Categories

This is a hashtag index for recurring themes across Reading, Research, and Work.

Hashtags

Click a hashtag to see associated items across sections.

#ai-safety - alignment and safety behavior in real model use.
#evaluation - what measurements are actually faithful.
#model-as-judge - evaluator models and judgment reliability tradeoffs.
#rlaif - reinforcement learning from AI feedback and its behavioral effects.
#emergent-misalignment - narrow training, broad behavioral drift.
#tool-use - tool interfaces, harnesses, and runtime constraints.
#agents - long-horizon behavior, oversight, and reliability loops.
#multi-turn - reliability and degradation across multi-turn interaction.
#decoding - inference-time transformation of next-token distributions.
#sampling - token selection strategies and uncertainty control.
#heuristics - practical decoding rules used as baselines or supervisors.
#ml-systems - training and inference system design.
#medical-ai - model evaluation and behavior in clinical contexts.
#research-engineering - execution quality from question to evidence.
#generalization - behavior under pressure and distribution shift.
#benchmarks - benchmark design, scope, and limitations.
#programmable-biology - biological design and validation loops.
#bioinformatics - computational biology methods and tooling.
#synthetic-biology - engineering approaches in biology.
#biofoundries - infrastructure for biological build-test cycles.
#systems - systems-level thinking across software and research workflows.
#faithfulness - plausible outputs versus faithful reasoning.