10x the context. Half the time.

Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.

Try Wispr Flow free

😲We Gave Away 12 Hours and Charged Nothing

When I was coming up, there was no one to ask.

No roadmap. No "here's step one." No room full of people a little further down the road who'd answer a dumb question without making you feel small. Just me, a screen, and that sick feeling of not knowing what to do or where to even start. If you've felt it — the overwhelm, the second-guessing, the hours torched on something that should've taken ten minutes — you know it doesn't just slow you down. It makes you want to quit.

I didn't have anyone in my corner back then. So I decided the people coming up after me would.

That's the real reason behind last weekend. We wrapped day three of the Build-A-Thon, and when I added it all up, we'd recorded close to 12 hours of video. Twelve hours. Free. No gate, no "enter your card to unlock module two," no dripfeed.

I know exactly what most of this industry would've done with that — chopped it into a $1,997 course, buried it behind a webinar funnel, made you sit through a 90-minute pitch to reach the part that actually helps. You've sat through that. So have I. It's why I built this the opposite way.

Here's what's on YouTube right now, forever, that you can watch tonight:

• The full Build-A-Thon — every session, start to finish, free to register and watch.

• The real builds: standing up agency sub-accounts with AI, and wiring the systems business owners use to get running and stop bleeding money.

• Doesn't matter which seat you're in — agency owner setting up client accounts, or a business owner just trying to get off the ground. It's all in there.

• Chaptered versions are coming next, so you can jump straight to the part you need.

We did it for two reasons.

One: so the people who actually need help finally have it. No excuse left. If you can't set up and run a business inside HighLevel after watching, it won't be because someone hid the good part.

Two — and I'll be straight with you, because I hate when people aren't. When you start your HighLevel under my affiliate link, GoHighLevel pays me a commission. Costs you nothing extra — same price either way. The difference is what I do with it: I don't take it and vanish. I hand it right back as support. Start fresh under me, or move your existing account over, and you get the top tier of the paid community included. Already locked in under another HighLevel affiliate? No problem — you can still grab the community on its own. Nobody gets shut out.

Because that's the part I never had. Not the software. The people. So the community is where it actually lives now — free if you want to learn it yourself, or $47/month to go VIP with hot-seat support and the group calls I'm personally on. Come in through my affiliate link, and that top tier's on the house.

The lesson, if there is one: the thing that nearly broke you coming up is usually the exact thing you're meant to build for someone else. I didn't have help, so I became it. Whatever you wish someone had handed you — that's your offer. Go build it.

Come hang out. 👇

🚀Point an Agent at a Lead and Walk Away

You know that automation you built with eleven steps, four if/else branches, and a wait timer you've adjusted six times? The one that breaks every time a lead does something you didn't predict?

You don't need it anymore.

HighLevel just dropped an AI Agent action inside Workflows (still in beta, so read to the end before you turn it loose). Here's the play: a trigger fires — form submission, no-show, Instagram comment — and instead of routing the lead through a step-by-step maze you built by hand, you hand the whole thing to an agent. One prompt. The agent figures out the steps and does the work.

Not "AI writes a reply." The agent actually does things in your CRM

We tested it live on a real website form. Someone submits, and with zero human touch the agent:

• Created the contact and pulled the existing CRM record

• Checked tags and skipped enrichment if they were already a customer or booked

• Created an opportunity in the right pipeline at the right stage

• Generated a one-time booking link

• Sent a personalized email and SMS — signed off and on-brand

• Added tracking tags and notes so the next workflow knows what happened

All of it. By itself. In about 15 seconds.

And it's not locked to forms. HighLevel pre-loaded a stack of agent templates you can fire off the same way:

• Form lead follow-up and no-show appointment recovery

• Stale deal nudge — pokes opportunities going cold

• Research/enrichment on new contacts automatically

• Instagram & Facebook comment-to-DM — someone comments "checklist," they get the DM with your link

• Quiz lead scoring + routing and call transcript summary with action items

• Lead pipeline tracker that uses memory — follows a lead through stages with periodic check-ins

You can also wire the agent into outside tools — Notion, ClickUp, contact actions, webhooks, even MCP connections — and pick which model runs the job. So this isn't a toy. You can get genuinely advanced with it.

💡Pro Tip — let Claude write the prompt, not you. Most people will open the default template, see all those yellow highlighted fields, and start fighting them one by one. Don't. The default prompt is decent but generic. Copy the whole thing, drop it into Claude (or whatever LLM already knows your brand — mine has my full brand profile, so it writes like me), and say "rewrite this for my business, ask me anything you need." It hands you back a fully personalized prompt. Paste it in, swap the calendar link, done. You go from "editing a template" to "deploying an agent that sounds like you" in about two minutes.

