98% isn't much

panny

I couldn't agree more. BTW, 98% of US users have JavaScript enabled in 2025. https://www.mediawiki.org/wiki/No-JavaScript_notes

zipy124

This concept is missed so much in AI research and is quite frustrating.

msephton

Reminds me of the Meat Loaf song “Two Out of Three Ain’t Bad” which was released in Japan as 66%の誘惑 “66% is Good Enough” etc https://www.discogs.com/release/8303076

arealaccount

I wonder how much traffic from bots is skewing OPs nested CSS calculations

atan2

That reminds me of an old comic where a guy picks a milk carton from the grocery store shelf and reads in the box: "Now with 0.01% less semen." and he does not know if he's happy or sad about it.

eknkc

I agree the general premise but do not agree when it comes to browser support. I feel like we should be building for the 98% or even 95% and force the remaining to upgrade their browsers. I've built for the IE6 - IE11 era for a painful and long time. I do not give a shit if you want to use a 3 year old browser at this point. Go look at a blank screen.

high_na_euv

Depends on the context

theragra

Pragmatically, often users without new browsers and OSses are not the best clients. In ideal world, sure, I want to support everyone. In a world with limited resources, I would better spend my time elsewhere.

vb-8448

If it's uptime it's definitively not much!

theandrewbailey

1% failure rate of a hundred might be acceptable. 1% failure rate of a million is not. Isn't that a named law?

amelius

This applies to AI too. Your classifier might be 98% accurate and it may sound like a lot. But if it sits inside a car, making thousands of decisions during every trip then you may be in deep trouble.

WaitWaitWha

The author seems to equivocate by comparing completely different domains. Whether 98% is acceptable, it depends on the cost of failure , not the percentage itself.

217

while true, the people who will read this and then think twice about implementing and applying things are exactly the people who already doing too much thinking

cantalopes

I am not exactly sure what is the article trying to point out

nilirl

> Truly robust engineering isn’t about what works for most; it’s about gracefully handling the edge cases. How do you justify this when you factor in cost and time?

BigRedEye

I think this single fact is a major source of enshittification in large software products, especially in the era of ML/AI. If your quality is 99%, it sounds like "you have solved your task", but in reality there is a long tail that over time affects nearly every customer. I've seen this so many times. 99% of search results are good (so within 100 queries you'll hit at least one bad result with p≈0.63), 99% of dashboard panes load normally (so a dashboard with 20 panes is broken in nearly 1 in 5 loads), and so on. If your LLM gets 99% of tool calls right, nearly every session will contain a malformed tool call. Probabilities are hard for humans, probably.

abap_rocky

> Can you imagine a venue refusing entry to former clients 2% of the time just because they’ve “improved their experience”? This reminds me almost precisely of the dynamics of pro sports in the US and how fans are getting priced out of attending games or even watching teams on TV as organizations shift to bespoke streaming platforms.

qarl2

It's just mathematical expectation. Don't look at the simple probability - look at probability * value.

wccrawford

Alternatively, 98% is plenty . If your business plan requires you to capitalize on more than 98% of the market, it's already a failure. It'll never happen. As always, it's an "it depends" situation. If your userbase is largely luddites, then maybe you need to support 10+ year old browsers that can't be updated. Otherwise, you can probably just worry about people who are using computers new enough to actually update their browser once a year or better. The tradeoff is code complexity and engineering time, vs having a larger market. And that's going to be an individual situation for every company.

mewpmewp2

There's likely always a line somewhere where effort becomes way out of proportion compared to getting that last mile effort. Arguably, if you only have a website, that won't work for anyone without access to the Internet. So then you should have a physical presence in each of those people's location, and arguably you shouldn't provide any improvements that give me more than physical presence does, so you should not have the website in the first place, since people without the Internet can't use it or you have to keep your website without any improvements over the physical office. If you only have a website, arguably 2+ billion people currently wouldn't be able to access it. And it seems odd to bring 150 mil people as an example, when the baseline should be at least 2+ billion with website only.