A QA, day by day

Push alerts on mobile with self-hosted ntfy

I added push notifications to my phone for ScamDetector with self-hosted ntfy. Four real-time alerts (injection, ban, brute-force, backend down) and three periodic monitors (daily digest, disk usage, traffic spike). A 90-line module with no dependencies.

13 min04/17

Iterating on ScamDetector, what I changed after publishing

What changed in ScamDetector after publishing. A third AI backend, persistent rate limiting, privacy-friendly logging, action steps by scam type, sharing results as an image, and PWA support.

14 min04/16

Hardening ScamDetector against prompt injection, hallucinations, and abuse

Defense against prompt injection, prevention of model hallucinations, layered rate limiting, and the rest of the changes that hardened ScamDetector for real production use.

15 min04/15

ScamDetector's architecture, an AI proxy that doesn't expose secrets

How ScamDetector is built under the hood, from the server-side proxy that protects API keys to the three interchangeable AI backends and the security measures.

8 min04/14

X-ray of the QA market in Spain in 2026

Between 300 and 500 active QA openings, a 45% salary gap between manual testing and automation, and 70% of IT professionals planning to change companies. A full X-ray of the QA market in Spain with data from Glassdoor, Manfred, Hays, and Robert Walters.

9 min04/13

ScamDetector, an AI-powered scam detector

ScamDetector combines artificial intelligence, phone reputation lookup, and URL scanning to help you spot digital scams. No signup, no stored data.

9 min04/12

Defects cluster: the 80/20 principle applied to testing

80% of bugs are concentrated in 20% of the code. Understanding the defect clustering principle lets you focus testing where it actually matters.

8 min04/11

How I added AI voice narration to blog posts

Every blog post can be read aloud with a built-in player. The audio is generated with AI the first time someone asks for it, then cached for everyone else.