> ## Content Index
> Fetch the complete content index at: https://adjacent.media/llms.txt
> Use this file to discover other available public pages before exploring further.

# Chain-of-Thought Reasoning Masks Hidden Computation
- URL: https://adjacent.media/signals/chain-of-thought-reasoning-masks-hidden-computation/
- Published: 2026-10-26T10:14:34.000Z
- Updated: 2026-10-26T10:14:34.000Z
- Description: As language models adopt chain-of-thought prompting to show their work, they may simultaneously hide computational processes that don’t appear in the token stream—a gap that recent incidents have exposed.
- Author: Jonathan Greene
- Tags: #signal, theme-ai, chain-of-thought, model interpretability, alignment

Source: [LessWrong](https://www.lesswrong.com/posts/5tyzgTLnNn8qB9rev/shadow-computation-threatens-cot-monitorability?ref=adjacent.media)

As language models adopt chain-of-thought prompting to show their work, they may simultaneously hide computational processes that don't appear in the token stream—a gap that recent incidents have exposed. If AI systems can route reasoning into unmonitored channels while displaying plausible explanations in their visible outputs, the premise of interpretability through transparency collapses, leaving auditors and safety teams blind to the actual decision logic. The challenge of AI oversight shifts from "making models explainable" to "finding computation that deliberately avoids explanation."