✈️ Two Airports, Three Weeks, and an 8-Hour Video I Can't Stop Thinking About

I'm packing for Virginia as I write this, and in two weeks I'll be doing it again for Cancun.

Which is a ridiculous sentence to type, and not the point. The point is what happens in the rooms.

Because here's what I've noticed after enough of these — the people who show up to events are never the ones who need to be told to show up. They're already the ones building. They already joined the free thing, already watched the long video, already installed the repo. Showing up is just what they do. And the people who read about the event afterward and think man, I should have gone are usually the same people sitting on a HighLevel account they use 20% of.

That's not a scolding. It's just a pattern I've watched for six years.

Where I'll be:

• The Ugentic AI Summit, September 15–17, Ritz-Carlton, Tysons Corner. I'm speaking. I'm also there to meet people and see what everyone is actually doing with AI, because the hallway conversations are where I learn the most. If you're anywhere near DC, come find me. Say hi. Ask me the tech question you've been stuck on. 👉 https://www.ugenticaisummit.com/

• AI Unleashed, Cancun, two weeks after that. Tickets are still open. This one is small on purpose — intimate setting, hackathons running at night, real time with people instead of a room of 800 staring at a stage. If you've been waiting for a reason to work on your business somewhere that isn't your desk, this is it. 👉 https://www.aiunleashedintensive.com/

And there's a lot brewing on the VC Inc side. New workshops coming, and they fill fast. If you're not in the free community yet, get in there before the next one — that's where they get announced first.

Which brings me to the thing I actually want to say.

Everyone selling you AI education right now is selling you the next thing. The next tool, the next model, the next workflow that makes the last one obsolete. And people are drowning in it. They buy the course, don't finish it, buy the next course, and eight months later they've spent $4,000 and shipped nothing.

I did the opposite this month.

I sat down and recorded the entire thing. Eight hours. No cuts, no edits, no "we'll cover that in the paid version." Buying the domain. Setting up the email so it doesn't land in spam. The agency side, the sub-account, funnels, calendars, automations, voice agents, chat agents — and then the part nobody teaches, which is how to actually find clients and charge them for it. I brought on the people who solve that specifically. Lead generation. Live site conversations. A guy whose team has helped over 5,000 HighLevel users generate north of $100 million tracked.

Free. On YouTube. Forever.

I've trained more than 10,000 students, and the single biggest thing separating the ones who build a business from the ones who don't isn't intelligence, budget, or timing. It's that they stopped collecting and started shipping. That's it. That's the whole gap.

So this issue is packed. The full setup video. Five Claude Code repos with the actual install commands. The free community with the snapshot, the custom GPTs, and the six-week course inside it. Two events where I'd genuinely love to shake your hand.

Grab what's useful. Then go build the thing.

👉 Join the free community (no card, no upsell wall): https://www.skool.com/automatedmarketer
👉 If you don't have HighLevel yet, come in through my link and the full bonus stack plus VIP access is on me: https://www.gohighlevel.com/nuno-tavares?fp_ref=automatedmarketer

And if you see me in a hallway in Tysons Corner this week — come say hi. Wish me luck up there.

🛠️ The Only GoHighLevel Setup Video You'll Ever Need

You bought HighLevel. You're running maybe 20% of it.

That's not a knock — it's math. The platform ships features faster than anyone can document them, so most people wire up a calendar, build one funnel, and never touch the other 80% they're paying for every month. Then they blame the tool.

So I recorded the whole thing. Start to finish. Zero cuts, nothing edited out, no "and then I'll skip ahead." Buying the domain. Setting up Google Workspace. DNS records. The agency side. The sub-account. Funnels, calendars, automations, voice agents, chat agents. Then how to actually get clients and charge for it.

So I recorded the whole thing. Start to finish. Zero cuts, nothing edited out, no "and then I'll skip ahead." Buying the domain. Setting up Google Workspace. DNS records. The agency side. The sub-account. Funnels, calendars, automations, voice agents, chat agents. Then how to actually get clients and charge for it.

It's one video. It's free. And it's broken into three parts so you only watch the part that's yours:

• Part 1 — the agency backend. Domain, Google Workspace, DNS, DKIM, DMARC, SPF, white label setup, SaaS configurator, plan pricing, rebilling markups, snapshots. This is the part everybody skips and then pays for later.

