Serguey Asael Shinder
A software engineer focused on LLM integration engineering practices, who has written technical articles on the problem of large language model output truncation and defensive strategies against it.
Timeline (last 90 days)
The original author, Serguey Asael Shinder, believes that the real bug is not the leak itself, but entrusting a 30% annual discount to a single sentence (the prompt) to guard.
The solution given by author Serguey Asael Shinder is to reorder the prompt by content change frequency: content that never changes goes first, content that changes per session goes in the middle, and content that changes on every request goes last.
Mechanically splitting documents by a fixed token count causes headings and the paragraphs they belong to to be placed in different chunks, stripping the paragraph of key disambiguating information such as which package it belongs to
Author Serguey Asael Shinder recommends reading the stop reason after every LLM call, just as one would read an HTTP status code.
All Facts (4)
The original author, Serguey Asael Shinder, believes that the real bug is not the leak itself, but entrusting a 30% annual discount to a single sentence (the prompt) to guard.
50%UnverifiedThe solution given by author Serguey Asael Shinder is to reorder the prompt by content change frequency: content that never changes goes first, content that changes per session goes in the middle, and content that changes on every request goes last.
50%UnverifiedMechanically splitting documents by a fixed token count causes headings and the paragraphs they belong to to be placed in different chunks, stripping the paragraph of key disambiguating information such as which package it belongs to
50%UnverifiedAuthor Serguey Asael Shinder recommends reading the stop reason after every LLM call, just as one would read an HTTP status code.
50%