News for hackers. Stuff that matters.
Re:Turbo Haskell by jerf on Friday October 02, 2026 @09:39AM (#49933395)
I briefly worked with him personally. He is 1. a perfectionist and 2. equipped with the intellectual and programming firepower to not get stuck in the resulting paralysis that a lot of people get stuck in but blast through tasks that should be impossible to do at all, let alone with a high degree of quality, successfully, through sheer intellectual firepower. The end result is that he is well known for already having not just one incredible Haskell library, but a number of them.

I can promise you he will not produce and release "slop". He would be an order of magnitude more offended at the idea than you are, I'd imagine.

Re:Turbo Haskell by ddellacosta on Friday October 02, 2026 @11:17AM (#49934473)
> not just one incredible Haskell library, but a number of them

...and this still understates his impact on Haskell, he helped shape how a lot of programs are written in Haskell through some of the stuff he worked on (see e.g. https://stackoverflow.com/a/2769840 as evidence of his influence on core functionality).

Re:Turbo Haskell by therepanic on Friday October 02, 2026 @11:07AM (#49934335)
in that case it is truly amazing, because I have never seen anyone even with current ai capable of being so productive and truly understanding what they are putting out.
Re:Turbo Haskell by solomonb on Friday October 02, 2026 @11:56AM (#49934980)
Ed Kmett is one of the smartest people I have ever met in my life. He is quite literally built different.

He completed undergrad and graduate programs in math, computer science, and a special program in ai all simultaneously in 3 years. I believe he was finishing programs before his scholarships could kick in.

Re:Turbo Haskell by edwardkmett on Friday October 02, 2026 @03:12PM (#49937329)
Long time no see!

In the original commenter's defense, I _did_ let the agents I have working on this get a more than a bit test harness crazy back when I was able to use GHC itself as a behavioral oracle. That worked fine through the initial build out, but then it rushed ahead with a broken CI due to misalignment in priorities, and then kept piling fixtures on top of fixtures and didn't properly track their cross-dependencies. My work on this in the last 48 hours has mostly been about getting that part under control and stable across my target platforms so I have better bedrock to build atop.