Now the part nobody tells you about beta AI features: the cost. This is a premium action, so it bills you. People panic about that. So we tested and checked the wallet — that entire form-to-booking sequence cost two cents. Mostly the outbound text. If you're on the AI Unlimited plan, bot charges basically disappear and you're left with pennies of usage. Test one live, watch your wallet, then scale it. You'll sleep better.

Here's why this matters more than the average HighLevel update: the work didn't get easier, it got delegated. You set the personality once, and the agent handles the messy middle — the part where leads go cold because you were asleep or slammed. Do it for yourself and your follow-up gets sharper. Do it for clients (or your white-label), and you're charging for a personalized concierge you set up in an afternoon and never touch again.

That's the build for this week: pick your leakiest automation — the one that loses leads in the gaps — and rebuild it as a single agent. Start with form follow-up. Test it once. Check the two cents. Then go fix the next one.

 🤖 Hermes Install: The 24/7 Agent You Talk To From Your Phone

Claude Code is incredible — until your laptop goes to sleep.

That's the wall everyone hits. You build a beautiful agent, you set a cron job, and then you close the lid and... it dies. Claude Code lives on your machine. No always-on. No Telegram out of the box. Great for building, not built to run the night shift.

So this build fixes exactly that. We installed Hermes — an open-source agentic system (think same family as OpenClaw and Paperclip, just lighter and, honestly, more stable in my testing) — on a rented server that never sleeps. Claude Code does the building. Hermes does the running. You text it on Telegram from your phone.

Here's the shape of the build (full step-by-step is in the SOP below):

• Rent the computer. A Hostinger VPS on the KVM 2 plan — 2 cores, 8GB RAM, 100GB NVMe. Plenty. Run Ubuntu 24.04 as the OS.

• Install into the root folder, not the one-click button. Costs you two extra minutes now, but it means later you can run a whole army of Hermes agents — one for personal, one per business — each with its own job, instead of cramming everything into one.

• Let Claude Code do the heavy lifting. Open the folder in VS Code, point Claude Code at it, and have it write the CLAUDE.md, build the folder structure, and manage the install with you. When the terminal gets finicky (and it will), Claude talks you through it.

• Pick the brain. You can't run the agent on your Claude subscription — but a $20 ChatGPT/OpenAI Codex sub works fine. We ran GPT-5.5. (If the device code comes up grayed out, flip on "device code authorization for Codex" in your ChatGPT security settings, then retry.)

• Wire up Telegram. Spin up a bot with /newbot, grab the token, and now you're talking to your agent from your pocket.

• Back it up to GitHub. A private repo, secrets excluded, plus a daily cron job so it commits itself every night. If the whole thing ever blows up, you reinstall, point it at the repo, and it rebuilds itself.

Then the payoff: a web dashboard with a Kanban board, installable skills, sessions, personalities — the works. I didn't build that by hand. I asked Claude Code to connect it, and it did.

💡Pro Tip — treat it like a new hire, not a toy. This is the step everyone skips and regrets. An always-on autonomous agent can hallucinate, wander, and torch tokens while you sleep. So before you hand it anything: give it its own credentials. Its own email (a Google Workspace inbox, not yours). Its own API keys per tool, never your master keys. And the second you connect Telegram, set an allow-list with your user ID (grab it from @userinfobot) so you — and only you — can talk to it. You'd never give a new employee your personal logins on day one. Same rule here.

One more thing I do on every big install like this: I build a second brain first. An Obsidian vault stuffed with the official docs, transcripts, and my own SOPs, indexed so Claude Code can pull from it the moment I get stuck. Hermes, OpenClaw, Claude Code, Co-Work — I keep one for each. It's the only way I keep up with the weekly chaos of AI updates without losing my mind.

The build for this issue is simple to say and a little scary to do: stand up one Hermes agent, give it one job — checking a single inbox, watching one feed, running one daily report — and let it run overnight. Tomorrow morning, see what it did without you.

I wrote the whole thing up so you're not guessing.

👉 Grab the Hermes Agent Install SOP — the exact build, in order, with the security gates baked in.

💰Claude 4.8, a Near-Trillion-Dollar Anthropic, and Video From a Doodle

30-plus things dropped this week. I'm not going to pick five and pretend the rest didn't happen — here's all of it, with a take on the ones that earn your time.

Start with the obvious. Claude Opus 4.8 shipped. If you're hoping it feels like a different brain, don't — Anthropic literally calls it "modest." The real upgrade is honesty: it flags when it's unsure instead of bluffing. Sounds boring until you've got an agent loose in your CRM, and then it's the whole game.

Same drop, Claude Code got dynamic workflows — it splits a big job across sub-agents that work in parallel and argue with each other until they agree, then hands you one answer. Your messy multi-step builds stop dying halfway through.

