Notes on Oxygen AI, memory systems, intent, drift, interfaces, and safer agentic software.
What if adapting a model did not mean rewriting its weights? WarpKernels treat memory as geometry: freeze the backbone, learn a tiny local metric, and tilt the representation space.
Next.js 16 makes the Rust-based toolchain feel less like an experiment and more like the default path for modern React apps.
A grounded look at why Signal feels different from other messaging apps: protocol design, metadata reduction, defaults, and tradeoffs.
A reading note on Lee C. Loveridge's geometric interpretation of the Riemann tensor, Ricci tensor, scalar curvature, and why physical intuition matters.
Sapient's HRM-Text-1B is a useful signal: architecture and training objective can still move the efficiency frontier, not just model size.
The next useful AI systems will not just be more capable. They will be easier to inspect, correct, and trust while they work.
The more capable AI systems become, the more their interfaces need to expose memory, intent, risk, and control with calm precision.
Compute keeps expanding, but the next useful frontier is not just bigger clusters. It is memory, agency, safety, and measurement.
A practical reading map for intent learning, grokking, memory, drift, and the safety problems that appear when agents persist over time.
Entropy, drift, attractors, and signal velocity are not just metaphors. They are useful tools for thinking about memory, intent, and safer agents.
A new home for notes on Oxygen AI, memory systems, intent, alignment, and the craft of building safer frontier software.