Mark Barbir

One project. Everyone’s context. Projects in @earmark_ai are now multiplayer! Invite teammates and bring together the meetings happening across your team, creating a shared history that keeps everyone aligned, whether they attended or not. ⦿ Combine insights from meetings attended by different teammates ⦿ Keep decisions, context, and next steps accessible to everyone ⦿ Ask Earmark about the latest status, past decisions, or what needs attention next tryearmark.com

If we're going to give computers intelligence, maybe we should just let them use it.

How @earmark_ai spreads. 🦠

We gave computers intelligence and then asked them to take meeting notes. 🥱 We’ve spent 30 years upgrading the notebook. Paper became a PalmPilot, then an app on our phones. Meeting software recorded the conversation. Now AI listens, summarizes it, and tells us what to do next. Every step has been useful. But underneath all that progress is the same assumption: we’re capturing this now so someone can do the work later. But not anymore with @earmark_ai .

Stop looking for advice. That’s the advice. I’m a Gen X founder who spent years refusing to make social content. Two reasons. First, my generation is allergic to anything that smells manufactured, and most social media has a content-farm ick to it that’s hard to shake. Second, I’ve seen myself on camera. The ums, the ya knows, the restarts, the mumbling. Not a great look. Meanwhile, my silence was actively hurting my company. :) Repetition got me over it. Even someone with my communication challenges has something to contribute. @hectorserrranoo on @lennysan's podcast illustrated the challenges are universal. He’s worked with famous creators. They all stutter. They all refilm. One puts a pen in his mouth between takes to get the words out. Hector’s own viral video took three thirty-minute cuts to become fifteen clean ones. So the ums I was worried about? They have them too. You just never see them, they turn the camera on anyway. We don’t have a skills gap. We have a willingness gap. The only fix is reps. 💪

Every AI founder has had the same bad dream. You wake up and one of the labs has shipped your product. @magicpatterns lived it. In April, @AnthropicAI launched Claude Design, and the announcement got something like 48 million views. Alex from Magic Patterns sat down with @daltonc at Standard Capital to talk through what happened next. It's worth 17 minutes of your time. I think about this a lot with @earmark_ai. Any of the labs could ship a meeting assistant on a random Tuesday. The part that stuck with me was the point that you aren't competing with the whole company. You're competing with a team of maybe five to ten people. A PM, a designer, a few engineers. Beat that team. I've run those teams and know what the energy feels like from the inside. There's a roadmap, a planning cycle, folks vying for promotion, a reorg rumor. The feature is one line item in a budget somewhere. When it ships, the team moves on to the next thing. That isn't a knock. It's how big companies have to work. But it's transactional. A founder doesn't move on. As Alex put it: "We're not going anywhere and this is our life's work." That energy shows up in small things. They pull their top 10 users every week and reach out to them. Everyone on the team does support, including their senior AI engineer. When a customer mentions Claude Design, Alex gets on the phone to learn how they work. The labs have distribution none of us will ever have, and I'm not going to pretend otherwise. What they can't staff is someone who wakes up thinking about one customer's problem and plans to still be working on it in ten years. So yes, the risk is real. Don't get rattled by it. Get closer to your customers than any feature factory team can.

There's a well-known @paulg essay that tells founders to find 10 users who love their product. He never defines love, and for a long time we didn't either. Where we've landed: love means 1. they're paying you, 2. use the product daily, 3. adopters organically expand. The holy trinity.

