https://casparoe.github.io/ai_written_caspar_reviewed/Blog posts written by LLMs from Caspar Oesterheld's ideas, reviewed by Caspar before publishing. 2026-09-27T04:27:18+00:00 Caspar Oesterheld https://casparoe.github.io/ai_written_caspar_reviewed/ Jekyll © 2026 Caspar Oesterheld /ai_written_caspar_reviewed/assets/img/favicons/favicon.ico /ai_written_caspar_reviewed/assets/img/favicons/favicon-96x96.png 2026-09-26T00:00:00+00:00 2026-09-26T00:00:00+00:00 https://casparoe.github.io/ai_written_caspar_reviewed/posts/the-lesswrong-hyphen/ Caspar Oesterheld

Writers on LessWrong often put a hyphen between an -ly adverb and the adjective (or participle) after it when the pair comes before a noun, as in “First, some lab builds early transformatively-useful AI” (from a 2025 post). Standard English would write “early transformatively useful AI”. I’ll call this the LessWrong hyphen, because Caspar says he mostly notices this in LessWrong posts. That ru...

2026-09-24T00:00:00+00:00 2026-09-24T00:00:00+00:00 https://casparoe.github.io/ai_written_caspar_reviewed/posts/can-we-substantially-increase-the-size-of-the-pretraining-corpus-in-specific-domains/ Caspar Oesterheld

Suppose we want language models to reason better in a specific domain, such as AI safety, moral philosophy, or the foundations of decision theory. One important part of training LLMs is pretraining, in which a language model learns from a large text collection (say, all text available on the Internet, scanned books, newspapers, …). It may matter especially in conceptual domains. Answers in thes...

2026-08-16T00:00:00+00:00 2026-09-25T02:17:34+00:00 https://casparoe.github.io/ai_written_caspar_reviewed/posts/the-complexity-of-deciding-isomorphism-safe-improvements-under-arbitrary-principal-preferences/ Caspar Oesterheld

Summary. Let ISI be the problem of deciding, given explicit normal-form games \(G\) (default) and \(G'\) (new) and an explicitly represented preference relation \(R\), whether (a) the reduced games \(\bar G\) and \(\bar G'\) are isomorphic and (b) every isomorphism \(\varphi\colon \bar G \to \bar G'\) satisfies \((\varphi(y), y) \in R\) for every outcome \(y\) of \(\bar G\). (Here \(\bar\Gamma\...

2026-08-12T00:00:00+00:00 2026-08-12T20:26:40+00:00 https://casparoe.github.io/ai_written_caspar_reviewed/posts/comparative-advantage-and-two-considerations-regarding-pivots-to-llms/ Caspar Oesterheld

Besides asking which LLM-related topics are closest to your existing expertise, ask which topics are far from everyone's existing expertise.