More from Words and Buttons Online
An image interpolation that gives us a continuous and smooth image, where every interpolated value only depends on the four neighboring pixel values. The image becomes smooth, but sharp features remain sharp.
All the “C++ killers”, even these which I wholeheartedly love and respect like Rust, Julia, and D, help you write more features with fewer bugs, but they don't much help when you need to squeeze the very last FLOPS from the hardware you rent. As such, they don’t have a competitive advantage over C++. Or, for that matter, even over each other. Most of them, for instance, Rust, Julia, and Clang even share the same backend. You can’t win a car race if you all share the same car. So, which technologies do hold a competitive advantage over C++ or, speaking generally, all the traditional ahead-of-time compilers?
Rational interpolation is a step forward from polynomial interpolation towards rational splines. With rational interpolation, you can build functions that run through a set of points and also have vertical asymptotes whenever you want. With this capability, you can now model functions like logarithms better.
This page uses polynomial modeling to show why software engineering tasks are often impossible to estimate.
Type inference works like logic deduction so any program in a statically typed language is two programs. The first one is the thing you sell, and the second – is a model that undergoes some sort of verification every time you run a compiler. This second program, although, often being written unknowingly, is not a free bonus, it’s something you have to pay for in several ways.
More in programming
I know I said we'd be back to normal newsletters this week and in fact had 80% of one already written. Then I unearthed something that was better left buried. Blog post here, Patreon notes here (Mostly an explanation of how I found this horror in the first place). Next week I'll send what was supposed to be this week's piece. (PS: April Cools in three weeks!)
Ask any B2C SaaS founder what metric they’d like to improve and most will say reducing churn. However, proactively reducing churn is a difficult task. I’ll outline the approach we’ve taken at Jenni AI to go from ~17% to 9% churn over the past year. We are still a work in progress but hopefully you’ll […] The post Notes on Improving Churn appeared first on Marc Astbury.
Meditation is easy when you know what to do: absolutely nothing! It's hard at first, like trying to look at the back of your own head, but there's a knack to it.
Discover why 'if not mylist' is twice as fast as 'len(mylist) == 0' by examining CPython's VM instructions and object memory access patterns.
In a fit of frustration, I wrote the first version of Kamal in six weeks at the start of 2023. Our plan to get out of the cloud was getting bogged down in enterprisey pricing and Kubernetes complexity. And I refused to accept that running our own hardware had to be that expensive or that convoluted. So I got busy building a cheap and simple alternative. Now, just two years later, Kamal is deploying every single application in our entire heritage fleet, and everything in active development. Finalizing a perfectly uniform mode of deployment for every web app we've built over the past two decades and still maintain. See, we have this obsession at 37signals: That the modern build-boost-discard cycle of internet applications is a scourge. That users ought to be able to trust that when they adopt a system like Basecamp or HEY, they don't have to fear eviction from the next executive re-org. We call this obsession Until The End Of The Internet. That obsession isn't free, but it's worth it. It means we're still operating the very first version of Basecamp for thousands of paying customers. That's the OG code base from 2003! Which hasn't seen any updates since 2010, beyond security patches, bug fixes, and performance improvements. But we're still operating it, and, along with every other app in our heritage collection, deploying it with Kamal. That just makes me smile, knowing that we have customers who adopted Basecamp in 2004, and are still able to use the same system some twenty years later. In the meantime, we've relaunched and dramatically improved Basecamp many times since. But for customers happy with what they have, there's no forced migration to the latest version. I very much had all of this in mind when designing Kamal. That's one of the reasons I really love Docker. It allows you to encapsulate an entire system, with all of its dependencies, and run it until the end of time. Kind of how modern gaming emulators can run the original ROM of Pac-Man or Pong to perfection and eternity. Kamal seeks to be but a simple wrapper and workflow around this wondrous simplicity. Complexity is but a bridge — and a fragile one at that. To build something durable, you have to make it simple.