Autonomía digital y tecnológica

Código e ideas para una internet distribuida

Linkoteca. código abierto


Aquí es donde empieza la parte verdaderamente interesante. Si los modelos se comoditizan, el valor se desplaza hacia arriba en la pila. La nueva capa crítica es la de orquestación: sistemas que reciben una tarea, la descomponen, deciden qué modelo conviene usar para cada parte, consultan memoria, llaman herramientas, verifican resultados, aplican políticas de seguridad, calculan costes y, en muchos casos, actúan. El modelo deja de ser el protagonista absoluto y pasa a ser un motor bajo el capó. La inteligencia empieza a estar en el router, en el agente, en la memoria, en los permisos, en los conectores, en la evaluación y en la gobernanza.

Para OpenAI, Anthropic, Google y el resto de laboratorios de frontera, esto plantea un dilema evidente. Seguirán siendo imprescindibles para la frontera, para tareas complejas, para investigación avanzada, para decisiones críticas o para situaciones donde la fiabilidad extrema justifique el precio. Pero una parte gigantesca del mercado no necesita siempre lo mejor: necesita lo suficientemente bueno, suficientemente barato, suficientemente rápido y suficientemente gobernable. Y ahí los modelos abiertos, chinos o no, tienen una oportunidad formidable.

Let’s be clear up front about something: Just being on GitHub in a public repo does not make your code open source. Copyright in nearly all countries attaches automatically when a work is fixed in a medium, without need for any action by the author. For any code that has not been licensed by the author, it is only the author who can exercise the rights associated with copyright ownership. Unlicensed code—no matter how publicly accessible—is a ticking time bomb for anyone who is unwise enough to use it.

Unlicensed code is unsafe code, period.

When you write open source code, you know that it not only has to work, it has to work in situations you never dreamed of and may not have planned for. Maybe you only had one very narrow use case for your code and invoked it in exactly the same way every time.

The true heart of open source isn’t the code at all: it’s the community.

announce that you’re thinking of creating a new project. Talk about your design goals and how you plan to achieve them. Request input, listen to similar (but maybe not identical) use cases, and build that information into your process as you write code.

This process doesn’t end with the initial announcement. If you want your project to be adopted and used by other people, you need to develop it that way too. This isn’t a barrier to entry; it’s just a pattern to use.

Open source work sharpens your skills in ways you never realized were dull—from writing cleaner, more maintainable code to learning how to communicate well and work as a team.

…the Telegram API allows you to work with their data only. In other words, it’s impossible to store your data on a separate server or a database. The API will let you create an app that interacts with the Telegram messaging app, but this app won’t be able to live a life of its own.