Viewing profile — aaronSong

Clink connects your existing AI coding agents (Claude Code, Codex CLI, Gemini CLI, Z.ai GLM) to build, preview, and deploy apps in isolated containers. No extra tokens needed. Why?…

thanks for using claudable before! for server costs, the sandbox previews have auto-stop functionality so the actual runtime is pretty short. keeps costs manageable. for deployment…

Oh I really like your product! you're right about how it works. the desktop app sends oauth tokens to connect to the cloud infrastructure. dyad looks really cool, love the local-fi…

Thanks for reaching out about the hosting platform! really interested in the European market, will definitely drop you an email once we get things more organized on our end. You ma…

ah, I think there's some confusion with the open source version! The OSS Claudable runs everything locally on your machine. This new Clink product runs in the cloud with Daytona sa…

Most app builders use lightweight coding agents and infrastructure. we're leveraging the full power of CLI agents (Claude Code, Codex, etc) as the engine, which means you can build…

Good question! while we use CLI agents as the coding backbone, our focus is really on the application building experience: live preview and seamless deployment. The key difference …

Ah, Vercel is what we use in the open source Claudable version! For Clink, we currently deploy static/frontend to Cloudflare Pages and full stack apps to Railway, both work great f…

For now, we support switching between models within the same provider, but we're currently adding the ability to switch providers mid-session. We're working on agent switching simi…

sandboxing is indeed the trickiest part. We're using Daytona for development environments, which gives us proper isolation out of the box. Each build runs in its own sandbox worksp…

Clink lets you use the coding agents you already pay for (Claude Code, Codex CLI, Gemini CLI, Z.ai GLM) to build → live-preview → ship apps in an isolated container. No token purch…

What I liked here is how unglamorous it is: tiny prompts, context only when needed, evaluate against real usage, and resist multi‑agent stuff until a single boring pipeline is stab…

Solid, practical write‑up on running an “agent” in production. Biggest takeaways for me: keep the base prompt tiny, inject context only when it’s actually needed, and eval on real …

that's great, will try

openai's crawling is the best. just following anthropic's way

nice naming. how is this different from something like Glass by pickle?

page looks like vibe-coded. is it?

Love the demos, but the bar for “digital restoration” should be provenance + uncertainty-by-default: clearly label what’s scanned fact vs. modelled conjecture, keep raw data revers…

I’ve been building AI tooling and noticing a shift: shipping is cheap; discovering demand and capturing value is the hard part. Hypothesis: as code/infra get commoditized, we may b…

I like how it integrated it