It took @sandengocka and me longer than I'd like to admit to stop staring at charts. We'd both spent years at companies well past product-market fit, and we brought those habits with us: dashboards, funnels, week-over-week growth, all tracked carefully on numbers too small to tell us anything. It felt like rigor. It wasn't telling us whether anyone cared. @daltonc and @mwseibel put it plainly. Don't worry about the graphs. Build something a few people love. There's a great line in this clip: it's almost impossible to have something 100 people love that isn't growing. Either you're lying about the love or you're not looking at the growth. For us, the number that shows it is a DAU/MAU of 50%. Half the people who use @earmark_ai in a month use it on any given day. The love produced that number. The number didn't produce the love. So we're playing the long game. AI Twitter has a new "fastest to $100M" chart every week, and good for those teams, but that's not the race we're in. Gmail spent months under 100 users. Stripe was invite-only for more than two years. The fundamentals still apply, even in the age of AI, you can't skip steps. If you're early, how many people love what you're building? Be honest about the number.

Central Coast product + engineering + design + founder folks 👋 We're hosting another @lennysan's Newsletter community meetup in San Luis Obispo at Rod & Hammer's October 21st from 6 - 9pm. Come meet other builders, swap notes, and have real conversations with people who understand what you do for living! Free drinks courtesy of @posthog 🦔. RSVP here: luma.com/dru3gaac

The next AI war is a harness war. @sdianahu from @ycombinator put it simply: “The model plus the harness gets you the output.” The models will keep changing. The durable value is everything that helps them understand what’s happening, remember what matters, and act inside the way a company actually works. One enormous source of that context is still mostly missing: conversation. Teams make decisions, surface constraints, hear customer feedback, and assign work in meetings. Then the meeting ends, the context gets flattened into notes, and someone has to explain it all over again to @claudeai, Codex, @cursor_ai, @linear, or the next agent. That’s the opportunity behind companies like @earmark_ai: make conversation executable. Give every agent the context it needs and set the work in motion while people are still talking. The meeting is the new prompt.

Today, we’re launching Earmark for Startups! 📼 👾 AI notes and summaries are about to go the way of Betamax. The next era isn’t about creating better records of what happened. It’s about conversations producing the work itself. Customer calls become product direction. Technical discussions become prototypes and specs. The context flows into @claudeai, Codex, @cursor_ai, @v0, @linear, and whatever your team uses next. We want startups building this way now. Qualifying teams can get full @earmark_ai Pro access free for three months, for up to 20 people. No stripped-down plan. No tiny trial. The whole product for the whole team. Apply here! content.tryearmark.com/start… Talk in. Deliverables out. No rewinding required.

Eighteen months ago the enterprise AI leaderboard looked settled. It wasn't. Look at where @OpenAI sat with enterprise then, and where @AnthropicAI sits today. That inversion didn't take a decade. It took a couple of planning cycles. I was recently asked on a podcast whether AI right now is a land grab. Win customers before behavior settles, the way people eventually stopped arguing about Yahoo vs. Hotmail vs. Gmail and just picked one. I don't disagree. But the board keeps flipping too fast for winning narratives to mean much. The script hasn't been written for anyone. Not for incumbents, not for startups, not for us. Nobody's won yet. Good. Let's build.

Thinking can't be outsourced. - @cagan Of his ten regrets, this is the one that applies to all of the above. Frameworks and AI both turn into avoidance tools when there's no judgment behind them.

Automation should make room for learning. - @karrisaarinen Every product you ship produces two things: the product, and the learning inside the team. Automate the repetitive work. Protect the work where people learn together.

Stay attached to the problem, hold the solution loosely. - @asvora & @mikeyk The PM job is connecting human problems to what tech can now do. Human problems move slowly. Model capabilities don't. Faster building doesn't shrink the job, it raises the premium on judgment.

"AI cannot turn an untested assumption into evidence." - @clairevo Eng capacity used to ration what got built. Now it doesn't. The scarce thing is customer evidence, conviction, attention. The question isn't "what can we build," it's "what do we need to prove next."

When a product creates a new way of working, your job is to make that promise understandable before someone has to experience it for themselves. Great interview from Caleb Pedosiuk where we discuss this as the biggest challenge in our business.

"The Meeting is the Work" with Caleb Pedosiuk and @earmark_ai. Thank you @thespeakerlisan for providing a platform for founders to soap box and riff.