't need to. The move is to build portable: keep your prompts, systems, and automations model-agnostic so you can swap whatever's winning this month without rebuilding anything. The operators who win in 2026 won't be married to one model. They'll route to whatever's best this week — and sleep fine when one of them gets pulled on a Friday.

Epic just shipped something that shouldn't be possible. Unreal Engine 5.8 dropped with an experimental MCP server — the same kind of connector you've been hearing about for Claude Code — except now it plugs your AI model straight into a triple-A game engine. You type what you want. The engine builds it.

And we're not talking "here's a pretty picture you're stuck with." We're talking real, editable, drag-it-around geometry.

And we're not talking "here's a pretty picture you're stuck with." We're talking real, editable, drag-it-around geometry.

• It built a whole city from prompts. In the demo they described districts, ran highways through them, dropped a forest around the edges — and Unreal generated actual procedural geometry you can open and edit, not a flat render. Start small (a sofa, a rug, a "reading nook" chair it furnishes for you) or go big (a full metropolis). Same workflow.

• It lights the scene for you. Say "overcast skies." Or get weird and say "Bogotá at 9:30 in the morning" and it works out the real sun position, direction, temperature, and atmosphere to match. Hand it a photo of Vancouver and it'll chase that look until it nails it.

• It does the stuff that used to eat months. A hazard zone — flashing lamps, rising steam, a danger meter on the HUD, all wired to fire the second a player walks in — the kind of thing that normally takes a team weeks of back-and-forth across five disciplines. Their artists built it in days by describing it.

• You're still the boss. Model gets it wrong (it misread the sky and botched the clouds at one point)? You see exactly what it did and fix it in the next sentence. Grab a road, drag it, the city reshapes around it. Collaboration, not a slot machine.

• The movie version is coming next. The same bridge hooks into image and video models (Nano Banana, Seedance, and friends), so you block out a film scene in Unreal and let the models generate the finished look. That half lands early next year — and it's aimed squarely at filmmaking and architectural visualization too.

Here's what makes this more than a flex: the wall between "I can picture it" and "I can build it" just got a lot shorter. You don't need a decade of node-graph wizardry anymore — you need to know what you want and how to say it.

The game-dev tools ship with 5.8 right now, free and open — point your own model at it and go. The film and architecture stuff lands early next year. Either way, the people who "can't design" are about to out-build the folks who spent ten years learning it the hard way. Go poke at it.

💲Deals at a Glance

🚀 Claude Code Workshop — 4 Days, 8 Hours, $67: Build apps, AI agents, and automations from scratch — hands-on, start to finish. 👉 Register Now

🎓 NEW — The Automated Marketer Skool Community (LIVE NOW): The room where the building doesn't stop. Twice-a-week group calls with me, plus courses on vibe coding anything — HighLevel, Claude Code, Lovable, AI, and automations. Introductory pricing for a limited time, with a lifetime option on the table. Two ways in:

• 👉 Join the Free Community — free access, free course, daily community

• 👉 Go VIP — $47/mo intro — hot-seat support, advanced templates, inner-circle access

🏗️ Agency Build-A-Thon — Free: A complete AI agency built from zero on GoHighLevel — domain, white-label SaaS plans, a client-ready website, voice + chat AI, the whole back end — recorded live, every step in order, nothing re-staged. Your start-to-finish blueprint if you sell to clients (or want to). 👉 Watch Free

🚀 Business Build-A-Thon — Free: Two real businesses built online with AI in one afternoon — domain, website, automations, and the AI that answers the phone, all from a blank screen. For launching your own business, not clients'. 👉 Watch Free

🛠️ Done-For-You — Built by Me and My Team at Rapid Active Marketing: Don't want to build it yourself? Skip the learning curve and have us build the whole thing for you — GoHighLevel setups, websites, funnels, automations, and AI voice + chat widgets — done right, by the people who actually teach this. We take on a handful of clients at a time, on purpose. 👉 See What We Build

💰 The Best GoHighLevel Affiliate Offer on the Planet: Everything you need to win with GHL, in one place. 👉 Get Started

🔥 Vibe Coding Incubator — FREE: Hands-on vibe coding, live weekly trainings, app templates, support, and Lovable how-to's. 👉 Join Free

😲 GoHighLevel Masterclass on Udemy — $20: 5,000+ students, 4.7★, 18+ hours of content. 👉 Grab It

Remember, there’s always something new coming from the Automated Marketer YouTube Channel—new content drops every Tuesday & Thursday. Subscribe so you never miss the next big idea.

Keep watching, keep learning, and keep moving forward!!

Nuno