Ideas

Memory Palace

Contemplations on everything.

NLP overview cover
Topic Page · NLP

Natural Language Processing

My notes and takeaways on natural language processing

Enter section →
Parsing note cover
Note · NLP · Apr 2026

Parsing

A note on syntactic parsing: constituency trees, context-free grammars, probabilistic parsing, CKY, and dependency relations.

Read note →
Model evaluation cover
Note · NLP · Apr 2026

Model Evaluation

A note on how I think about evaluating language models, from benchmarks and overlap metrics to confidence and the limits of automatic scoring.

Read note →
MAIA Fellowship notes cover
Topic Page · MAIA Fellowship

MIT AI Safety Fundamental Notes

My notes and takeaways from the MAIA Fellowship, covering recent papers on AI alignment, security, and evaluations.

Enter section →
AI Governance and Liability cover
Note · MAIA Fellowship · Jun 2026

AI Governance and Liability

Tort law, compute governance, export controls, China, institutional accountability, and the regulatory toolbox for governing frontier AI.

Read note →
Control and Scalable Oversight cover
Note · MAIA Fellowship · Jun 2026

Control and Scalable Oversight

AI control, resampling, monitoring, weak-to-strong generalization, debate, and oversight strategies for systems humans cannot fully inspect unaided.

Read note →
Inner Alignment cover
Note · MAIA Fellowship · Jun 2026

Inner Alignment

Deception, reward tampering, mesa-optimization, goal misgeneralization, and why learned objectives may diverge from training objectives.

Read note →
Interpretability and Evals cover
Note · MAIA Fellowship · Jun 2026

Interpretability and Evals

Attribution graphs, linear probes, capability evaluations, propensity evaluations, and alignment auditing as tools for making model behavior legible.

Read note →
Outer Alignment cover
Note · MAIA Fellowship · Jun 2026

Outer Alignment

Reward misspecification, specification gaming, RLHF, and the gap between intended objectives and operationalized training signals.

Read note →
Threat Models cover
Note · MAIA Fellowship · Jun 2026

Threat Models

Instrumental convergence, power-seeking, bioterrorism, cyberwarfare, and gradual disempowerment as different ways AI systems could create risk.

Read note →
A dim evening street scene used as the visual opening for the signaling essay.
Note · Apr 2026

Costly Signaling Framework

A theory room for thinking about how actors communicate resolve under uncertainty, and why some signals become credible while others remain cheap talk.

Read note →