I'd be interested to know when that Opus 4.6 baseline is from given their recent recognition of performance issues. Do you have a paper posted on this review?
There's another cheeky example of this where you select a pseudo-random seed that makes your result significant. I have a personal seed, I use it in every piece of research that uses random number generation. It keeps me honest!
One of the big problems I have with cursor is that it ignores the rules frequently. For example, working in the front-end it will sometimes totally ignore all the components that I have explicitly told it to use. Would this... fix that?
This article is great but it doesn't even talk about the assistant. The assistant is an LLM powered version of Kagi. It's been my daily driver for months and the productivity boost is incredible. I rarely ever use kagi's actual search functionality anymore unless I'm looking for something very niche. I love that you can also swap models. Happy customer here!
The Assistant product is incredible. It's been my daily driver (+Cursor), since September. I don't even bother with normal search results unless I really can't find something (like a specific GitHub issue that hasn't been cached yet)
I love that you can pin, raise or lower results from specific websites. They also set it up so you can see popular raise/lower and pins. It's wonderful. Trust me too, it's worth it. ~850 searches a month here.
Linux desktop has come a long way. KDE, Gnome and newer shells that layer on top like Pop Shell are really a joy to use. I very rarely use Windows anymore except for gaming where there are still some compatibility issues. Steam has done great work here too. Battery life on a laptop is still an area where you might prefer to go the WSL route.
The product is great. I love that I can pin, raise or lower the importance of specific websites. I average about 800 searches a month, and I have only had to go to another search engine maybe once or twice a week. This product feels like what the web should be.
Fractional scaling works well, but only when all your monitors are the same size. If you're a laptop user connecting to a differently sized screen it can be a challenge.
It's been a solid daily driver for me at work and home for the last 4 years. I would also be interested in knowing what you feel is completely broken. The window tiling system they came up with is really quite good.
Love this comment because it translates so well to any mentally taxing endeavor. Writing a long-running program? Set a few assertions up front so that it fails immediately before wasting your time. Think about those "failure" condition assertions up front and save your time on practically any experiment. Even chess players do this by surrendering early when they know a game has reached a conclusion.
Whatever you do, focus on a deep specialization. Find something you really love then become a profound expert in it. Education is absolutely critical in a world where you are either telling machines what to do or being told what to do by a machine.
S3, S4, R6, and reference classes. To be fair they are situational and not one size fits all. The stricter ones are mainly used in biostats where significant metadata makes more sense in OO. S3 is nice and easy, primarily just a list with dispatches. Everything else is less so.
I can't think of a PDF but there are images like this. "Lenna". There are also really famous datasets that are used for examples in similar ways (iris).