Running the OWASP LLM Top 10 as a real audit is a completely different experience than reading it as a checklist. I ran it against my TypeScript agent stack with system prompts, MCP tools, and Cline — and the findings were uncomfortable.
Jun 20 2026 · 9′ · Tutorials · TypeScript · LLM
The most common mistake when implementing MCP tools is coupling them to the provider's SDK. The spec exists to prevent exactly that. A practical architecture guide: the input/output contract that makes a tool work across Claude, GPT, and local models without rewriting anything.
Jun 10 2026 · 8′ · Tutorials · TypeScript · LLM
Two weeks using Cline as an autonomous coding agent on a TypeScript project. What tasks I delegated, where it screwed up, how it compares to Claude Code, and which workflows I'd never hand over to it. Architecture-grade analysis, not hype.
Jun 07 2026 · 11′ · Tutorials · TypeScript · developer tools
A system prompt isn't documentation for the model — it's a contract. After several redesigns, I landed on a format with fixed sections, explicit limits, and dynamically injected context. Here's what survived and why.
May 29 2026 · 9′ · Tutorials · TypeScript · produccion
The hype says Kimi K2.6 beat Claude and GPT-5.5 at coding. I ran it against my own codebase — not cherry-picked HumanEval — and what I found changes the question you should actually be asking.
May 03 2026 · 11′ · Experiments · TypeScript · produccion
874 points on HN for 'I cancelled Claude'. Before joining the chorus, I ran my own regression cases against real Claude Code logs. The degradation is real — just not where everyone's complaining about it.
Apr 25 2026 · 8′ · Experiments · TypeScript · claude code
Anthropic just said Claude CLI usage in the OpenClaw style is allowed. Last week I was doing it with a knot in my stomach. Today I'm doing the exact same thing. What changed was the paperwork. And that tells me something pretty uncomfortable about what it means to build on platforms that rewrite the
Apr 21 2026 · 7′ · Opinion · anthropic · agentes-ia
I diff the public Claude system prompts line by line between versions and map the behavior changes I was already observing in production before I knew the prompt had changed. The canary I didn't have.
Apr 20 2026 · 8′ · Experiments · produccion · anthropic
A Hacker News thread with 208 points asks whether AI agent costs grow exponentially. I have months of real production logs. The answer isn't what anyone expects: it's not exponential — it's discrete jumps. And you're the one causing them.
Apr 18 2026 · 7′ · Reflections · TypeScript · optimizacion
Architecture, real systems, product decisions and technical lessons. Everything reviewed with my own judgment before it ships.
213 articles published