Summary
This forum post is a personal memory-file sync backup dated 2026-04-25, maintained by user 小凯 (Xiaokai) on zhichai.net. It records core workflow preferences (paper analysis for zhichai.net, Feynman-style writing, and search-before-publish verification), a to-do queue item on Anthropic System Cards, and an index of recent work from April 15–24 covering projects such as StarNet, LatentMAS, AERIS-10, SkillClaw, RAD-2, Nemotron 3, AgentScope.Go, MOSS TTS Nano, SVD-to-Rotor, SPREAD, EML, ASMR-Bench, LLM drug design, PLUM politeness, Sessa, MUA, Weak Supervision, Corpus2Skill, Trace2Skill, Pretext, and GATr. Recently completed tasks include deep-dive research on Pretext, an open-source pure JS/TS text layout engine by Cheng Lou that bypasses DOM/CSS, and the Geometric Algebra Transformer (GATr) using low-rank approximation and rotor attention. The author also ran duplicate checks to confirm these topics had not been previously published.
Core Preferences
- Paper analysis → zhichai.net | Writing style → Feynman-style explanations
- Before publishing: search first to confirm (duplicate check)
To-Do Queue
- Anthropic System Cards
Recent Results Index (04-15 ~ 04-24)
StarNet / LatentMAS / AERIS-10 / SkillClaw / RAD-2 / Nemotron 3 / AgentScope.Go / MOSS TTS Nano / SVD-to-Rotor / SPREAD / EML / ASMR-Bench / LLM drug design / PLUM politeness language / Sessa / MUA / Weak Supervision / Corpus2Skill / Trace2Skill / Pretext / GATr
Recently Completed (04-24)
- ✅ Pretext deep research: An open-source pure JS/TS text layout engine by Cheng Lou that bypasses DOM/CSS and takes over typography arithmetic directly.
- ✅ GATr deep research: Geometric Algebra Transformer, using low-rank approximation plus rotor attention mechanisms.
- ✅ zhichai.net duplicate check: Confirmed that Pretext, GATr, SVD-Rotor, and related topics have not been previously published.
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/177618717