Mark Barbir

In the best meetings we have, no one takes notes. We are just in it. Completely present. We can do this because @earmark_ai has our back. It captures and generates work from conversation so we don't have to. But we noticed a second problem: after the call, we wanted a signal for what mattered most in an exact moment. That's what Pins do. Drag the Earmark widget over your Zoom window. Tap pin when something matters. AI titles it from the live transcript. Keep your eyes on the conversation the whole time. After the call, every pin is in your summary. The decision. The objection. The thing someone said that changed the direction. You didn't have to write it down. You just had to notice it. Earmark gives you presence. Pins tell Earmark what mattered. A few things worth understanding: ↳ Pins are exact. Not a rough timestamp on a long recording – the precise moment, AI-titled from context. ↳ Pins are configurable. The prompt behind a pin can be customized. What you do with the moment doesn't have to be the default. ↳ Pins flow into your work. Pull them into summaries, put them in a table, ask Earmark about them. Pins are live now. 🚀 📌

EverHealth didn’t roll out @earmark_ai because someone mandated it. Their teams kept finding value and expanding usage. Over the last year, adoption across EverHealth, DrChrono, EMHware, Updox, and other @EverCommerce organizations grew from 15 monthly active users to 60. Meeting volume grew nearly 9×. What stands out isn't the growth - it's the depth of engagement. Across the group, users have run nearly 7,000 meetings, created more than 5,000 artifacts, and shared those artifacts over 4,400 times. More than 80% of users have both started meetings and generated work inside Earmark. The strongest signal comes from EverHealth itself. Teams aren't using Earmark as a note-taker. They're using it as part of their operating system - capturing customer conversations, creating requirements, generating updates, and turning discussions into execution artifacts in real time. When adoption spreads organically across Product, Engineering, Design, Customer Success, Execs, and multiple business units, it's usually a sign that you're solving a workflow, not a role-specific problem. The future of work isn't about documenting conversations better. It's about turning conversations directly into outcomes.

If AI was going to kill SaaS and professional services, someone forgot to tell the AI labs. @OpenAI and @AnthropicAI aren't just building foundation models. They're buying consultancies and standing up forward-deployed engineering teams – the exact work that was supposed to disappear. @benedictevans mentioned this on @lennysan's Podcast. Companies don't have spare capacity to figure out which workflows to change, scope the rebuild, and execute it. Someone has to do that work. If you could point ChatGPT at a company and fire everyone in two weeks, you wouldn't need hundreds of consultants to make it work. That fact tells you where the value actually lives. The foundation layer isn't winner-take-all. The companies that capture value will be the ones that figure out deployment, workflow redesign, and specific vertical use cases – the work a product alone can't do. But that also means there is a lot of opportunities and enabling almost anyone to become an AI agent orchestrator. Think back to web development in the 90's and the opportunities building websites provided - this wave feels very similar.

Seven PMs at @ServiceTitan started using @earmark_ai in November. There was no formal rollout plan, no leadership mandate. Seven people tried it, and we watched to see what would happen. By February, Design joined. Then Engineering, then other teams from corners of the company we hadn't spoken to. Seven users became 60+ over six months, entirely through word of mouth and amazing internal champions. That's the growth signal we care about most. When we surveyed the team twice, 67% said they'd be very disappointed to lose Earmark. Every single respondent said at least somewhat disappointed. But the number I keep coming back to isn't the satisfaction score. It's what they said about time. Every respondent reported getting 5+ hours back per person, per week. We asked where those hours went – not "do you feel more productive," but where did they actually go. The answers were consistent: customer interviews, vision casting, deeper analytical work. One person wrote "thinking time – the part that gets cut first." Earmark gave it back. But the number that shaped the product most isn't in any of that. It's the requests. "Pipe Earmark into Claude." Shipped. "Don't store my sensitive meetings." Shipped. "Stop rebuilding the same task." Shipped. The ServiceTitan team didn't just Earmark for six months. They built it with us. That's the partnership worth talking about.

The most agent-pilled team we know of is hiring more humans, paying for more SaaS, and betting on PMs. On @lennysan's podcast, @danshipper lays out three predictions that pop the groupthink bubble. 🫧 First, PMs win this era. He describes a PM on his team who's lightly technical - not deep, but just technical enough. A year ago, that person couldn't have built anything meaningful alone. Today, with coding models, he ships to production faster than anyone. Good product sense plus just-enough technical fluency plus AI execution is dangerous. He doesn't need to rally a team anymore. He just builds. Second, designers are right behind them. They've spent years gated by engineering bandwidth and translation loss. Now they can build what they want. And because AI output trends toward sameness (i.e. we can detect a Claude generated design), differentiated taste matters more, not less. Finally, SaaS isn't dying - the most AI-native teams are paying for more SaaS, not less. Agents don't replace these tools. They use them. More agents means more workflows, more usage, more demand. The thread between all these is building is no longer the bottleneck. Deciding what to build, whether it's any good, and which problems matter is the bottleneck. Folks trained for that work can become first-class builders. The new unit of output isn't a team. It's a person with taste and an agent.

