LLMs Will Replace 8-Track Duplication Engineers
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
8-track cartridge duplication required engineers to solve a 4-way balanced number partitioning problem — an NP-hard optimization — by hand with pencil and paper. Using 6,463 deduplicated solves scraped from Discogs and MusicBrainz, the author shows that the median 1970s duplication engineer landed just 5 seconds from the provable optimum, beating both the greedy LPT algorithm and the Karmarkar–Karp differencing method (which wasn't even published until 1982). 27.5% of human solutions were provably optimal. The dataset is then used as a novel LLM benchmark: top reasoning models like Fable 5 (max effort) achieved perfect scores, while models without reasoning performed far worse than the human experts. The piece also explores house styles across duplication shops — some preferred splitting tracks, others repeated songs — and calculates the real economic value of good optimization (saving tens of thousands of dollars per album run).
