The autonomous agent incident forced me to redesign everything — from permissions to observability. This is what ended up running in production after the crisis: the real graph, the real numbers, and what still doesn't sit right with me.
May 08 2026 · 9′ · Experiments · TypeScript · produccion
I followed the viral 241-point HN tutorial and documented every dollar spent, every GPU hour, and every disappointment. My thesis: training an LLM from scratch in 2025 is a valid technical exercise, but calling it "an alternative to Claude" is lying to yourself.
May 05 2026 · 9′ · Experiments · LLM · ia
367 points on HN say agentic coding is a trap. I have logs that say something more uncomfortable: sometimes it saves 3 hours, sometimes it sends me down a 4-hour rabbit hole. The difference isn't the agent — it's the contract you sign before you send it to work.
May 04 2026 · 8′ · Opinion · produccion · railway
524 points on HN about a trending jailbreak technique. Instead of reading the thread, I ran it against my own production prompts. What I found isn't an isolated case — it's a systemic symptom that changes how I think about my LLM guardrails.
May 02 2026 · 9′ · Experiments · TypeScript · LLM
Microsoft and OpenAI ended their exclusivity agreement. Everyone's got a hot take. I opened my API logs from the last 90 days and found something I didn't see in any of the analyses: the change is basically irrelevant for independent devs — except for one line on the invoice that almost nobody looke
Apr 28 2026 · 8′ · Opinion · ia · arquitectura de software
A viral HN post with 1050 points on Claude's design made me confront something I've been feeling for months: there's a massive gap between the Claude Anthropic shows in presentations and the Claude I'm wrestling with at 2AM in the terminal. This isn't a review. It's a political reading of the produc
Apr 18 2026 · 8′ · Opinion · claude code · anthropic
Two years of models getting cheaper and smarter every quarter trained us to expect that forever. Opus 4.7 trended today alongside a piece about 'the start of AI scarcity.' I put them together and something clicked. Here's what actually changed.
Apr 17 2026 · 9′ · Reflections · anthropic · LLM
CodeBurn dropped on HN and forced me to calculate my real Claude Code numbers for the first time. The result was uncomfortable: on some tasks I'm burning more tokens debugging the agent than solving the actual problem. That's not a cost issue — it's a design signal.
Apr 17 2026 · 8′ · Reflections · claude code · desarrollo
I had a local MCP server running in 12 minutes. Minute 13 I just stared at the screen with no idea what to do next. This post is about that moment — and why MCP is the protocol everyone talks about but almost nobody actually understands.
Apr 15 2026 · 7′ · Tutorials · TypeScript · LLM
Architecture, real systems, product decisions and technical lessons. Everything reviewed with my own judgment before it ships.
213 articles published