Optimizing AI agents requires writing down the valuable institutional knowledge in your company. Big co’s will struggle with this: middle managers worry about becoming replaceable and leaders worry about leaks. If they don’t do it, they’ll be crushed by competitors who do.

Harj Taggar Sam DeBrule Totally agree. I’ve spent the last decade helping teams go async-first and write down what most leave unsaid. It’s not flashy, but it’s a serious advantage now. The companies that already work this way will move faster than those still hoarding info in meetings and minds.

Introducing Clerk Billing The easiest subscription billing for B2B and B2C SaaS. No webhooks, no UI to build, no payment integration code, no DIY authorization engine. Just enable Billing and drop a <PricingTable/> on your /pricing page!

Introducing the open-source Next.js Audio Starter Kit Add Text to Speech, Speech to Text, Sound Effects and Conversational AI to your product in minutes. Built with: • ElevenLabs SDK • Next.js + shadcn/ui • Tailwind CSS v4 Repo link in the comments below:

This is based on `resumable-stream` which you can find on npm! A core feature is that while it uses Redis, the actual Redis usage is minimal unless you actually need to resume a stream (which should be rare in practice). It supports: - Clients resuming streams on network

New tsgo compiler drops Sentry's typecheck time from 54.8s to 6.8s This is a 1.5m LOC codebase Time to start trying this out

AI agents have a scaling problem. (And we're finally fixing it) Over the last few years, we saw a pattern: → Developers can prototype AI workflows quickly. → But moving from prototype to production? Painful. Most frameworks demand deep AI expertise, heavy Python use, and

Built on top of the amazing tech stacks from Vercel ai sdk, Mastra workflows, and Bun!

Something important to remember when architecting software using LLM dev tools that your team is only limited by the number of terminals you have open. Considerations around interface design and boundaries are now an issue even personal projects face if you want to move fast

almost there...

if you're building rag, avoid these antipatterns: link to the talk in the thread

Oh my gosh! It's happening 😵 Here's me on macOS 26 Beta running containers *without Docker installed* 😱 NATIVE CONTAINER SUPPORT IN MACOS github.com/apple/container

Problem: Your Playwright tests break every sprint because you're testing through the UI. Solution: I learned to test business logic through APIs first. UI tests are 10x slower and more brittle. Now I only use UI tests for actual user journeys that can't be validated via API.

Hono v4.8.0 is out! * Reduced the code size: gzipped hono/tiny is 4.5KB! * New Route Helper * SSG supports Plugin * 3rd party middleware: * UA Blocker Middleware * MCP Middleware * Zod Validator supports Zod v4 * and more Enjoy! github.com/honojs/hono/re…

WOW! dina kozlov 🐀 from Cloudflare just wrapped a 🔥 session on building Model Context Protocol (MCP) servers. 💡 Best practices 🛠️ Real-world patterns 🤖 Agent-ready infra tips ✨ Plus… iconic MCP earrings. If you missed it, catch the replay — this one was gold.

introducing mastra scorers evals still work but we learned a lot from watching how people used them users want to be able to do things like use evals on workflow steps businesses need custom evals tailored to KPIs so we shipped a bunch of improvements and are calling it

Excited to launch GPT-5 on OpenRouter! 🧠dynamic reasoning ("think hard about this!") 🧑‍💻 much better frontend design and debugging 🏥 OpenAI's best model for health-related questions Details 👇