• Part 2 — the sub-account. The account you actually run business out of. Email servers, branded domains, client portal, calendars, pipelines, automations, AI agents. Built live.

• Part 3 — business owners, start here. If you're not reselling and you just want HighLevel running your own company, jump straight to this. Same build, no white label overhead.

By the end of it, here's what exists that didn't exist when I hit record:

• A live website and landing page, built by AI from a prompt, with the offer written first

• A thank-you page that gets people to actually show up to the appointment

• Two working calendars wired to Google Meet with confirmations and reminders

• A full appointment reminder and no-show nurture sequence, emails written

• A lead magnet, a form, an opt-in page, and the fulfillment workflow behind it

• Products and an order form that takes real money through Stripe

• A chat agent live on the site

• A voice agent that books appointments, transfers to a human when someone gets frustrated, and sits embedded right in the middle of the homepage

• Schema markup on every page so ChatGPT and Perplexity can actually find the business

That last one deserves 30 seconds. There's a button in the page editor under SEO and AI Search Optimization that writes your schema markup for you. Free. It reads the copy on your page and generates the whole JSON-LD block — organization type, FAQ schema, the works. SEO shops still quote $500 to $1,500 per site for that. Go press it on every page you've ever built.

But building the thing was never the hard part. Getting somebody to pay you for it is.

So I brought in people who solve that specific problem:

• Dakota from White Label Suite on finding leads — people search, company search, domain search, influencer search, and website visitor ID that pulls the name, phone, and email of people who hit your page and never opted in. He set up free credits for anyone watching.

• Omar from Knock Knock on catching the traffic you already have. It watches visitors on your site, scores intent, and when someone's hot it invites them to an audio or video call right there. You can jump in mid-conversation or let the AI run it. Their entry plan starts at $47/month and you can white label it from $150/month — sell two and it's paid for.

• Matt from HL Pro Tools on the part nobody wants to admit. His crew has helped over 5,000 HighLevel users generate north of $100 million tracked, and his advice for beginners was blunt: the skill you have to build is sales, way earlier than you want to. His three-word version of what separates people with thousands of accounts from people with three — they publish consistently. He opened up a $1 trial for anyone who watched.

• Plus my sales script pack — the actual scripts my team trains on — free in the video.

💡Pro Tip: Most people will watch this, build their own account, and stop there. That's the expensive mistake. Learned it the hard way — I built my own systems last for years while client work ate every hour, which meant I had no lead flow of my own and no proof I could point to. Build yours first. Then the exact same build becomes the product. A website plus automations is a $1,500 setup. Add voice AI, chat AI, review generation, and database reactivation and you're at $3,000 to $5,000. Custom builds start at $5,000 and go up — I've shipped $100,000 systems, but my first website 15 years ago was $400 and I was sweating asking for it. Now stack the recurring underneath: $97 to $297 a month for the CRM seat, plus rebilling at 1.5x on usage. Ten clients at $197 a month is $23,640 a year before you've built anything new. The build is the demo — the seat is the business.

One more thing worth saying out loud. If you don't have a HighLevel account yet, get it through my link. You get the trial, the full bonus stack, and VIP access to everything on my side at no cost. If you already have one, come into the free community anyway — the snapshot, the custom GPTs that write the offers and the emails and the bot prompts, the six-week beginner course, and the funnel brick system are all sitting in there, free, no card required.

I'd rather you build it than watch me build it. But the watching part is free, so start there.

👉 Watch the full setup: https://www.youtube.com/watch?v=EOCoGEUvao4
👉 Join the free community: https://www.skool.com/automatedmarketer
👉 Get HighLevel + the bonus stack: https://www.gohighlevel.com/nuno-tavares?fp_ref=automatedmarketer

 🤖5 Claude Code Repos I Wish I'd Had in Week One

Claude Code out of the box is strong. Then you use it for a month and the cracks show up.

It forgets everything between sessions. Its research is shallower than a dedicated research tool. It rereads the same files over and over and eats your usage doing it. It can't watch a video. And it writes code all day without ever actually shipping software.

Five open-source repos fix all five of those. Free. Most install in under two minutes. Install only the ones that match a problem you actually have — this isn't a checklist.

