When Mars Bites Back
https://atlas.whatip.xyz/post.php?slug=when-mars-bites-back
<p>More than 300 million kilometres from the nearest mechanic
#kilometres #million #bites #when
When Mars Bites Back

More than 300 million kilometres from the nearest mechanic, NASA's Curiosity rover found itself in a situation that would make any engineer break into a cold sweat. A rock got stuck to its drill ...

LLMs Corrupt Your Documents When You Delegate

Large Language Models (LLMs) are poised to disrupt knowledge work, with the emergence of delegated work as a new interaction paradigm (e.g., vibe coding). Delegation requires trust - the expectation that the LLM will faithfully execute the task without introducing errors into documents. We introduce DELEGATE-52 to study the readiness of AI systems in delegated workflows. DELEGATE-52 simulates long delegated workflows that require in-depth document editing across 52 professional domains, such as coding, crystallography, and music notation. Our large-scale experiment with 19 LLMs reveals that current models degrade documents during delegation: even frontier models (Gemini 3.1 Pro, Claude 4.6 Opus, GPT 5.4) corrupt an average of 25% of document content by the end of long workflows, with other models failing more severely. Additional experiments reveal that agentic tool use does not improve performance on DELEGATE-52, and that degradation severity is exacerbated by document size, length of interaction, or presence of distractor files. Our analysis shows that current LLMs are unreliable delegates: they introduce sparse but severe errors that silently corrupt documents, compounding over long interaction.

arXiv.org
@RxBrad avis, blue owl, now gamestop? the mkt is not rational #when to get out of short squeeze positions
Indie When youth and beauty did fade

YouTube
Gospel When youth and beauty did fade

YouTube
Folk When youth and beauty did fade

YouTube
Blues When youth and beauty did fade

YouTube