How the “You might also like” cards on this blog get made
Not Algolia, and not a chat that invents links. A local embedding, a JSON file, and Hugo. The same numbers draw the blog map.
Leer artículoRead article
GPT-6 Astra: the day nobody answered
OpenAI ships GPT-6 Astra: anything you can do on a computer. This afternoon ChatGPT, Claude and Grok went down together. The 6, the price, and what doesn’t change.
Leer artículoRead article
Fable 5.1: smarter, cheaper, more leash
Anthropic ships Fable 5.1 and Mythos 5.1: the same model, two locks. Cache gets cheaper, science jumps, and the leash gets more precise. A 5.1 doesn’t bring me back to Max.
Leer artículoRead article
They called a cut a raise
Anthropic announces a permanent +25% on Claude Code. Versus today that’s −17%. The headline, the second tweet, and why it fits cancelling Max.
Leer artículoRead article
OpenAI is leaving Cursor
OpenAI cuts direct access to Cursor on November 12. SpaceX bought the editor. The statement, the replies, and why the editor is no longer a Swiss Army knife.
Leer artículoRead article
Today Cursor opened a GitHub. The usual one was down
Cursor’s Origin forge is out in beta. The same day GitHub had the API half-down and Copilot in red. What it is, what it isn’t, and why I’m going to try it.
Leer artículoRead article
I cancelled Claude Max. Not because of the model
Anthropic’s latest moves —watermarks on text and files— were the last straw. I’m sticking with Grok, for now.
Leer artículoRead article
The EU cooked up another dumb idea. Now it’s watermarks on AI text
Invisible watermark + C2PA metadata on the output of providers who play by the rules. On paper: transparency. In reality: it doesn’t work, punishes compliance, degrades the product, and creates false security. Another stone in the road.
Leer artículoRead article
I’ve started coding with Grok (after months on Claude)
I’ve spent months coding with Claude. This week I put Grok Build into the daily mix: what feels different, what I still do with Claude, and a menu bar app I built so I’m not blind about credit usage.
Leer artículoRead article
I cloned my voice for the podcast, and found out TTS engines swallow whole sentences
I went looking for a cheaper TTS for my podcast and ended up cloning my own voice. Along the way I found out that voice generators skip entire sentences from the script and fill the gap with gibberish.
Leer artículoRead article
They switched Fable 5 back on: the leash is still on
After 19 days switched off by a US government order, Fable 5 is back. With an apparently happy ending that, looked at closely, confirms the very lesson of the shutdown.
Leer artículoRead article
Claude Sonnet 5: what it brings and when to pick it
Anthropic ships Claude Sonnet 5: 1M context, launch pricing of $2/$10 and performance close to Opus. What changes and when to pick it over Opus 4.8 and Haiku 4.5.
Leer artículoRead article
I moved my app's AI from Claude to Apple's cloud
My fitness app used the Claude API for all its AI. Apple is giving small developers its cloud AI for free, so I spent these days swapping Anthropic for Apple's models. What I migrated, what I didn't, and what I learned.
Leer artículoRead article
They switched off Fable 5: the tool on a leash
Three days ago Fable 5 was the most capable AI model available to the public. Today it's gone: the US government ordered it switched off. What that means for anyone building on top of it.
Leer artículoRead article
Why I shut down my daily AI podcast
Automating a daily AI news show was the easy part. Keeping it relevant wasn't. Why I shut it down, how to close a podcast properly, and what I learned.
Leer artículoRead article
How I turned my podcast into YouTube video without recording anything
I built a pipeline that turns every audio episode into a YouTube video —animated cover, waveform and synced subtitles— with no camera and local tools. Here's how, step by step.
Leer artículoRead article
Claude Opus 4.8 lands: better at code, more honest and with sub-agents in parallel
Anthropic released Claude Opus 4.8 today at the same price as 4.7. It improves agentic coding, reasoning with tools and knowledge work, and ships with dynamic workflows for sub-agents in parallel.
Leer artículoRead article
How I code in 2026: my stack in the age of agents
My real programming stack in 2026: a Mac mini at home, Claude Code as the editor, plain PHP, Hugo, git hooks that deploy by themselves. And the other side: how I write Moodle plugins at UDIMA, where I do touch code line by line.
Leer artículoRead article
Gemini 3.5 Flash: speed over intelligence for AI agents
Gemini 3.5 Flash is 4× faster and 10× cheaper. Why speed matters more than intelligence for AI agents, and why it will end up powering Siri.
Leer artículoRead article
An AI-generated podcast: how we built it and what we've learned
How we built El podcast de Sergio and El informativo using AI for the entire process: scripting, voices, and automatic publishing to Apple Podcasts.
Leer artículoRead articleFrom Recorded Class to Moodle Quiz in One Click
mod_jitsi automatically converts a recorded class into a quiz, summary and timestamped transcript using Gemini 2.5 Flash. No copy-pasting, no extra work.
Leer artículoRead article
What's new in the Jitsi plugin v5.0
mod_jitsi 5.0 introduces a Developer Portal with opt-in telemetry, a full attendance report with recording heatmap, real-time presence and email invitations from the activity.
Leer artículoRead articleWhat's new in the Jitsi plugin v4.6
mod_jitsi 4.6 brings scalable recording with a pool of Jibris on Google Cloud, AI-generated summary/quiz/transcript (Gemini 2.5 Flash), improved private sessions and a brand-new tutoring-hours system.
Leer artículoRead article