PinnedHugo Sequier·May 13Learning the Result Forgetting the ProcessFrom school to AI, I was often taught to pass the test before I was taught to understand what I was doing.
PinnedHugo Sequier·May 10The Productivity Trap of AIWe learned to produce faster, but I am not sure we learned what that production is doing to us.A response icon1A response icon1
Hugo Sequier·Apr 15Why AI Agent Harnesses Matter More Than Model ChoiceGemma and Qwen keep getting better, but most agent failures still come from bad execution architecture.
Hugo Sequier·Apr 7How I Cut Token Usage in Half by Redesigning My Agent ArchitectureBigger models, same bill. The fix is in how you architect.
Hugo Sequier·Apr 2How I Built a Second Brain for Claude CodeAfter using Obsidian as my own external brain, I realized the real bottleneck was the agent doing the work
Hugo Sequier·Mar 17Autoresearch Beyond ML: How I Adapted Karpathys Loop to Trade Prediction MarketsFix the evaluator. Search the strategy space. That’s the whole idea.
Hugo Sequier·Mar 12The Impostor Loop: Why Vibecoding Made Me Feel Like a FraudI ship faster than ever. I’ve never felt less competent.A response icon1A response icon1
Hugo Sequier·Mar 3How I Keep Claude Code Fast With Zero PluginsNo MCP servers, no extensions, no magic. Just markdown files and a 4-step workflow.
Hugo Sequier·Feb 24How I Turned OpenClaw into a MVP AgencyI set up 5 AI coding agents with OpenClaw on a cheap VPS, wrote a Python script to orchestrate them, and now I ship working apps while…
Hugo Sequier·Feb 20How to Use Kimi K2.5 Free on OpenClawNVIDIA NIM gives you Kimi K2.5 for free. Here’s how to set it up, and what they don’t tell you about latency
