Daria Y
443 posts
Daria Y
@daria7y
Building in public @trymira, personal AI agent in Telegram, 1M MAU tech entrepreneur, ML engineer, ex Google
Day 15 building in public 540,700 / 1m MAU Our creator contest just wrapped, and we've got a new wave of ambassadors. Next week we launch another one to grow that group several times over. At our scale, growth comes from people showing the product, not from paid clicks.
Day 14 building in public 538,500 / 1m MAU Every month, half a million people ask Mira to do something: search, memory, planning, shopping, building apps from a prompt. That's distribution most agents never get. So if you build a great MCP, API or vertical agent: you bring the solution, we bring the people. Already live with Granola and others. DMs open.
we launched mira × aesty.ai 😎 mira (500k MAU) is now in stylist mode - answers your questions, runs try-on, figures out color type the collab idea came from @ykvlv_ai not a full MCP yet but a first step in that direction! try it here: t.me/mira?start=aes…
Day 13 building in public 541,300 / 1m MAU Up 37k today. First real green day of the challenge.
Day 12 building in public 504,400 / 1m MAU While the user chart does its thing, another number moved the right way. Shipped Pro Max and annual billing. Average check up 20%.
Day 10 building in public 521,900 / 1m MAU A question for consumer AI founders: was there a feature your users started using in a completely different way than you intended?
AI is moving into the physical world. We're excited about a new wave of startups rebuilding the systems that power the real world, from education and healthcare to defense, finance, infrastructure, and work itself. ycombinator.com/rfs
Day 9 building in public 559,100 / 1m MAU A viral wave is the worst time to judge a product. Growth hides everything. Every weak spot in the product. The number goes up regardless, so you learn nothing. The plateau is where you finally see it. Who stayed, what they actually use, where it leaks. That feedback doesn't exist at the peak because the peak drowns it out.
Day 8 building in public 577,300 / 1m MAU The open weights debate keeps getting argued at the policy level. For us it's the daily job. The way we think about models inside @trymira: a reminder and a spreadsheet analysis should never cost the same to serve. Fusion is the part we're building. The Mixture-of-Agents paper showed it back in 2024: open models answering together, one model combining them, 65.1 on AlpacaEval 2.0 against 57.5 for GPT-4o. Frontier quality without frontier infra. At a million people, getting this layer right is the difference between a product that works and a product that quietly bleeds. Nobody sees this layer. It's most of the work.
Day 7 building in public 595,600 / 1m MAU A week ago I started posting this number publicly every day, which most founders would tell you not to do while it's moving the wrong way. Doing it anyway. It keeps the work honest and it makes the wins mean something when they come. Week two starts tomorrow.
Cognition is acquiring @Interaction, the makers of Poke. If you've ever used Poke, you know why we're so excited.
Day 6 building in public 599,148 / 1m MAU Skills are the new apps. The library is the new app store. Used to be: ship an app to the app store. Then: vibe code a landing page, bring your own tokens, figure out the architecture. Now: build on infrastructure that already works and ship a workflow in one tap. Meeting prep, daily briefing, expense tracking. One click to install, one click to share, monetize later. Building without building skills.mira.tg
