https://krishghosh.com/ Krish Ghosh builds the software that makes AI agents reliable: running them, testing them and serving the models underneath. Plus plain-English writing on why software breaks and how to fix it. en-GB Thu, 17 Sep 2026 12:00:00 GMT https://krishghosh.com/writing/autoresearch-parts https://krishghosh.com/writing/autoresearch-parts Thu, 17 Sep 2026 12:00:00 GMT A system that runs its own experiments has about seventeen moving parts, and each one has a way of quietly lying to you. Here is every mechanism I found that stops it, in the order the loop runs them. AI agents Evaluation LLMs https://krishghosh.com/writing/149-research-agents https://krishghosh.com/writing/149-research-agents Wed, 16 Sep 2026 12:00:00 GMT Systems that run their own experiments, read their own results and keep the changes that worked. The ones that have discovered anything real share one thing, and it isn’t the search algorithm. It’s who marks the homework. AI agents Evaluation LLMs https://krishghosh.com/writing/why-agent-runs-fail https://krishghosh.com/writing/why-agent-runs-fail Fri, 14 Aug 2026 12:00:00 GMT Almost every failed agent run I’ve debugged came down to a timeout, a retry that did something twice, or context that got cut off without warning. The model was fine. The system around it wasn’t. AI agents Reliability LLMs https://krishghosh.com/writing/where-the-money-goes https://krishghosh.com/writing/where-the-money-goes Sun, 09 Aug 2026 12:00:00 GMT These days, much of an LLM bill goes on words you never read: the model thinking out loud, charged by the word. Here’s where the money goes, why more thinking isn’t always better, and how to stop paying genius rates for easy questions. LLMs AI agents Cost https://krishghosh.com/writing/scale-is-the-boring-part https://krishghosh.com/writing/scale-is-the-boring-part Mon, 03 Aug 2026 12:00:00 GMT Seven years took language models from 124 million parameters to 2.8 trillion, and the block diagram barely moved. The interesting part is quieter: what a model keeps while it reads, what it throws away when it runs out of room, and who gets to decide. LLMs Attention Architecture https://krishghosh.com/writing/checkout-is-the-easy-part https://krishghosh.com/writing/checkout-is-the-easy-part Mon, 15 Dec 2025 12:00:00 GMT A subscription isn’t a row with a boolean on it. It’s a long-running state machine that outlives every process and most of the services that touch it. Here’s how the real ones are built, and three problems that only show up once there are hundreds of millions of them. Distributed systems Payments Reliability https://krishghosh.com/writing/the-cost-of-consensus https://krishghosh.com/writing/the-cost-of-consensus Fri, 18 Jul 2025 12:00:00 GMT Every consensus protocol has the same speed limit, a round trip to most of your servers. So the real engineering happens in where you put those servers and what you’re willing to call a failure. Distributed systems Consensus Latency https://krishghosh.com/writing/backpressure-is-a-design-language https://krishghosh.com/writing/backpressure-is-a-design-language Fri, 02 May 2025 12:00:00 GMT Every queue in your system is a decision about who waits when things get busy. Most outages I’ve read about happened because nobody made that decision on purpose. Distributed systems Reliability