And the number nobody can actually process: Anthropic raised $65B at a ~$965B valuation. Most valuable startup ever, past OpenAI. The company behind your stack is now a near-trillion-dollar bet. Build like it's not going anywhere.

The rest of the week, fast:

Microsoft went on a tear

• MAI Image 2.5 — now the #3 image model on the planet, and it nails the two things every other model botches: readable text and product/brand mockups. One-line prompt spit out a full event flyer.

• 365 Copilot redesign — bigger prompt box, inline formatting, and it now pulls from your emails, files, and meetings to answer, dropping charts right inline.

• Perplexity Computer is now inside Word, Excel, PowerPoint, and Outlook for the heavier multi-step stuff Copilot won't touch.

• Microsoft Build is next week — so all of this is just the appetizer.

Agents leveled up

• Hermes (yeah, the one we installed up top) has a self-improving loop — builds new skills from past jobs and keeps memory between sessions. Your agent stops starting from zero every time.

Creative tools that change your content costs

• ElevenLabs Music V2 — full songs, trained on licensed data, cleared for commercial use. No lawsuit roulette.

• ElevenLabs Dubbing V2 — dubs your existing videos into other languages in your voice, emotion, and expressions. 30 min free. Your back catalog just went multilingual.

• Leonardo AI image-to-3D — turn a flat image into a rotatable 3D model. Product spins, game props, ecommerce.

• Google Gemini Omni — draw a route on a map screenshot, get first-person driving video. Sketch a flight path, get drone footage. Establishing shots without owning a drone.

Stuff that'll hit creators

• YouTube is moving AI labels front and center — under the player on long-form, overlaid on Shorts.

• And it's adding automatic AI detection. Right now you self-report (nobody does). Soon YouTube flags photorealistic AI for you. Matters if your intros are AI and the rest isn't.

The "where's this going" file

• The Pope put out one of his rare official letters — on AI — comparing it to nuclear weapons that need "disarming," with an Anthropic co-founder beside him. The co-founder admitted every lab feels pressure to ship that fights doing the right thing, and that we need outside critics with no skin in the game.

• Sam Altman walked back the jobs-apocalypse talk — says AI hasn't gutted entry-level work like he feared. (Convenient, with an IPO coming.)

• Jensen Huang called the "we laid people off because of AI" excuse lazy and irresponsible. Demis Hassabis said the same. The reality: bloated post-COVID companies cutting overhead and blaming the robot — and investors aren't buying it.

Quick hits

• Erin Brockovich (yes, that one) launched a crowdsourced map of AI data centers near you — built, building, or proposed. brockovichdatacenter.com.

• Apple quietly added a genai.apple.com subdomain right before WWDC. Something's coming.

• And from the future-is-dumb department: a $118 collar in China claims to translate your dog at 95% accuracy, and AI robot barbers are scanning skulls and cutting hair for under a buck. (Both get the full treatment in Cool Shit — they earned it.)

💡Pro Tip: Don't adopt this whole list. Pick the one tool that kills your biggest recurring cost this month. Re-recording content for other markets? Dubbing V2. Paying a designer per flyer? MAI Image. Lock one in, get it working, then look at the next. The people who lose this race are the ones chasing all of it at once.

The thread through all of it — the layoffs, the IPO spin, the Pope, the flood of tools — is that AI isn't quietly deleting operators. It's stacking the deck for the ones who use it. You don't need to fear the list. You need to pick from it.

😲 Anthropic ran its first big European dev conference — and the agent news is the real story

Code with Claude London, May 19–20. Two days, split clean: enterprise on day one, founders and solo builders on day two. What's worth knowing:

• Self-hosted sandboxes — run Claude agents on your own infrastructure instead of someone else's servers. Same instinct as the Hermes VPS build up top: keep your keys, memory, and client data on hardware you control.

• MCP tunnels — let agents reach your internal systems without exposing anything to the public internet. For anyone wiring agents into a real business, that's the security piece that's been missing.

• A built-in frontier memory system for managed agents — they hold context, organize info like a file system, and get sharper from past tasks. Memory was the weak spot in every agent. This is Anthropic closing it at the source.

• And the line everyone's repeating: a Stripe team converted 50,000 lines of Scala to Java in four days with Claude. Four days. That's not "AI helps a bit" — that's a quarter of work gone in a long weekend.

Day two was all builders: prototype-to-revenue talks, giving agents memory, writing evals, stacking multi-agent systems, shipping working software with no traditional eng team. Recordings are up on Claude's London events page if you want to see what you missed.

Why it matters for you: the whole event circled the same theme running through this entire issue — agents that remember, run securely, and do real work. The tooling to run a business on agents isn't coming. It's already shipping.

💲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

🛠️ 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