https://www.kumarharsh.in/writing/Notes by freelance AI engineer Kumar Harsh on building AI systems that stay up: WhatsApp agents, LLM failover, RAG and how long an AI build takes.enhttps://www.kumarharsh.in/writing/whatsapp-ai-assistant-cloudflare-workers/https://www.kumarharsh.in/writing/whatsapp-ai-assistant-cloudflare-workers/Sun, 01 Nov 2026 00:00:00 +0530The architecture behind Honey, a WhatsApp AI assistant on Cloudflare Workers and D1 that runs for ₹0 a month: failover cascade, memory and cron routines.https://www.kumarharsh.in/writing/llm-provider-failover/https://www.kumarharsh.in/writing/llm-provider-failover/Sun, 01 Nov 2026 00:00:00 +0530How to design LLM failover: fallback cascades, timeouts, retry budgets, circuit breakers and chaos testing, with a TypeScript sketch.https://www.kumarharsh.in/writing/how-long-to-build-an-ai-agent/https://www.kumarharsh.in/writing/how-long-to-build-an-ai-agent/Sun, 01 Nov 2026 00:00:00 +0530Typical timelines for AI agents, RAG systems, AI MVPs and reliability work, and the factors that make a build shorter or longer.