Agentic AI, LLM engineering, and the models behind modern automation — multi-agent systems, LLM evaluation and comparisons, RAG, fine-tuning, AI infrastructure, security, and production AI engineering.
From the blog
How SIP REGISTER and INVITE work end-to-end, why your AI agent platform needs to handle 401 challenges and Record-Route correctly, and the failure modes that bite production builds.
Langfuse was acquired by Clickhouse in January 2026. Helicone is the right default for most production teams. Here is the 2026 observability picker.
Score sentiment from –1.0 to 1.0, lead intent from 0 to 100, and extract structured entities from every call. Async pipeline with NATS, gpt-4o-mini, and a Postgres analytics table.
Moveworks ships 100+ prebuilt Workday agents inside Slack and Teams. Workday promises 14 more agents in 2026. Here is what HR self-service chat actually deflects.
MV3 broke a lot of things, but it left WebRTC alone. Here is how teams are shipping side-panel voice agents in Chrome and Edge in 2026, and what to watch for.
Use Supabase Auth, RLS-protected Postgres, and Deno Edge Functions to ship a multi-tenant voice agent. Real working SQL policies and TS code.
A complete walkthrough for building a custom MCP server: FastMCP for Python, the official TypeScript SDK, the JSON-RPC spec, stdio vs streamable HTTP, and OAuth 2.1 with PKCE.
Every PR that touches a prompt, tool, or model should run the eval suite and block on regression. Here is the GitHub Actions setup we ship across 37 agents.
The 2026 vector DB market consolidated to four serious products. Pinecone is fastest to deploy. Qdrant has the best price-performance. Weaviate wins on hybrid. Chroma is the dev sandbox. Here is the buyer's guide.
