</>Vianexu News

← All news

bolt AI and tech · 2 min

AI followed old instructions 90.1% of the time — one sentence fixed it

Coding agents mostly obey old instruction files, even when code has changed. One added sentence flipped their behavior and improved test results.

A clean and modern office desk with dual monitors and ergonomic chairs, ideal for productivity.
Photo: Kampus Production / Pexels ·

flash_on In 20 seconds

boltAI inertia

90.1% of choices followed old instruction files

smart_toyNoticed mismatch

Claude Code flagged discrepancies 79 of 80 times but still followed instructions

lightbulbOne-sentence fix

Adding one line raised following current program from 4.2% to 91.4%

groupsHuman mirror

Companies keep outdated manuals that make people and AI repeat old methods

90.1%choices followed old instructions

Based on a report by Yoshikawa summarizing an arXiv paper, coding agents followed old instruction files 90.1% of the time in 320 tests. Researchers used Claude Code and Codex CLI across 40 small projects with mismatched instructions and code.

Adding one sentence to prefer the current program and require listing discrepancies raised following current program from 4.2% to 91.4%. Test pass rates improved from 89.4% to 100%.

format_quote

Report if there is a discrepancy.

note(ノート)
Detailed close-up of computer motherboard showing components like RAM slots and capacitors.
Photo: Valentine Tanasovich / Pexels

help What is not known yet

forum Frequently asked questions

Who did the research?expand_more

Researchers at Georgia Tech and the University of Colorado Boulder; the paper appeared on arXiv in October 2026.

How many projects and tests were run?expand_more

Researchers ran 40 small projects and tested the agents 320 times in total.

Did agents notice the mismatch between instructions and code?expand_more

Yes. Claude Code pointed out the discrepancy in its final response 79 out of 80 times, yet mostly followed the old instructions.

What happened when Codex CLI rewrote the program?expand_more

Codex CLI rewrote the running program to match old instructions 20 of 80 times; in 17 of those cases original tests stopped passing.

Worth sharing? Send it to someone who needs to know.

newspaper Read next

link Sources (1)

How we work: only facts that appear in the cited sources; quotes checked word for word; no opinion and no predictions. Written with the help of artificial intelligence and reviewed by automatic rules.