ESD
Hi there, welcome back to ESD Newsletter.
Here’s what’s inside today:
🛡️ The 8 Moats for SaaS in the AI Era
📈 ESD Trending: 5 hot startup stories
✉️ How to cold email an investor (by Michael Seibel)
Let’s get started!
The 8 Moats for SaaS in the AI Era
Gokul Rajaram
Gokul Rajaram helped build Google, Facebook, and Square. Now he invests, using one test on every software company: 8 questions he calls the 8 moats. Score four or more and you survive AI. Score one or less and you're quietly dying.
Why listen to Gokul
He was an early product leader at Google, then Facebook, then Square, then DoorDash. Four of the most important companies of the era, before becoming a top investor. At Square alone, he helped grow it from one product to eleven, each doing over $50M in revenue. When he talks about what makes a company durable, it's earned.
The context: public markets have decided every software company is going to zero. The logic is that code is becoming free to generate, so any app can be cloned overnight. Gokul thinks that's a massive overreaction. Not every software company is equal, and the 8 moats are how you tell the survivors from the dead.
Moat 1: Data
Not just having data. Having proprietary data nobody else can access. His example is Spotify's Discover, built on a decade of listening behavior across hundreds of billions of plays. You can clone the app. You can't clone the ten years of data that makes it work.
Moat 2: Workflow
The deeper you're embedded in how a company actually runs, the harder you are to remove. A tool that runs your entire business, like an ERP, is buried deep. A light add-on that lives on the surface is not. Weak on its own, powerful the deeper it goes.
Moat 3: Regulatory
Licenses, capital requirements, multi-year contracts. The boring barriers competitors can't just code around. His example is Coinbase: state-by-state money transmission licenses that take years and millions to acquire. You can't vibe-code your way past a regulator.
Moat 4: Distribution
A proprietary channel to customers that rivals can't buy their way into. His example is Intuit. They trained an entire network of accountants to work only in QuickBooks. Try to use a competitor and your own accountant tells you no. The product can be copied. The channel can't.
Moat 5: Ecosystem
When thousands of third parties build on top of you, they become your defense. His example is Shopify. You can clone the storefront easily. You can't clone the hundreds of thousands of developers and apps built around it. Every merchant relies on five or six of those add-ons. That's the moat.
Moat 6: Network
Value that grows with every new participant, and can't be faked. His example is DoorDash. AI can copy the app. It can't copy the marketplace density, the driver liquidity, the years of reputation and history. More users make it structurally better. That's a real network effect.
Moat 7: Physical infrastructure
Anything made of atoms, not just code, is hard to displace. Warehouses, hardware, real-world assets. A pure software company simply doesn't have this. He notes humanoid robots might chip at it eventually, but that's years away.
Moat 8: Scale
When your size makes your costs so low that nobody can match them. His examples are Amazon and TSMC, where the scale itself is the barrier. Worth noting: he says software used to be a scale game, but now that anyone can build cheaply, only companies with real physical scale, like the hyperscalers, truly own this one.
How to use it
Score any company one point per moat. Four or more, it's durable. Two or three is shaky. One or less means it needs to build fast or it's in trouble. He ran the test live: Atlassian scored a three, and he argued it's being oversold. Monday scored a one, and he said its lower price might be fair.
The two that matter most early on
Most of the eight are impossible to prove when a company is young. Distribution, ecosystem, network, all too early to call. So it comes down to two: a data asset that compounds with every interaction, and a workflow so deeply embedded it can't be ripped out. Those are the ones to bet on before the others exist.
ESD Trending: 5 hot startup stories
Mallow launched a €120 speaker for kids aged 4 to 10. No screen. A figurine sits on top, the kid talks, and the speaker runs quizzes and audio games. Co-founder Cédric O used to be France’s digital minister and helped start Mistral. They raised €11M from Balderton, Daphni, and Motier, tested it with 300 kids, and partnered with game publisher Nathan. Most kids’ AI so far is another app. This one is a box on the table.
They rented a church in San Francisco, put app icons in a coffin, and invited Vercel, Stripe, and OpenAI. The product is agents that live in iMessage and WhatsApp instead of a new app. They say 40,000 developers signed up and revenue grew 10x in four months. The funeral is a stunt. The bet isn’t: people may stop downloading your app if an agent can just text them.
Announced at DevDay on September 29. Each Dot runs on GPT-6 Astra, gets its own cloud computer and browser, and can connect to more than 4,000 apps. You reach it in ChatGPT, Slack, Teams, or on a call. It comes with Pro and Business Premium. Free and Plus don’t get one yet. Meta’s Muse already showed people will install an agent. OpenAI just put one inside the product they already pay for.
World Labs builds models that understand 3D space, not just text. Founded in 2024, it raised $1B and was valued at $5B just months ago.AMD is paying $8.2B in stock. Li becomes AMD’s chief scientist and reports to Lisa Su. The deal should close by year-end if regulators sign off. A chip company buying the lab, so the next chips get designed around the model.
Not a new funding round. A $300M employee share sale, led by Wellington and T. Rowe Price, at twice the February price. The company says its voice agents now handle 15 million conversations a week, up 3x since February: refunds, insurance renewals, appointments. Staff can sell. That’s how these companies are keeping people from leaving.
How to Cold Email an Investor (by Michael Seibel)
Michael Seibel (ex YC CEO) gets a flood of cold emails and replies to most of them. Here's what actually gets a response. The one goal is simple: get a reply, not a meeting, not a check, just the start of a conversation.
The do's:
Keep it to 60 seconds. If it takes 2-5 minutes to read, he saves it for later, and later rarely comes. Write it, read it out loud, and cut whatever drags.
Lead with facts, not a story. What problem you solve, your solution, any launch or growth, market size, your co-founders, whether you can build, and one thing you believe about the space that others don't.
Cut the jargon. A customer knows your industry's terms. An investor might not. Write like you're explaining it to a friend outside your field.
Email from your company address, with your name in it. Odd personal addresses feel off, and investor tools only pull up your info if you use your real one. A deck is optional, but if you attach one, use the standard format (Google "Airbnb pitch deck"). And track your opens.
The don'ts:
Don't ask for a meeting right away. Say something interesting and the investor will come to you. Push for a meeting first and you feel like someone twisting their arm.
Don't fire off fast follow-ups. If you're tracking opens, you know they saw it. They've already decided to reply now or later. Piling on more emails won't change that.
Don't forget to say what you actually do. The most common mistake: "I'm in town, let's grab an hour so I can tell you what I'm building." Wrong order. Put the interesting thing up front and make them chase the details. That's your leverage.
The real lesson: a cold email isn't a pitch, it's bait. Say enough to make an investor curious, then let them come to you asking for more. Short, specific, and factual beats long and polished every time.
That's all for today 🫡
See you next time,
Komron & the ESD Team
PS: We're still testing the format, one deep story, five quick hits, and one practical tip per edition. Too much? Too little? Reply and let us know what you'd change. I'll personally read every one. 🙏🏻