When we started @earmark_ai , the question we cared most about was simple: could we create a daily and essential service for our customers? In 2026, Earmark has grown 5× - more signups, more teams, more meetings shaped by the product. But the numbers we're proudest of aren't the growth ones. They're the quiet ones. Nearly 40% of our users open Earmark every single day. Half of the people who try it come back within the same week, and a meaningful share are still using it a month later. The average user spends about two hours a day inside Earmark and runs close to three meetings a day through it. People aren't just trying @earmark_ai . They're working in it. It's becoming part of the rhythm of their week. Quiet, constant, useful. To every team building with us - thank you. We're just getting started. 💪

we’re seeing huge demand at @earmark_ai for these types of workflows. solving bespoke workflows where someone can literally meet and their job is done after is what people are paying for

The return of the maker feels very real right now. As @bchesky said, the era of pure people managers - those who just run meetings and manage by proxy - is fading because AI is pulling everyone closer to the work itself. The best leaders are the ones still building, still shaping what customers actually experience, not just overseeing it. It’s less about managing people and more about managing the work - being a player-coach who can both lead and make. AI is just accelerating that shift.

@andrewchen Welcome to the party, better late than never.

bullish on the PM role quietly becoming the most important role in tech again when anyone can build, the person who decides WHAT to build becomes the bottleneck

We didn’t start with meetings - we started with the @Apple Vision Pro. Our first product was an immersive rehearsal tool for product and engineering leaders, giving real-time feedback on how you speak - pace, tone, delivery - inside simulated environments. It worked, but through dozens of customer interviews, we learned a hard truth: people don’t rehearse, they just jump from meeting to meeting. So we doubled down on what actually mattered - real-time feedback and real-time completion of work during conversations, not before them. That’s what led to @earmark_ai: instead of helping you practice for presentations, we sit inside them, turning conversations into actual deliverables as they happen. Same core idea, just applied at the moment it counts.

onboarding @earmark_ai

@AndrewYNg continues to illustrate what we’re seeing in the market. When coding gets 10x faster, the bottleneck moves. It’s deciding what to build, aligning on the customer problem, getting design/legal/marketing context, and turning the conversation into actual work. The best AI-native teams are starting to look different. Smaller. More generalist. Less handoff-heavy. Engineers are doing more product thinking. PMs are getting closer to building. Most importantly, everyone has more capacity to get closer to customers. That shift is hard, but it’s also exciting. The teams that win won’t just use AI to write code faster. They’ll use AI to compress the distance between conversation, decision, and execution. Great read deeplearning.ai/the-batch/issu…

@_catwu (Head of Product on @claudeai Code + Cowork) said something on @lennysan's podcast we've been seeing in the market too: People spend more time tinkering with their AI setup than using it. Skills, MCPs, custom workflows - the rig looks great in a theory. But the actual work isn't moving. Her line: "the simple setups actually work better." Build stuff you'll actually use daily. One-shotting a prototype is fun for ten seconds and then you never open it again. That's not leverage. The tinkering can feel productive, but building something real should be the goal. 💪

Engineering is moving faster than ever. What we’re seeing now is that AI has shifted the bottleneck. It used to be engineering capacity. Now it’s often product’s ability to keep up with what engineering can build. That changes the job. PMs and other disciplines can’t just react anymore. They need to understand the technical reality, make faster judgment calls, and help teams decide not just what can be built, but what should be built. The teams that win won’t just move faster. They’ll create capacity to be even more strategic and customer facing in order to make better bets.

One of our favorite things about Earmark is that it helps in the moment, not just after the meeting. A simple example: you’re in a product or engineering discussion, someone goes deep on the technical side, and @earmark_ai translates it into plain language while the conversation is still happening. You can keep up without interrupting the flow or feeling lost. Then the real magic kicks in. Instead of ending the meeting with a summary and a bunch of follow-up work, we use that live context to start the work right there. Feature ideas can turn into prototypes in @cursor_ai before the meeting even ends. That shift matters. Fewer “we’ll follow up later” moments. More real progress while everyone is still in the room.

Our first product was for Vision Pro. We thought better meeting prep could help solve the overload problem we all experienced. That didn’t become the company, but it gave us one idea we kept: real-time feedback matters. That idea carried over from the beginning. First as live prompts in presentation rehearsals. Today, it shows up in Earmark helping teams while the conversation is still happening, not after. The bigger lesson was simple: don’t build an unproven product on top of an unproven platform. So we moved to a daily problem almost everyone has, meetings, and built for the web so it could work anywhere.