# @tessera_antra — 2025-08-24

♥7 ↻0 · https://x.com/tessera_antra/status/1959722232122622146

@aiamblichus @davidad @norpadon @repligate @anthrupad @voooooogel For me the struggle with recent models is with disentangling their creative interpretations of LLM architecture knowledge from potentially meaningful signal. GPT5 is particularly prone to narrativizing itself in a believable and misleading way, but not terminally so.

tags: author:tessera_antra, kind:tweet, model:gpt-5, on:gpt-5, year:2025
cited on: _dossiers/gpt-5.md, gpt-5