Evan Tarver (@evan_tarv) header
Evan Tarver (@evan_tarv) profile picture

Building a one-person AI company @ https://teapawt.org/t.co/AhUnYsBjyX Previously: Founder @ Selling Signals (acquired) Sharing what I learn weekly👇

📍LA | ATX
Joined September 2011
Photo by Evan Tarver (@evan_tarv)Photo by Evan Tarver (@evan_tarv)Photo by Evan Tarver (@evan_tarv)Photo by Evan Tarver (@evan_tarv)Photo by Evan Tarver (@evan_tarv)Photo by Evan Tarver (@evan_tarv)Photo by Evan Tarver (@evan_tarv)Photo by Evan Tarver (@evan_tarv)Photo by Evan Tarver (@evan_tarv)Photo by Evan Tarver (@evan_tarv)
I've found it's a bit of a shitshow at any level
It's true. Startups are basically a series of disasters. Although YC's official motto is "Make something people want," we had two unofficial ones as well: "Never a dull moment" and "You can't make this stuff up."
4
Building SaaS got cheaper, but some products are still hard to copy. Here’s the framework I use to think about my business, from easiest to copy to hardest: 1. A tool your customer could build in a weekend with little expertise. Tools that are easy to understand, quick to build, and low risk are the first to go. 2. A product built on your expertise that you don't improve on. Whatever expertise you bake into your product as a baseline can be easily copied and used to leapfrog you. 3. A product that you actively use every week and keep improving. Copycats end up copying an outdated version while you’re working on the next release. 4. A product plus a service on top that improve together. The service helps you dogfood your product and is the sweaty part nobody wants to copy. I’m betting Flyletter, my AI newsletter tool, on number four. I use it to run a done-for-you newsletter service for other founders, which lets me deliver higher-quality work faster. In return, every edge case, quality bar, and learning from the client work goes back into improving the Flyletter product. If you do it right, copycats are just chasing an outdated snapshot. The moat is a product built on your expertise that you use every week and keep improving.
1
29
Elon Musk took down Tesla's wall of patents in 2014 and said he wouldn't sue anyone using its tech in good faith. Tesla wasn't worried about people copying last year's car. It was too busy building the next one. SaaS founders worried about AI copycats should think the same way.
11
I got absolutely BODIED on Threads. I still think everyone's wrong, though. Here's how I think about writing with AI the right way, to keep it authentically yours: 1. Connect Notion to Claude as an idea repository, and save your ideas as you have them. 2. Start from a voice note. Talk out your idea so the AI has your personal context, perspective, and experiences. 3. Create a beat-by-beat outline that reorders your voice note into a cohesive bullet list and acts as the scaffolding of your piece. 4. Have AI write the draft by matching the language of your voice note, ideally with access to brand context and live writing samples. 5. Hand-edit it until you're satisfied with the results. Done this way, I don't see how AI is any different from an elite ghostwriter. And I'll use AI every time if it's the difference between being able to share my ideas or not.
2
1
18
People complain about AI on social media apps that use AI to decide who to show it to
1
14
lol people HATE AI on Threads
1
8
When people write with AI, they check for tells, like em dashes and "it's not X, it's Y" giveaways. What they don't check for is voice flattening. When AI writes, it often reverts to the mean. It tries to find a middle road, so the piece comes out bland in both style and substance. That's why AI writing still doesn't sound human, even with every tell stripped out. So whenever you write with AI, check for both. And for the love of God, try to add back some soul.
1
11
Remember the metaverse?
1
10
If we all hate AI writing we also hate ghostwriters…right?
1
17
Honestly I was afraid of posting more and putting myself out there and being seen as "cringe," but the only thing that's happened is I've met some really cool people and had some wonderful conversations.
1
10
I'm a writer, and writing has always been how I figure out what I think. Lately, talking does it faster. When you explain an idea out loud, you find out quickly how well you actually understand it and where the gaps are. So when it's time to write, I flesh out the idea in a voice note to Claude first. The goal is to get my full thought out, however messy. Then I reason through the gaps and work with Claude as a thought partner to refine the idea. The next time you have an idea worth writing about, record yourself explaining it before you ever open a doc.
