Guardamos checks AI-generated fitness and training plans against published sports-medicine guidelines before they reach the people who follow them. General-purpose AI assistants are good at producing plausible-sounding training programs, but they don't reliably account for context that matters clinically — such as a long break from training, which changes what's an appropriate starting point. Missing that context is a documented factor behind real injuries. Guardamos doesn't generate plans or claim to guarantee safety. It's a guardrail and explainability layer: it flags what a plan is missing, explains why in plain language sourced back to the relevant guideline, and — when a plan is appropriate — explains why that's the case too. The name comes from the Portuguese (and Spanish) verb "guardar" — to protect, to keep. Early-stage, solo-built, open about its rules and its limitations. Beyond catching hallucinations, Guardamos is built around the kind of explainability and audit-trail requirements that regulations like the EU AI Act are pushing toward — every verdict traces back to a specific guideline, not a black box. That lets the AI teams we work with focus their own effort on what they do best (the generative, probabilistic side of their product) instead of building this layer themselves. guardamos.dev