Threads
Four questions the work keeps returning to. Papers, essays, and projects are filed by what they are, and linked by what they ask.
Platforms and mediation
When a platform or an answer engine supplies the first account of an issue, what happens to attention, disagreement, and everyday life.
- Aug 2026 Preprint When AI Becomes the First Pass: Source Selection, Conflict Retention, and Novel-Case Classification in AI-Mediated Interpretive Pipelines
- Mar 2026 Essay Occupied Lives: Platform Convenience and the Dissipation of Social Energy
- Mar 2026 Essay Generative Search and the Platformization of Sensemaking
Agents and audit
How AI agents come to act in the world, and whether anyone can reconstruct afterwards what they did.
Models and evaluation
Small, measurable questions about how language models change, and the tools that make local experiments possible.
- Sept 2026 Software gguf2mlx-stream: Bounded-Memory GGUF to MLX Checkpoint Transcoding
- Aug 2026 Preprint When AI Becomes the First Pass: Source Selection, Conflict Retention, and Novel-Case Classification in AI-Mediated Interpretive Pipelines
- Nov 2025 Study Small Factual Updates in LLMs: LoRA-SFT / LoRA-DPO Evaluation Study
Interactive worlds
Characters, scenarios, and serial stories made with AI and held in place by designed constraints.