How a team of three people and Claude took Nitrosend to 1,200 signups and 750,000 weekly email sends, one rough iPhone video at a time.
Melbourne, a home office. A founder props his iPhone on a shelf, clips on a decent mic, and talks straight to camera about a feature that he shipped that morning. No studio. No crew. Just him, the phone, and about ten minutes.
That rough clip is the raw material. Once George Hartley hands it to Claude and a freelance editor, it turns into a launch that pulls somewhere between 100,000 and a million impressions on X. He has run that same motion over and over since June. It has carried Nitrosend to 1,200 signups and roughly 750,000 weekly email sends with a team of three people and Claude.
I wanted to know exactly how he does it, because George clearly knows what he is doing. In fact, this is his third email company across 13 years of building. He co-founded SmartrMail, which generated over $350 million in extra sales for Shopify and BigCommerce stores before Relay Commerce acquired it in 2022. He co-founded Bluethumb, now Australia’s largest art marketplace. He mentors founders at Startmate, Australia’s answer to Y Combinator. So when he told me his launch spikes are the single thing that drives Nitrosend’s early growth, I asked him to break the whole sequence down.
What Nitrosend actually is
Nitrosend is AI-native email that runs inside your AI agent. For most of George’s users, that means inside Claude, Codex, or Cursor. He describes it as a headless replacement for Mailchimp, Klaviyo, Resend, and Instantly. It is the full email stack that you’d need to run and grow a brand, and he built it for agents rather than for people who click around a dashboard.
One number shows the bet is landing. Nitrosend has logged 23,000 user sessions since June. 94% of them never touched the app’s interface at all. Agents drove the product. George’s whole company is built on the assumption that his users will rarely, if ever, look at his software directly.
I’ve spent years launching data and AI products before their markets were obvious, for companies from SingleStore to IBM, so I am quick to recognize a founder who’s building ahead of the curve. George’s story is particularly worthwhile because he has paired that early bet with a launch engine that anyone can copy.
Launches as software releases
George doesn’t wait for a big annual moment to “do marketing.” He treats every meaningful thing that his team ships as its own launch event.
He started from zero and asked a simple question: what’s the easiest way to get to 100 users, then 1,000? He picked the two channels where his buyers already hang out, X and LinkedIn, looked at what actually earns attention there, and landed on producing short videos about what he’s building and releasing. Then he mapped it to his product roadmap. Every time the team could ship something interesting, they packaged it as a little launch.
Last week’s example was the agentic flow builder. It had existed since June. George’s team improved the design output enough that it produced a genuinely better result, so they made that the launch theme. The upgrade became the event.
So how does he turn a shipped feature into a spike that pulls a million impressions? Well, here’s the sequence George runs, and it is worth stealing.
Reproduce This Win: George’s lean launch sequence
- Write a punchy script with Claude, and train it on your own signal. George gets Claude’s help to draft the script. He feeds it his brand assets, plus what his customers are actually saying about the product. He mines his own inbox and his socials for the real language, then gets a few drafts and picks the sharpest one. The selling points come straight from his customers.
- Film rough and ready, straight to camera. An iPhone in the office and a decent mic. No production value, on purpose. The realness is the point.
- Build a reusable animation base once. For his first launch, George paid a US agency to produce the animation, and kept the working files. Every launch since then reuses that same base, so the polish becomes a one-time cost that carries across every future launch.
- Cut it together two ways, depending on budget. The cheap way is to have Claude edit the video with Remotion and Claude Fable, and generate any missing assets through an image model via the API. The better way is a freelance product editor in the USA who works fast off the existing animation base. Pick your lane by how much the moment matters.
- Distribute through a clipper bounty. George’s editor uses clipping platforms Clipur and Content Rewards, where you post a bounty for a target number of impressions. The output looks organic in the feed, which is exactly what earns attention on X and LinkedIn.
- Time it so your warm network sees it the second that it’s live. Most founders skip this step. George blasts his user base and email list the moment that the launch goes live, and he has Claude pre-draft one-to-one outbound to a hit list of about 100 warm people in his network. Immediate launch engagement is vital here as the algorithm looks for signals right away. The next most important part is follow-up where either you work manually by liking and responding to all comments and reposts on X and LinkedIn, or by running Claude (with a strict tone of voice and guardrail skills) to do it on your behalf afterwards.
Key Takeaway: The engine here is reuse. George built the expensive parts once (the animation base, the brand voice, the outbound list) and reuses them on every launch, so each new launch costs him a script and an afternoon. This reuse is how three people ship launches at the cadence of a much bigger team.
The results back this up. Those spikes have run between 100,000 and a million impressions each, pulled Nitrosend to 1,200 signups fast, converted a share of them to paying, and kicked off word of mouth. George told me his bigger paying customers have all arrived through that word of mouth in the last three weeks. He’s now easing off the frequent-launch pace and saving the motion for when the team has something real to ship.