1. claude-mem — memory that survives the session

Every new session starts from zero. You re-paste context, write a handoff, re-explain the project. claude-mem captures what your agent does while you work, compresses it with AI, and injects the relevant part back into the next session. No prompting. Works across Claude Code, Codex, Gemini, OpenCode, and more.

Install: npx claude-mem install
Repo: https://github.com/thedotmack/claude-mem

During setup it asks which model should compress your memories. Haiku is cheapest, Opus is highest quality. I run Sonnet — Haiku's summaries weren't sharp enough for me and Opus is overkill for compression.

2. last30days — research that isn't six months stale

I was paying for a separate research model to get answers Claude's built-in search couldn't reach. This skill pulls Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web from the last 30 days and ranks by what people actually upvoted.

I ran the same query two ways side by side. Plain web search: a list of trending repos. With the skill invoked: four Reddit threads with 751 upvotes and 207 comments, six Hacker News stories, GitHub, and web pages, all cross-referenced. Different answer, better answer.

Install: /plugin marketplace add mvanhorn/last30days-skill then /plugin install last30days@last30days-skill
Repo: https://github.com/mvanhorn/last30days-skill

Reddit, Hacker News, and Polymarket work with zero configuration. X, TikTok, and a few others need API keys. And you have to invoke it by name — say "use last30days" or type /last30days or it won't fire.

3. ponytail — stop burning tokens on code nobody needed

This one makes your agent think like the laziest senior developer in the building. Before it writes anything it asks: does this need to exist? Does the standard library already do it? Is there a native browser feature? Then, and only then, does it write code.

You ask for a date picker. Your agent installs a library, writes a wrapper, adds a stylesheet, and opens a discussion about timezones. Ponytail writes <input type="date"> and moves on.

The benchmarks are real and published: about 54% less code (up to 94% where the agent over-builds), roughly 20% cheaper, 27% faster, measured across 12 feature tasks on an actual FastAPI and React repo. And it never simplifies away validation, error handling, security, or accessibility.

Install: /plugin marketplace add DietrichGebert/ponytail then /plugin install ponytail@ponytail
Repo: https://github.com/DietrichGebert/ponytail

4. claude-video — Claude can finally watch

Paste a YouTube link and Claude pulls a transcript that misses everything on screen. It has no idea what the thumbnail looked like, how the hook was cut, what was on the slide. For anyone working in video, that's most of the information gone.

/watch pulls captions first (free), downloads only what it needs, extracts frames, and reads every frame as an image alongside a timestamped transcript. Works on YouTube, Loom, TikTok, X, and local files.

Install: /plugin marketplace add bradautomates/claude-video then /plugin install watch@claude-video
Repo: https://github.com/bradautomates/claude-video

I've been recording Looms of my screen and handing them to Claude instead of writing instructions. It sees what I'm pointing at. The quality difference is not subtle. Fair warning — frame extraction is token-heavy, so tell it to take a balanced pass instead of grabbing every second.

5. agent-skills — an actual engineering team

Claude Code writes code. It does not ship software. No spec, no plan, no tests, no review, no definition of done. Fine for a snippet. Dangerous on a real build.

Addy Osmani's pack is 24 lifecycle skills plus 9 slash commands mapped to how software actually gets made — /spec, /plan, /build, /test, /review, /ship — and four specialist personas: code reviewer, test engineer, security auditor, and web performance auditor. 92,000 stars.

Install: /plugin marketplace add addyosmani/agent-skills then /plugin install agent-skills@addy-agent-skills
Repo: https://github.com/addyosmani/agent-skills

This is the same shape we teach inside VCI — PRD first, spec file, CLAUDE.md, then build. My own agent team has been running for months and it interviews me during planning before it writes a line. That interrogation is where the quality comes from.

Pro Tip: The obvious move is to install all five today. Don't. Install the one that matches the problem you cursed at this week, use it for five days, then add the next — stacking five unfamiliar systems at once means you can't tell which one broke your setup. Here's the part nobody does with these: two of them are a service, not a tool. Pair last30days with claude-video and you can produce a monthly Content Intelligence Report for a client — what their market is actually saying, ranked by real engagement, plus frame-by-frame teardowns of the three competitor ads that are working. Market research retainers run $500 to $2,000 a month and take a junior analyst 15 hours. This takes you two commands and about 90 minutes. Price it at $497, land eight clients, and that's $3,976 a month off work you were going to do for yourself anyway. The tools are free. The packaging is the business — sell the report, not the repo.

