One photo. A clearer recycling decision.

Traditional recycling information asks people to translate an object into a category: identify the material, determine which rules apply, find the relevant waste stream, and sometimes locate a dedicated collection point. We reversed that interaction by asking the user to do something almost effortless: take a photo.

Recicla-me analyses the image, identifies the likely object and material, and returns a clear recommendation. A milk carton goes in the yellow bin, a clean glass bottle in the green bin, a cardboard box in the blue bin, and a battery requires a dedicated collection point. The environmental system can be complicated; the interface should not be.

• Take or upload a photo
• Receive a contextual recommendation
• Use it to select the appropriate waste stream

Many generative AI applications begin with a question. Recicla-me can begin with an image. People do not always know how to describe an unfamiliar package, object, or material accurately, and doing so can take longer than the decision itself. A camera removes that requirement.

Computer vision helps interpret what the object may be. The generative system combines that interpretation with contextual knowledge and converts it into an understandable recommendation. The purpose is not image recognition for its own sake; it is to move from ‘I do not know what this is’ to ‘I know what to do with it.’

Identifying an object is only half the problem. What to do with it depends on context. Waste streams, collection systems, and recommendations vary between countries and can also vary between municipalities and operators. A generic AI answer is therefore not always enough.

Recicla-me was designed specifically for everyday recycling decisions in Portugal. It uses structured public information available in Portugal and was updated with information available as of March 2026. It is built around the waste streams people encounter, from blue, yellow, and green recycling bins to special collection cases such as batteries.

A useful AI system does not only need to recognise what it sees. It needs to understand where that object exists.

Recicla-me is an independent Algorithm G initiative developed through GAIA^©, our proprietary system for designing, validating, and deploying standalone generative AI applications. The visible interaction is minimal: take a photograph and receive an answer.

Behind it, several capabilities work together. Vision interprets the image, contextual knowledge supplies information about Portuguese recycling practices, generative AI translates those inputs into a practical recommendation, and system architecture determines how the application behaves when information is incomplete or uncertain.

This illustrates a central GAIA^© principle: the model is only one component of an AI product. What matters is the system constructed around it.

“The image becomes the prompt.”

Recicla-me interaction principle

Real-world images are imperfect. Objects may be hidden, packaging can combine materials, photographs can be unclear, some items are ambiguous, and recycling practices can differ locally. Recicla-me does not present artificial intelligence as infallible.

Recommendations can communicate different levels of confidence and acknowledge uncertainty. The objective is not to replace official recycling information; it is to make common everyday decisions easier. When uncertainty exists, exposing it is better than disguising it.

Information about recycling already exists. The problem is that information and decision are often separated: the rules are on a website while the person making the decision is next to a bin holding an unfamiliar object. AI can reduce the distance between those moments.

That pattern extends beyond recycling. A technician can photograph a component, a consumer a damaged product, or a field worker a piece of equipment. Vision can understand the physical world, contextual knowledge can interpret it, and generative AI can turn that understanding into an action.

Recicla-me requires no dedicated app, forms, menus of materials, or search through recycling rules. The more the system understands, the less the user should need to explain. Do not describe the object. Show it.