Here’s why this works, in my language… George has built a Minimum Viable Growth System, which is the smallest repeatable engine that produces predictable pipeline. Most founders that I talk to reach for scale before they have that system, so they pour money into paid channels on top of an engine that doesn’t exist yet. George built the engine first, then let it compound. The order of those moves is the whole game. It is what I spend most of my time helping technical founders get right.
What didn’t work (so you can skip it)
I always ask founders what to avoid, because learning from these failures is what inevitably save my readers the most money. George ran two experiments alongside the launches. Both were duds this early.
- Remarketing ads flopped. He ran light-touch retargeting across Meta, the Google display network, and LinkedIn to people who’d already been pixeled. They didn’t work at all. He isn’t sure whether it was the creative or a lack of brand trust. Either way, he called it a waste of money for now.
- Cold Google search ads didn’t convert. He tested a batch of open keywords and got an acceptable cost per click. None of it turned into paying customers.
I agree with his read on both of these results… when your company is this young, there’s a trust gap that paid traffic can’t close yet. Borrowed attention from a launch beats bought attention when nobody knows your name.
I’ve lived the other side of this equation. At SingleStore, I generated 1,100 organic marketing-qualified leads at about $50 each without spending a dollar on paid acquisition. The trust that actually converts early comes from how useful you are in public, and that does not require a retargeting pixel following someone around the web. George arrived at the same conclusion from the opposite direction, when he tested paid channels first and watched them fall flat.
The AI-native operating layer underneath it all
Here’s what makes George’s cadence possible. His team is three humans and Claude. Claude is effectively handling the launch execution end to end: the scripts, the edits, the outbound drafts.
We got into something that I’ve been saying for a while. The real edge comes from building the elements of your brand into the AI, so it can retrieve them on demand. This is the core of the work that I do now in engineering growth for technical companies. A person can’t hold every case study, every proof point, and every customer quote in their head at the moment that they need it. An agent can. As George put it, humans run out of memory, patience, inspiration, and work ethic. Agents don’t.
He’s building that same idea into the product itself. When a brand onboards to Nitrosend, a one-click flow scans their website, pulls in their brand voice and design, and creates an AI memory for that brand. That memory stacks up learnings over time, and is specific to what you send and how your audience responds, so engagement should improve the longer you use it. He’s also betting hard that agents will discover him, which means generative engine optimization (GEO) and answer engine optimization matter more to him than classic search.
Key Takeaway: George’s launch speed comes from one place. He built a system, in his tools and in his product, where the AI carries the context so the humans only have to make the decisions.
Where his real edge comes from
When I asked George where his biggest growth breakthroughs over 13 years came from, he didn’t point to anything on the internet. He said the breakthroughs that truly moved the needle came from talking to other founders one-to-one, usually in person. In George’s perspective, the alpha on X is already gone by the time you read it.
His freelance editor is the perfect example. George found him through another startup, some Australians that he met in San Francisco because a mutual contact said the two teams should talk. One coffee led to the editor. The editor led to the animation agency. None of that was searchable.
I live on Koh Samui island, in Thailand. My business partner Jonathan is on a tropical island in the Philippines. So we know exactly how the “just have dinner with the right operators” advice can feel totally out of reach. Here’s what I’ve found: you can replicate that edge without being in the room. My newsletter and Jonathan’s podcast are how we meet and learn from sharp founders every week. This very conversation is how I found out about Clippur, which I’m now going to go try.
Key Takeaway: Build a reason to be in ongoing conversation with builders, and you create the same lucky introductions that used to require a plane ticket.
Two more things George would tell a technical founder who is starting an AI-native company today:
- Have the hard conversations with your co-founders up front. The biggest killer that he sees in new startups is co-founders blowing up with each other. Sit down early and work through how it could fall apart, then get the agreements in place before you need them.
- Build something that you’re the user of, or close enough to stand in their shoes. When you understand the buyer that well, you shortcut the early product and marketing decisions. You make fewer mistakes. Every mistake that you skip is a bit more time that you keep.
The last point lands hard for me, because it is the same principle behind how I work. I sit in one seat as both the technical translator and the growth lead, which is what I call the Technical Marketing Hybrid.
When you understand the product and the buyer at that depth, you don’t need to trial-and-error so much because you have experience that informs what will work.
George has that same hybrid advantage. He’s built email startups three times now, so he knows this buyer in his bones and can move at a speed that the rest of us have to engineer our way toward.
Want to see what AI-native email feels like? You can start Nitrosend free and let your agent drive it at: nitrosend.com.
P.S. George also mentioned he buys sleepy little websites on Flippa now and then, the kind that a founder built and then stopped loving, to acquire an audience and a live test bed for new ideas. His very first email company, Sendicate, came that way. It taught him how to build smarter mail in the first place. I love that. There’s more than one way to avoid starting from scratch, and the founders who know the ins and outs of bringing things to market can clearly see those gems everywhere. Buying that momentum is the same instinct behind everything George builds: he finds the momentum that already exists, then engineers on top of it. What’s not to love?
