When I code AI assisted Cursor and OpenCode GPT 5.5 I get way better and faster results if I start with the UI than with the Backend. Backend @fastifyjs @trpcio or @honojs @middleapi (oRPC) @DrizzleORM @PostgreSQL @Docker Storage @Minio with Docker for testing @Cloudflare R2 for production Files SDK files-sdk.dev/ Frontend @reactjs @ReactRouter or @tan_stack. UI @shadcn or @base_ui Hosting @Hetzner_Online @getdokploy

Adapters shipped for 🚢 - @honojs - @elysiaJS - @UseExpressJS - @fastifyjs - @nextjs - @nestframework

Just launched Sangria! It's an open-source project that allows developers to accept agent payments, in under 3 lines of code! Come with adapters for most of the popular frameworks, @nextjs @fastifyjs @honojs @fastiapi express. powered by x402 (@coinbase)

1 MB. Zero dependencies. evlog is everything your logging library should be. ✨ Wide-events native Log entire request contexts, not just strings. One log per request, full context. 🤖 @vercel AI SDK integration Track LLM calls, tokens, costs, tool calls, streaming metrics automatically. Wrap your model and get full AI observability. 🔐 @better_auth integration Every wide event includes who made the request. userId, user profile, session metadata. Zero manual work. 📋 First-class audit logs Tamper-evident audit trails. One enricher, one drain wrapper, one helper. 🧹 Auto-redaction Smart masking for credit cards, emails, IPs, phone numbers, JWTs, PII. Before console output and drains. 🔌 Drain to anything @AxiomFM, @datadoghq, @PostHog, @getsentry, @BetterStackHQ, @hyperdxio, @grafana / @honeycombio via OTLP, custom endpoints, file system, @nuxt_hub self-hosted storage. ⚡️ Every framework @nuxt_js, @nextjs, @sveltejs, @nitrojsdev, @tan_stack Start, @nestframework, Express, @honojs, @fastifyjs, @elysiajs, @remix_run, @astrodotbuild, @cloudflare Workers, @awscloud Lambda. 🎯 Structured errors Errors that explain why and how to fix. Actionable context for humans and AI agents. 🎨 Client logging Browser events with the same API. Auto console styling, user identity, optional server transport. 🔥 Tail sampling Never miss errors or slow requests. Head sampling drops noise, tail sampling rescues critical events. 📦 Typed fields Compile-time type safety for wide events. Prevent typos, ensure consistency. 🛠 @vite_js plugin Build-time optimizations, auto-init, debug stripping, source location injection. 🔍 Agent Skills AI-assisted code review and migration guide. Digging through logs is not observability. It's hope evlog.dev

Last week i managed to make @fastifyjs work on @Cloudflare workers. Had to patch a few things but as of now i managed to make fastify, ajv, typebox, pino & light-my-request work all together 👀

httpxy 0.5 is now WAY faster 🔥 Special thanks to @matteocollina and @fastifyjs ❤️ Performance optimizations were inspired by analysis of fast-proxy.

Nest is for noobs who don't know how to structure projects and they need help so Nest forces you to follow their pattern, i personally hate it, I use @fastifyjs and now @honojs and very happy and much faster easier to maintain code and i use @DrizzleORM @kysely_

You write a skill, test it once, and it seems fine. But do you actually know if it works? In his latest post, Simon Maple ( @sjmaple ) shows how misleading that assumption can be. He ran a well-written Fastify (@fastifyjs ) skill through an eval pipeline. It had a perfect structure score. Clear instructions. Good triggers. Everything looked solid. And it still failed in real scenarios. That’s the gap most people miss. A clean SKILL.md doesn’t mean the agent follows it. Sometimes it helps, sometimes it does nothing, and sometimes it even makes things worse. The interesting part is how he fixes it. Instead of manually writing evals, he uses a single command to: • generate real scenarios • run the agent with and without the skill • measure the actual impact • detect regressions • suggest fixes • re-run and verify improvements No guesswork. Just numbers. In his example, the skill jumped from ~67% to ~94% success rate after one optimization loop. More importantly, it exposed issues that would’ve gone unnoticed in production. The takeaway is simple but important: If you’re building with agent skills, intuition isn’t enough. You need a feedback loop that tells you whether your instructions actually improve outcomes. Otherwise, you’re just hoping your skill works. Read the full blog here: tessl.io/blog/stop-guessing-…

Can you open issues in the repo for the skills that needs improving? This is indeed very useful!

Model bake off! Which Claude code model most effectively uses @matteocollina’s @fastifyjs agent skill? We tested Haiku, Sonnet and Opus using a @tessl_io evals and a new skill we built. tessl.io/blog/your-skill-wor…

Glad to hear that. World getting back to normal 🙏

Dealt with that a while back. So the app started as a full stack nextjs with server actions. As we figured out exactly what we wanted, it was clear we needed a separate backend so we moved that to @fastifyjs . The Ul was the only thing left as we did not need SSR anymore really

👀 API looks super clean! Does it work natively with @nextjs?

Strong pitch

Will definitely check out

@boristane 👀

I must say, @evilrabbit_ You hire incredibly talented folks. Every single one of your engineers is exceptional.

pure 🔥

Wow, the landing page is really good!