Polybee perspectives
Insights.
Research, markets and context.
Ideas to understand AI
and participate in building it.
Original research
A critical reading of the results, architectural choices and open questions in Polybee’s research.
Fitting the examples is not enough: the generalization challenge in F3-AGI v4.0
Our first research study opens a discussion about choosing rules, measuring transfer and recognizing what a benchmark result still leaves unresolved.
Read articleOriginal research
Continual learning, retention and cost: examining the limits and possibilities of selective adaptation.
Learning without updating everything: memory, plasticity and cost in F3-AGI
How to decide when to learn, preserve what has already been learned and measure the cost of that choice—while distinguishing an algorithmic metric from electricity consumption.
Read articleOriginal research
Agents that investigate: distinguishing progress, information gain and unproductive repetition.
Exploring with purpose: action, information and memory in F3-AGI
An agent needs to discover what is worth trying—and recognize when repetition no longer teaches it anything. The interactive study raises questions about exploration, memory and evaluation.
Read articleAGI and research
Research, progress and open questions about general intelligence.
AGI in 2026: how to interpret advances in AI agents
Agent progress raises a more precise question: under what conditions does a demonstrated capability become reliable?
Read articleGlobal market
The forces shaping AI infrastructure and industry.
The global AI market in 2026: who controls the capability to build?
Access to models expands possibilities. Infrastructure and dependency decisions determine how much control builders retain.
Read articleAI in Brazil
Local capability, Portuguese and building from Brazil.
AI in Brazil: Portuguese, evaluation and the capability to choose
Building from Brazil means turning language, context and knowledge into concrete research and engineering decisions.
Read article