November 2, 2025
Bug math, paywalls, and spicy memes
I ****Ing Hate Science
Coders roast fake studies and paywalls, yet admit science is the least-worst
TLDR: A developer dug into “bug cost” research and found the famous IBM chart is fake, old studies dominate, and paywalls make real evidence hard to find. Commenters split between “science is flawed but best we’ve got” and “this is pseudoscience,” sparking debate over how to trust software claims.
Today’s software drama: a developer dove into whether catching bugs early actually saves money and discovered the internet’s favorite chart—the “IBM Systems Sciences Institute” bug-cost graph—basically doesn’t exist. The piece slams dusty 1970s studies, broken citations, and journal paywalls, then cheekily praises Sci-Hub (“Alexandra Elbakyan did more for society than the entire FSF”). The community grabbed popcorn.
Top mood? Disillusioned but pragmatic. ironmagma quips that science “sucks” yet beats every alternative. artyom calls classroom “bug-cost” experiments pseudoscience and a funding grab. rramadass reminds everyone the scientific method (how we test ideas) is itself empirical, but you still need brains to interpret results. whatshisface throws shade at fields that don’t self-correct, warning bad studies can stick like gum under a desk. Jtsummers drops receipts with past flamewars: thread 1, thread 2.
Jokes and memes flew: people riffed on “napalm trains,” the author’s line about defects vs. bugs (“one’s a bug, the other’s a defect”), and the eternal hunt for papers dubbed “scrobbling.” For non-insiders: Empirical Software Engineering means studying coding with data; Formal Methods are math-heavy ways to prove programs correct. Verdict from the crowd? The research world is messy, but read the primary sources and distrust neat charts. Stay skeptical.
Key Points
- •The article challenges a widely shared bug-cost chart attributed to the IBM Systems Sciences Institute, citing research that the study does not exist.
- •It notes that much of the bug-cost discourse relies on 1970s-era sources (e.g., Barry Boehm, COCOMO) and that secondary sources often distort primary findings.
- •Access to primary research is costly due to paywalls, with per-paper costs around $40; Sci-Hub is mentioned as a workaround.
- •Finding relevant papers is harder than paying for them, due to noisy search, fragmented academic databases, and inconsistent terminology (e.g., “bug” vs. “defect”).
- •The article outlines a practical literature search method: start with seed terms in suitable journals, skim abstracts, follow citations and references, discover better terms, and build a network of key review papers over time.