I’m excited to share some big news: the company I co-founded, Mastra, just got into @ycombinator's Winter 2025 batch!

If you think you're late to AI, you're wrong. Here's what I learned from Series A founders. - Series A startups are only NOW adding AI features after customer demand became overwhelming - Customers are flooding in from AI tools like Claude, Bolt, and Lovable - forcing teams to adapt - Tech startups (the early adopters) are just beginning their AI integration - other industries haven't even started - "AI companies" will soon be as redundant as saying "cloud companies"

RT @alanwells: 🧵 How I built a founder-mode customer support bot by stealing ideas from self-driving cars instead of adopting traditional C…

RT @zenorocha: Today, we're excited to announce @newdotemail. It's a new way to build beautiful, responsive, and cross-platform emails usi…

@resend @AnthropicAI Yes!

RT @calcsam: sometimes the best feature you can ship is removing one which is why we've taken out the @mastra_ai model routing in favor of…

Register for the Workshop here:

@kaushikm_ @stripe @googlesheets @X @GIPHY @resend I would like to make it easy for our internal agents to access Linear and Notion!

RT @pennepastaa: a classic for founders 👀 Thank you @mastra_ai for the book! If you’re wanting to get your AI agents where your users are…

RT @kaushikm_: Turn the Google Sheets API into an AI Agent <2 min with agents.json - create new sheets - update existing ones - extract i…

The next evolution of product development isn't about prettier interfaces or better APIs - it's about building for AI agents. - Current APIs were built for human comprehension, not machine efficiency - Web interfaces are visual-first, making them challenging for AI agents - Why we need open web standards for agent-to-agent interaction - The critical gap between how machines and humans process information - Why @biilmann believes AX standardization is the next big challenge Want to shape this future? Join the conversation at

@pelaseyed Not sure my pockets are deep enough to actually run this on any legitimately sized knowledge base, but the accuracy benchmarks have to be off the charts. Now if costs keep decreasing and speed keeps increasing... you just might be onto something here.

@abhiaiyer @ehindero_ Legend

Register here:

RT @tryfondo: And we're live! W25 + Alumni Co-working & Headshots Day is in full swing: Founders, headshots, coffee — all day. 🚀 Big sho…

RT @abhiaiyer: What I’m Learning About AI Agents at @ycombinator 🧵

RT @calcsam: Managing agent memory means feeding it the right context from previous interactions. So: both recency and semantic similarit…

When coding, naming things is hard. One of my favorite uses of AI is to give it a few examples and then say "you decide" It picks the best option most of the time. Even if it doesn't, it convinces me it's the best option so I'm okay with it.

Most AI Agents eventually go off the rails, but there's a pattern to help prevent it. We've watched dozens of teams build agents. The ones with the most success all follow a similar process. - Letting AI agents make all decisions leads to unpredictable (and often bad) outcomes - The critical parallel between human workflows and successful AI agent design - Adding structure and "constraint" to your agents counter-intuitively leads to better results