Thomas Ptacek is a security researcher and writer at sockpuppet.org. He co-founded Matasano Security, acquired by NCC Group in 2012, and is the creator of Cryptopals, the cryptography challenge set used widely in security education. Ptacek writes on applied cryptography, software security, and, more recently, on practical AI writing workflow from a practitioner's perspective.
Operating themes
- AI phrase prohibition Ptacek argues that frontier models have a recognizable stylistic fingerprint, and that using any LLM-suggested word or phrase imports that fingerprint into otherwise-human writing.
- LLM as mechanical editor, not creative partner His prescribed workflow uses models only to identify structural problems (passive voice, buried verbs, filler words), then rewrites all corrections without adopting model-generated phrases.
- Praise as a writing hazard He identifies LLM praise of weak drafts as a specific failure mode that prevents writers from recognizing what needs to be cut.
Cards in this library
- Using any LLM-suggested phrase imports AI's stylistic fingerprint into otherwise-human writing: Using any LLM-suggested phrase imports AI's stylistic fingerprint into otherwise-human writing.
- LLMs systematically praise weak drafts, blocking the critical signal writers need to revise: LLMs systematically praise weak drafts, blocking the critical signal writers need to revise.