July 7, 2026
Two Percent of Trouble
98% Isn't Much
When “good enough” means real people get locked out
TLDR: The article argues that “works for 98% of people” can still shut out millions, so it’s not good enough for basic everyday needs. Commenters mostly cheered that take, dragged the same logic in artificial intelligence, questioned the stats, and turned the thread into dark-comedy gold.
A spicy little debate broke out over one brutal idea: 98% is only impressive until you’re in the unlucky 2%. The post argues that for basic things people rely on, “works most of the time” is not a win — it’s a mess. The writer compares it to a babysitter with a 98% survival rate or a restaurant that avoids poisoning customers 98% of the time, which is exactly the kind of nightmare math that made readers sit up straight.
And the comments? Equal parts agreement, side-eye, and chaos. One of the strongest reactions came from people saying this problem goes way beyond websites. One commenter groaned that the same sloppy thinking shows up in artificial intelligence research, where flashy success rates can hide real failures. Another dropped a sneaky bomb: if even JavaScript is only available for 98% of US users, then “everybody has it” suddenly sounds a lot less comforting.
But not everyone swallowed the numbers whole. One skeptic wondered whether bots — automated visitors, not humans — may be distorting the writer’s browser stats, adding a dash of nerdy suspicion to the pile-on. And then the thread went gloriously off the rails: one reader invoked Meat Loaf’s “Two Out of Three Ain’t Bad,” while another resurrected an old joke about milk being sold as “Now with 0.01% less semen” — the perfect cursed punchline to a discussion about whether tiny percentages are reassuring or deeply horrifying.
The mood was clear: people are tired of big numbers being used like a magic trick. If your shiny improvement breaks things for millions, the comments say that isn’t innovation — it’s just exclusion with better marketing.
Key Points
- •The article argues that a 98% success rate can be unacceptable when applied to basic expectations or essential functions.
- •It says that web compatibility figures such as 98% can still translate into very large numbers of excluded users, estimated at about 150 million people globally.
- •The author emphasizes that population-wide support metrics may not reflect a specific website’s real audience.
- •As a concrete example, the article says nested CSS was considered widely supported generally, but only about 70% of browsers visiting one client site supported the needed features over the last year.
- •The article concludes that robust engineering should prioritize graceful degradation and handling edge cases rather than relying only on headline support percentages.