Building a self-planning data pipeline agent, and measuring whether it's actually rightHow Sluice repairs drifted data files with LLM agents, a locked-down sandbox and exact ground truth, on free-tier models TL;DR — Sluice is an open-source agentic data pipeline. A file arrives with dr6h ago · 12 min readpranavkrishna.hashnode.dev
Building MergeSignal: Pre-Merge Intelligence for GitA few months ago, I spent an afternoon debugging a production incident that should not have happened. Two pull requests merged cleanly. CI passed on both. The merge queue approved them. The combined cSep 21 · 10 min readpranavkrishna.hashnode.dev
Building a Thinking-First Multi-Agent Room (Not Another Fake Agent Demo)Most multi-agent demos feel the same. They start with a fixed cast: Builder Critic Reviewer Then they run a fixed pipeline, often with web search on every request, and call that “orchestration.” Jul 25 · 4 min readpranavkrishna.hashnode.dev
I Gave an AI Agent Access to a Production Database. Here's What Actually Happened.Subtitle: A real account of using Model Context Protocol in production — the architecture, the failures, and what I'd do differently. The Setup Last year I was working on a system that needed to do so1BMay 19 · 8 min readpranavkrishna.hashnode.dev
Why Developers Forget Decisions — And Why AI Won’t Fix ItEvery developer knows this moment: You remember that a decision was made.You remember it was important.You just don’t remember where it happened — Slack, Zoom, Teams,Gmeet, or a random document. You scroll.You search.You reconstruct history. This is ...Jan 24 · 2 min readwhy-developers-forget-decisions.hashnode.dev
