Tommy

近况Now

灵感来自 nownownow.com — 我注意力的快照。 Inspired by nownownow.com — a snapshot of where my attention is.

更新于Updated 2026-05-02

当前在动的事。

在做

  • 这个站点(Astro 6 + Tailwind 4 + Obsidian 当 CMS)。目前在快速迭代 期——搜索、反向链接、wikilinks、更聪明的 404、默认 OG 卡、 标题锚点都在过去一周落地。见 Obsidian → Astro publish pipeline
  • EEG attention trainer (concept) —— 慢炖的副业。已经把 Muse 2 上的每用户基线校准回路跑通了;下一步是双 N-back 的任 务引擎。

在读

  • 《主人与他的使者》(重读,部分;McGilchrist)。笔记落在 书房
  • The Beginning of Infinity(Deutsch)。第一遍。
  • 《死神永生》(重读,慢)。大概是第三遍。
  • 在认真重过 Friston 2010 的那篇 (书房条目)。

在想

  • Transformer 里的 “attention” 与认知神经科学里的”注意”能不能 撑起一个严肃的类比,还是只是一个有用的类比。 (Vaswani et al. 2017 vs Attention as relation, not state。)
  • 当指标与目标分离时,EEG 驱动的注意训练的伦理。
  • 为什么用 LLM 辅助写作降低了我对自己草稿的信任——以及为什么 那是一个健康的信号。(为什么写作。)

想找通信对象的开放问题

  • 有没有人在任何闭环生物反馈装置上做长期 Goodhart-aware 的奖励 设计?
  • 有没有不依靠 McGilchrist 那套文化叙事的、对半球特化论 的严肃哲学辩护?
  • 当代对 Friston 自由能原理的最佳批评——不是”它是同义反复”那种 ——有哪些?

如果这里有你的领域,欢迎从 about 页联系我。

What’s currently active.

Building

  • This site (Astro 6 + Tailwind 4 + Obsidian-as-CMS). Currently in rapid-iteration mode — search, backlinks, wikilinks, smarter 404, default OG card, heading anchors all landed in the last week. See Obsidian → Astro publish pipeline.
  • EEG attention trainer (concept) — slow-cook side project. Got the per-user baseline calibration loop working with the Muse 2; next is the dual-N-back task engine.

Reading

  • The Master and His Emissary (re-read, partial — McGilchrist). Notes landing on the library.
  • The Beginning of Infinity (Deutsch). First read.
  • 死神永生 (re-read, slow). Probably the third pass.
  • Working through Friston’s 2010 paper carefully again (library entry).

Thinking about

  • Whether “attention” in Transformers and “attention” in cognitive neuroscience can sustain a serious analogy or just a useful one. (Vaswani et al. 2017 vs Attention as relation, not state.)
  • The ethics of EEG-driven attention training when the metric and the criterion diverge.
  • Why writing with LLM assistance has lowered my trust in my own drafts — and why that’s a healthy signal. (为什么写作.)

Open questions I’d love correspondents on

  • Anyone running a long-term Goodhart-aware reward design in any closed-loop biofeedback rig?
  • Is there a serious philosophical defence of strong hemispheric specialisation accounts that doesn’t lean on McGilchrist’s cultural narrative?
  • Best contemporary critique of Friston’s free-energy principle that isn’t just “it’s tautological”?

Reach out via the about page if any of these are your turf.