Sahil Patel

our go-to backend for database, and real-time workflows.

vercel ai sdk made it easy to build fast ai features and ship iterations without worrying about infra.

MiniMax M2 surprised us. it’s fast, cheap, and very strong at coding and tool use. We’ve seen it outperform Gemini 2.5 Pro, Claude Sonnet 4.5, and Qwen3 on several benchmarks, and it works really well for agentic tasks.

Kimi K2 is one of the best models we’ve used for writing. It handles long context really well, stays coherent, and produces clean, readable drafts without much prompt tweaking.

Vercel makes building and shipping products ridiculously fast. From frontend hosting to edge functions and AI SDKs, it removes a ton of infra friction and lets teams focus on shipping.

We’ve been really impressed with DeepSeek. R1 and V3 showed that open models can genuinely compete with top proprietary models, esp on reasoning and coding. The level of transparency in their research is rare and super valuable for the ecosystem.

GLM 4.6 is incredible for coding. it performs similarly to claude sonnet 4.0 but at much lower cost

we've been using posthog to track how users interact with the product, fix bugs they report, and improve usability