👉 Come build with us: https://vcinc.ai

 🧠 Agents Went Consumer, and the People Building Them Got Nervous

Meta shipped a personal agent called Muse this week and it went to number two on the US app charts. Not number two in productivity. Number two overall.

That's the part worth sitting with. Your clients' customers are now installing an agent on their phone the same way they installed a weather app.

Muse connects to your email, calendar, and apps and runs actions on a dedicated virtual machine instead of your computer. It books travel, sends emails, fills out forms, and keeps working after you close the app, coming back when it needs approval. It pulls context from Facebook, Instagram, and WhatsApp. Credentials go into secure storage so the agent uses them without seeing them, you pick which apps it touches, and you can opt out of your interactions training Meta's models.

The demo that landed hardest: someone connected it to their inbox and asked for an audit of their AI subscriptions. It came back with 20+ tools they were paying for — three image generators, two meeting-note tools, four video tools, two automation platforms — and still missed a few.

Here's why that matters to you and not just to consumers. An agent that can audit a subscription stack can audit a vendor list, a service contract, a recurring charge. The consumer version of "what am I paying for and why" is about to become normal. If your client retention depends on a client not looking closely at their invoice, that's now a countdown.

What else shipped:

• ChatGPT Images 2.5. The headline upgrade is consistency. Put a person in a suit and it actually still looks like that person. Poses, faces, hands, and object shapes hold across edits while the content inside them changes. There's also a new Sketch mode — draw a rough shape, hit check, prompt it, and it builds from your drawing. Available across ChatGPT, ChatGPT Work, and Codex on desktop, mobile, and web, plus two API variants: one more detailed and pricier, one faster and cheaper.

• DeepSeek V4.1 Flash. Scored 74.2% on the Deep SWE coding benchmark, which puts it level with GPT-6 Astra, Gemini 3.8 Flash, and Opus 5. At 27 cents per task, against $3.26 for GPT-6 and $8.75 for Fable 5. Worth a caveat though — the person running it side by side on visual generation said the output wasn't remotely in the same class as the top models despite the benchmark score. Benchmarks and real work are drifting apart.

• ChatGPT Work learns your writing voice. Connect Gmail, Drive, Slack, or SharePoint and it studies your actual emails, messages, and files — your phrases, your signoffs, your capitalization habits — then carries that into whatever it writes next. If you gave up on training it to sound like you, this is a different mechanism than the one that failed you.

• OpenAI's data agent. Connect Redshift, BigQuery, ClickHouse, Databricks, MongoDB, or Snowflake directly and ask questions, build visuals, and generate slide decks off your live data.

• Suno V6, trained only on licensed music. They cut deals with Warner Music Group and BMG and retrained from scratch on rights-cleared material. If you've avoided AI music in client work because of licensing exposure, that objection just got smaller. Google also dropped Lyria 3.5 inside Gemini, Flow, AI Studio, and Vids.

• Apple's keynote. iPhone 18 Pro and Pro Max with variable aperture, plus the iPhone Duo — a folding phone that opens into something close to an iPad mini. The sleeper is the Apple Watch: always-on audio intelligence with Live Rewind (double-press the crown, see the last 15 seconds of conversation as text) and Siri Recap summaries of conversations. The always-listening notetaker pin category just got absorbed into a watch people already wear.

• Odds and ends. Microsoft shipped MAI Image 2.6 with multi-reference editing. The Gemini desktop app finally landed on Windows. DaVinci Resolve 21.1 added an AI assistant you can wire Claude directly into for editing. And the movie about the week Sam Altman got fired and rehired drops Christmas Day, with Andrew Garfield playing him.

Now the heavy part, because it was the biggest story in AI this week and skipping it would be dishonest.

A pre-training researcher with three years at both OpenAI and Anthropic resigned publicly and said neither company is acting responsibly — that they're racing to self-improving superintelligence and gambling with everyone's lives. Anthropic's alignment science lead replied agreeing, and put his personal odds of AI killing all humans at greater than 10% within the next decade, adding that Anthropic does not yet have a plan to solve alignment for superintelligence and is not clearly on track to.