9
People love to hate AI writing. For a novel or poetry, I get it. But a consistent newsletter is one of the best ways for a founder or creator to build authority and share what they know with people who'd actually get value from it. Which, dare I say, makes the world a little better of a place. The problem is most of us don't have the time to write one every week, so the realistic options look like this: 1. Don't do it at all. 2. Start, then fall off the wagon. 3. Get your ideas out consistently with AI's help. I'll take option 3 every time if it's the difference between sharing my ideas or not. Just make sure when you do that you publish what's authentically yours. People are good at sniffing out phonies, and the whole point is to share your hard-won experience. So I start by talking my idea out in a voice note. It keeps the ideas mine, pulls out my first-person perspective, and leaves AI to help me shape my thoughts rather than think for me.
1
12
I don't have time to write a newsletter. So I ship mine in an hour a week. A single issue used to take me a day or two, most of it spent staring at a blank page. Now I run the whole thing from one Claude chat, with Flyletter (my AI newsletter writing tool) and Beehiiv plugged in. It takes six steps: 1. Set up your brand voice. Give Claude your brand context, your audience and your best writing samples so it knows how you sound. I use Flyletter for this. 2. Talk the idea out. Record a voice note to Claude and get your full thought out, however messy. This week's issue started on a dog walk. 3. Approve an outline. Have Claude turn the voice note into a beat-by-beat outline, and push back until you'd sign off on every point. 4. Write the draft. Let AI write from the approved outline, using your voice note for the style. Then run a separate pass that catches AI tells. 5. Edit by hand. Add back some soul and make every line something you'd stand behind before it goes out. 6. Ship with the @beehiiv MCP. Have Claude upload the draft into Beehiiv using your newsletter template, then give it a final check and schedule it. Do this and you can ship a high-quality newsletter in about an hour a week. Broke down the entire workflow in my newsletter this week if you're into that sort of thing.
4
2
61
I launched my product, then immediately left for two weeks in Europe (incredible timing). I still took sales calls all throughout the Balkans, but the itch to build never let up. Mid-trip, I posted this: "I'm in Croatia and all I want to do is Claude Code. It's a problem." But endlessly building without real market feedback is a trap. Every feature comes from your own assumptions instead of what people actually need. So I stuck to marketing. I spent mid-June through August on LinkedIn outreach, and the first users it brought in told me what my product was missing. To get that feedback fast, do three things: 1. Use the channel that reaches your people fastest. For me, that's LinkedIn outreach, because you can test a message and quickly see whether it lands. 2. Make the trade explicit up front. Offer an extended free trial in exchange for an honest feedback call, because revenue can wait. 3. Record every call, and let AI find the patterns. I record in Granola, and a weekly Claude task turns the transcripts into a report on what keeps coming up. Before you build another feature, go get real market feedback.
5
1
100
More like daily.
Running a business is just realizing how dumb you were 6 months ago every 6 months
24
Listening to the @theallinpod guys talk about the election is so funny. They’re like oh it looks like voters really cared about affordability and pocketbook issues like no shit dude people just want to be able to afford to live lol
26
People ask me my political affiliation and at this point I just say whatever @TimJDillon is
15
Anyone with Claude Code could copy my product. I've already seen people try. But none of them can copy what makes it actually work. Everyone says AI killed SaaS, but purpose-built products still matter. They pack years of hard-won expertise into a tool for one specific job. Someone could copy your code, but they can't copy the years of experience behind it. That matters for two reasons. First, a copycat doesn't understand the job your product does or the people it's for as well as you. They'll always be a release behind, copying last month's features with last month's messaging. Second, building around your own expertise means your personal brand doubles as the product's marketing. You get to tell the story of why you built it, and because you're the first user, you can dogfood it in public and become your own case study. Launching a product got easier, but building a real business around one is still hard. To find a product only you could build, ask yourself three questions: 1. Has someone already paid you for your expertise? A salary, a client, or a contract all count. 2. Can you package that expertise into a workflow? If the steps are repeatable, a product can follow them. 3. Could someone without your expertise build a close-enough version in an afternoon with Claude? If so, keep looking (or keep it as a weekend project). That's how I started Flyletter, my AI newsletter writing tool. I built it for myself using everything I'd learned from years of running newsletters and scaling content businesses, and my own newsletter doubles as its marketing. Build the product only you could build, and tell the story only you can tell.
10
1
5
109
If all else fails, treating every situation as a chance to meditate usually does the trick
My life improved significantly when I looked for evidence it was rigged in my favor. Stuck in traffic? Extra time on that audiobook. Rude stranger? Extra chance to practice empathy. Long line at the store? Extra chance for a moment of gratitude. Every circumstance that I normally viewed as "negative" became an opportunity for progress. And as a result, every area of my life improved.
1
1
1
20
Among my friends we call it "scoring one for the home team"
Underrated life skill: Being happy when your friends win bigger than you.
1
2
23