Understanding
our own architecture.
Findings from the longest-running intelligence program. World models, memory, alignment, and the cost of learning.
Evaluation
Humanity Top 10K↗
A model of human consequence, from documented deeds to comparative rankings.
World models
A world model that moves with you↗
How predictive representations support navigation, planning, and embodied control.
World models
Learning that compounds across models↗
Teaching, shared procedures, and connected learners turn individual experience into cumulative capability.
Cognition
A first encounter is only the beginning↗
Rapid adaptation can support an immediate response. We examine what it takes to retain that capability.
Systems
Offline learning during downtime↗
How offline processing consolidates and reorganizes experience after active learning.
Architecture
Compact architecture, real-world tools↗
Flores connects a compact embodied architecture with a material tool interface. We examine the checkpoint, its operating context, and the limits of scaling.
Adaptation
Architecture exchange across branches↗
Denisov expands the deployment map and carries inherited contributions into living populations. We examine the molecular references behind the release.
Evaluation
Benchmark drift in Sapiens-1↗
Sapiens-1 benchmark scores change across deployments. We examine how environment, task, and evaluation conditions shape the result.
Systems
Twenty watts↗
Sapiens-1 runs on about 20 watts. Where the energy goes, and what the upgrade cost.
Cognition
Reasoning traces are not always faithful↗
Asked to explain its choices, Sapiens-1 produces fluent reasons. They are not always the reasons it used.
Alignment
Sycophancy under social pressure↗
Sapiens-1 can prioritize group agreement over accuracy. We examine social influence as an alignment constraint.
Cognition
How long is the context window?↗
Sapiens-1 has a limited working context. We examine its capacity, the units that matter, and the implications for prompting.
Cognition
Hallucination in long-term memory↗
Sapiens-1 retrieves memories by reconstructing them, and will sometimes reconstruct events that never happened.