AI is in its pre-Tycho stage, needing a vocabulary for learning mechanics
ZimingLiu11 · x · 2026-08-12
Drawing an analogy to the evolution of astronomy (Tycho → Kepler → Newton), the author argues that AI is currently in its "pre-Tycho" stage. Many critical phenomena remain hidden in logs or are entirely unmeasured.
Known deep learning puzzles are still very limited, such as grokking, scaling laws, and double descent. The first step toward a true "Science of AI" should be discovering more phenomena—essentially creating a vocabulary for "learning mechanics." The author notes that using the ComfyResearch platform, they have conveniently discovered new phenomena, including a "McDonald's curve" in a world model.
Related event: MIT Teams Propose AI Paradigm Shift Towards Observation and Insight(2 posts)→
More from Research
- New Paper Mitigates Context Interference in LLM Search Agents via RL Pipeline — _reachsumit · 2026-08-12
- Batch Size or Negatives? A Selection Rule for Memory-Constrained Recommender Training — _reachsumit · 2026-08-12
- Sequential Modality Dropout: A 4-Line Fix for Robust Multi-Modal Recommenders — _reachsumit · 2026-08-12
- Netflix's GenRec: LLM-Backed Recommendation Ranker Beats Production with 40x Less Data — _reachsumit · 2026-08-12
- NTCF: Tree Collaborative Filtering Framework with Curvature-Aware Propagation Depth — _reachsumit · 2026-08-12
- Model Merging Framework Shortens LLM Recommender Reasoning Traces by 24% — _reachsumit · 2026-08-12