The blog is not the final product. It is the open laboratory: a place to expose assumptions early, connect literatures that rarely meet, and let strong ideas accumulate enough evidence to become research.
Active research directions
Open-ended evolutionary search
What measurable conditions let novelty keep accumulating instead of collapsing into a fixed objective?
Target output: survey essay → benchmark → preprint
Morphogenesis as distributed inference
Can error correction, bioelectric coordination, and generative models share a useful formal language?
Target output: concept map → simulation → position paper
Scale-free agency
Which operational tests distinguish an agent from a stable pattern at cellular, organismal, and social scales?
Target output: taxonomy → thought experiments → workshop paper
Evidence ladder
- Observation — record a surprising phenomenon and the competing explanations.
- Synthesis — connect the primary literature and state a narrow, falsifiable claim.
- Artifact — publish code, data, simulations, and negative results.
- Preprint — freeze the claim, methods, limitations, and reproducibility package.
- Peer review — submit where the actual contribution belongs, not where the topic merely sounds fashionable.
Research integrity
Every substantive piece will distinguish evidence from speculation, link primary sources, disclose important revisions, and publish reproducible artifacts when computation is involved. This is independent work; nothing confidential or proprietary to NVIDIA belongs here. Employer review and disclosure requirements come before any public research release.
If you work on artificial life, evolutionary computation, morphogenesis, complex systems, or mathematical models of agency, send a note. Concrete criticism and disconfirming evidence are especially welcome.