Hands-on tutorials for Claude Code, AI agents, cloud and system design
A look at how ChatGPT works. ChatGPT served 100 million weekly active users within two months of launch, the fastest consumer-software adoption in history. The architecture that scaled to that size wasn’t new technology; it was the careful application of patterns every ML inference system has used for years, plus one trick: the per-session KV-cache… Continue reading
GPT-5.6 now runs on Amazon Bedrock. OpenAI’s current frontier family is available in more than 25 AWS Regions, reachable through cross-region inference profiles. This guide shows you how to enable the models, call them three different ways, read the real price sheet, and size quotas without surprises. GPT-5.6: What You’ll Learn GPT-5.6 arrived on Amazon… Continue reading
TL;DR 30 sec8 min coredeep dive optional Prompt caching on Amazon Bedrock charges 25 percent more to write a cached prefix and 90 percent less to read it back, so any context you reuse within the TTL bills at a fraction of the standard input rate. Cache reads cost 90 percent less than standard input… Continue reading
TL;DR 30 sec7 min coredeep dive optional Tiny-LLM is a hands-on course where you build an LLM serving engine from raw array operations on Apple Silicon: KV cache, continuous batching, and paged attention, ending with a coding agent running on the engine you wrote. Four weeks: Qwen3 from matrix multiplies, a faster single request, a… Continue reading
TL;DR 30 sec7 min coredeep dive optional Zero to SGLang is the official SGLang course, built with the Datawhale community, that takes you from running your first inference server to building a mini SGLang engine and then into the real one’s advanced internals. Part 0 gets a Qwen3-0.6B server running on your GPU in the… Continue reading
TL;DR 30 sec6 min coredeep dive optional Mooncake is Moonshot AI’s open-source serving architecture that treats the KV cache as the center of the system: prefill and decode run on separate clusters, and a disaggregated cache pool built from spare CPU, DRAM, and SSD moves context at RDMA speed. Best Paper at FAST 2025, and… Continue reading
TL;DR 30 sec6 min coredeep dive optional Cloudflare cut the inference cost of AI code review from $2.43 to $0.77 per review by moving one prompt cache breakpoint, deleting model calls nobody needed, and stepping models and effort down only where quality held. The cache fix alone lifted first-request cache hit rates from 14 to… Continue reading