HydraDB: Building Memory That EvolvesLast time I wrote about memory, I shipped a minimal Mem0-style system — vectors, LLM-as-judge, ADD/UPDATE/DELETE on extracted facts. It worked. People liked it. But the deeper I went, the more I realiMay 26 · 28 min readkarthiknadar1204.hashnode.dev
son of COT->Self-ConsistencyIf you’ve spent any time prompting large language models in 2025–2026, you’ve probably used (or at least tried) Chain-of-Thought (CoT) prompting. You give the model a few examples that show step-by-stMar 7 · 4 min readkarthiknadar1204.hashnode.dev
Tree of Thoughts: Making LLMs Actually Think Before They SpeakIn my last post I wrote about Chain-of-Verification — a dead-simple way to catch hallucinations without any external tools or retraining.Today we’re moving to the next pain point: tasks where the modeMar 7 · 8 min readkarthiknadar1204.hashnode.dev
Building MemoryBank: Bringing Long-Term, Human-Like Memory to AI CompanionsAI memory—being a subset of self-improving agents—has been my area of curiosity for quite some time now. When I came across this paper MemoryBank: Enhancing Large Language Models with Long-Term MemoryMar 7 · 9 min readkarthiknadar1204.hashnode.dev
Why One-Shot Answers Are the Real Hallucination ProblemIn my last post where I went deep on AI memory systems, that was all about helping models remember better.Today, we’re tackling the other side of the pain spectrum: hallucinations. You’ve seen it. You ask ChatGPT or Grok something simple, and it repl...Jan 24 · 4 min readkarthiknadar1204.hashnode.dev
