IDEAS. EVIDENCE. POSSIBILITIES.
A closer look
at what comes next.
Research worth understanding. Questions worth testing. Explore our thinking on intelligence, the systems around it, and the work of making it useful.
7 research analyses2 company explainers
FEATURED ANALYSISTraining · 3 MIN READ
GRPO efficiency starts before the kernel
A practical investigation of sparse rewards, loss normalization, and what a faster training step actually proves.
Read the analysis ↗Explore the journal
9 stories · including the featured analysis above
AgentsA coding agent is a system, not a score
How to compare coding agents without confusing model quality, scaffolding, test contamination, and retry budget.
RoboticsRobot foundation models need a data contract
What GR00T and π₀ teach us about the gap between a general model and a reliable policy on a particular robot.
Spatial AIA map should know what it does not know
A proposed architecture for LiDAR–camera mapping that keeps geometric confidence separate from semantic guesses.
RetrievalRAG should earn the right to answer
A claim-level design for retrieval, contradiction handling, citations, and useful abstention.
SafetyPrompt injection is a permission problem too
Why an agent’s ability to read a document should not grant that document authority to authorize an action.
SystemsLong context has a memory budget
A worked KV-cache calculation, and why FlashAttention, GQA, and PagedAttention solve different problems.
CompanyInside the WEBSPACEAI Trust Center
A reader’s guide to the policy directory, its cross-references, and the questions the documents are intended to answer.
CompanyHow our safety risk levels work
The published AI Safety Framework describes four risk levels. Here is how to read the escalation and release controls without confusing a policy with a test result.
HOW WE WRITE
Evidence before
a headline.
Our research analyses distinguish published findings, our interpretation, and experiments still to be run. Worked calculations are identified as calculations. A proposed method is not presented as a measured breakthrough.
Company explainers link to the underlying website documents. Sources and the scope of the work are listed in each article.
Download the analytical examples ↗Follow the RSS feed ↗