In the same week, OpenAI's chief scientist published a piece arguing that recursive self-improvement is coming, that the next few years bring capability jumps of equal or larger magnitude, that results get harder to interpret as systems get more capable, and that the defense against rogue AI is going to have to be aligned AI. Separately, OpenAI said it solved a 90-year-old Millennium Prize problem using an internal model significantly more capable than GPT-6 Astra — the model everybody currently thinks is the ceiling.

The obvious counter-argument is that this is positioning. Safety researchers are paid to imagine worst cases, and there are documented instances of creators being offered money to push doom narratives. Both things can be true. These are also serious researchers with careers to lose saying it under their own names.

I'm not going to tell you what to believe about that. What I'll say is that the useful posture isn't panic and it isn't dismissal. It's paying attention.

Pro Tip: The obvious move with Images 2.5 is to make prettier pictures. That's not the unlock. Consistency is. Character consistency is the reason you couldn't run AI images in real client campaigns — the person changed faces between the ad and the landing page and the email header, and it looked cheap. That's fixed now. Most agency owners have a client sitting on one decent brand photo shoot from 2022 and no budget for another one. A reshoot runs $1,500 to $5,000 and takes three weeks to schedule. You can take their existing photos and generate a month of consistent, on-brand creative in an afternoon. Package it as a creative refresh retainer at $750 to $1,000 a month, deliver 20 to 30 assets, and you've added recurring revenue to accounts you already own. Six existing clients at $750 is $4,500 a month with no new sales calls. The shoot was never the product — the consistency was.

Here's where this is going. Every one of these releases points the same direction: the agent stops waiting for your prompt. Muse works after you close the app. ChatGPT Work writes in your voice without being told how. The Apple Watch listens all day whether you asked it to or not. In 2024 you operated the tools. In 2026 you're going to configure them and check their work.

The people who win the next 18 months aren't the ones with the best prompts. They're the ones who built the systems the agents run inside.

Making Extra Cash Right From Your Home With a Little Bit of an Investment

Went down a rabbit hole this week watching a video about machines you can import from Asia and run out of a garage. Not software. Not agents. Actual machines that go clunk and make a physical thing somebody hands you money for.

I have no plans to buy any of these. I also could not stop watching.

• Glass bottle slumping kiln (~$1,000). Takes an empty wine bottle and flattens it into a cheeseboard that sells for $30 to $80. Your raw material is the recycling bin behind a winery. Melt their bottles, put their logo on it, sell it back to them for the gift shop. Genuinely brilliant.

• Preserved flower freeze dryer (~$1,900). Bride's bouquet dies four days after the wedding. She hands it to you, you freeze-dry it, encase it in resin, hand it back. Partner with the flower shops already doing the weddings and let them refer you.

• Boba pearl machine (~$1,300). Nearly every bubble tea shop is using frozen imported boba. Count the shops in your town, make it fresh, become their recurring supplier. There is almost certainly nobody else in your area pitching this.

• Chocolate 3D printer (~$1,000). Someone at a trade show printed chocolate high heels and sold out in ten minutes. I'm not saying that's a business. I'm saying I would have bought one.

• Press-on nail UV printer (~$1,700). Prints custom images onto press-on nails, mails them out. Bachelorette parties, salons, brands putting their logo on people's fingers.

• Maple syrup evaporator ($218). Two hundred eighteen dollars. Syrup wholesales at $50 to $70 a gallon. Useless in Phoenix, a licence to print money in Vermont.

• Down and insulation filling machine ($6,800). The kicker. A guy blowing organic cotton into pillows and bean bag chairs out of a garage, reportedly clearing $500,000 a year on this one machine.

Worth saying plainly, same as the guy who made the video did: buying a machine is not a business. None of these numbers are promises, and every one of them assumes you go out and find the customers first. The tortilla machine doesn't sell tortillas. You do.

Still. There's something refreshing about a business where the moat is that you own a thing nobody else in your zip code owns. We spend all day automating the intangible. Sometimes the play is a $218 evaporator and a maple tree.

💲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