Ashwindevelops·Oct 8Your AI Agent Does Not Need More Memory. It Needs Better Context.An agent starts a task well.A response icon2A response icon2
Ashwindevelops·Sep 4Your AI Agent Does Not Need More Markdown. It Needs the Right Markdown.Most teams start with one AGENTS.md file.
Ashwindevelops·Aug 27Turbopuffer Made Me Rethink RAG: A Fast Index Does Not Mean Correct RetrievalMost teams building RAG ask this first:
Ashwindevelops·Jul 26vLLM Teardown Part 3: The Layer Nobody BenchmarksYour agent works. Tools get called, JSON comes back clean, tests are green. So you turn on the speed features from Part 2, watch throughput…A response icon1A response icon1
Ashwindevelops·Jul 18vLLM Teardown Part 2: The Virtual Memory for LLMsIn Part 1, we looked at the basic math of how many users fit on a GPU if everyone is the same size. But in the real world, traffic is…
Ashwindevelops·Jul 12vLLM Teardown: How Many Users Can One GPU Really Hold?Getting an open-source model to talk back takes about five minutes. You pull meta-llama/Llama-3.1–8B-Instruct, run vllm serve, send one…A response icon3A response icon3
Ashwindevelops·Jun 13Stop Storing Everything: Hard Lessons From Building an AI Agent With MemoryI used to think the hard part of building an AI agent was the model.
Ashwindevelops·May 26The Skill Engineers Are Quietly Losing While AI Makes Them FasterThere’s a strange thing happening to a lot of engineers right now, and almost nobody is naming it. We’re getting more productive every…
Ashwindevelops·May 9The Rise of AI Skills: Why Prompting is No Longer EnoughI used to think the secret to better AI output was a better prompt. So I’d spend an hour refining a single sentence hoping that if I just…
Ashwindevelops·Apr 12Agentic AI: The Shift from Chatbots to Autonomous WorkersI used to think that the peak of AI was a chatbot that could write a decent email or explain a complex concept.