Latest blog posts and updateshttps://rajapatnaik.com/en-ushttps://rajapatnaik.com/blog/2026/04/07/hermes-agent/https://rajapatnaik.com/blog/2026/04/07/hermes-agent/I've been running OpenClaw and Hermes side by side, and I think that's actually the right move for most people. They solve different problems. Here's the full breakdown on Hermes — what it is, how it's different from OpenClaw, and how to get it running.Tue, 07 Apr 2026 12:00:00 GMThttps://rajapatnaik.com/blog/2026/04/05/adversarial-debate-gepa/https://rajapatnaik.com/blog/2026/04/05/adversarial-debate-gepa/Most prompt optimization demos show a single agent getting better at a single task — classify this email, answer this question, extract these fields. That's useful, but it sidesteps the harder problem: what happens when multiple agents interact, and improving one changes the game for the others? I built a multi-agent adversarial debate system with DSPy and GEPA to find out.Sun, 05 Apr 2026 12:00:00 GMThttps://rajapatnaik.com/blog/2025/10/23/langgraph-dspy-gepa-researcher/https://rajapatnaik.com/blog/2025/10/23/langgraph-dspy-gepa-researcher/Build an agentic research system with GEPA prompt optimization at every stepThu, 23 Oct 2025 12:00:00 GMThttps://rajapatnaik.com/blog/2025/10/20/sql-generator/https://rajapatnaik.com/blog/2025/10/20/sql-generator/Evolving an NL → SQL Generator with DSPy + GEPA (no labels required)Mon, 20 Oct 2025 12:00:00 GMThttps://rajapatnaik.com/blog/2025/10/17/dspy-fact-checker/https://rajapatnaik.com/blog/2025/10/17/dspy-fact-checker/Build an LLM that self‑corrects instead of 'RAG and pray.'Fri, 17 Oct 2025 12:00:00 GMThttps://rajapatnaik.com/blog/2025/10/14/dspy-gepa-deidentification/https://rajapatnaik.com/blog/2025/10/14/dspy-gepa-deidentification/Most “privacy filters” are vibes. They look redacted… until an email or phone number slips through in production.Tue, 14 Oct 2025 12:00:00 GMThttps://rajapatnaik.com/blog/2025/10/10/intro-to-dspy/https://rajapatnaik.com/blog/2025/10/10/intro-to-dspy/A technical deep dive into DSPy, Stanford's framework for programming—not prompting—language models. Learn how to build reliable AI systems with signatures, modules, and optimizers through practical code examples including classification, chain-of-thought reasoning, RAG, and multi-step agents.Fri, 10 Oct 2025 12:00:00 GMThttps://rajapatnaik.com/blog/2025/09/02/claude-code-lighthouse/https://rajapatnaik.com/blog/2025/09/02/claude-code-lighthouse/Specification for a new Claude Code feature I call Lighthouse that makes large codebase edits transparent and safe: the agent states intent in a Change Contract, shows blast radius via an Impact Map, and executes work in Checkpoints with snapshots and one‑click rollbacks.Tue, 02 Sep 2025 12:00:00 GMThttps://rajapatnaik.com/blog/2025/09/01/openai-realtime-rust/https://rajapatnaik.com/blog/2025/09/01/openai-realtime-rust/For anyone wanting to try OpenAI's real-time API, I created a repository that provides a minimal Rust implementation, enabling bidirectional voice conversations through WebSocket connections. The implementation handles real-time audio streaming, automatic speech detection, and live transcription display.Mon, 01 Sep 2025 12:00:00 GMThttps://rajapatnaik.com/blog/2025/08/22/omarchy/https://rajapatnaik.com/blog/2025/08/22/omarchy/Your operating system is holding you back, and you probably don't even know it. Whether you're trapped in Apple's ecosystem, suffering through Windows' bloatware, or settling for Ubuntu's safe mediocrity, you're missing out on what computing in 2025 should actually feel like. DHH's Omarchy made me realize we've all been expecting far too little from our machines. If I'm going to start a blog, I might as well begin with the revolution that made me give a damn about my OS again.Thu, 22 Aug 2024 12:00:00 GMT