Remember when Netflix was good? Or Uber? Or Facebook? Or MoviePass? Or Google? They have all since been enshittified, the process whereby the first version of a service is cheap and good funded by investors to lure in users, then the service gets overrun by ads to lure in advertisers,
Development often deals with backlogs of tasks and the problem of how to schedule them to achieve the desired goal. To do that, we need to know in which order to execute the tasks, and to translate that into a plan we need to know how long they take. Here I introduce a missing piece to achieve that.
The Pareto principle is being upended by AI, so that you no longer spend 20% of the time to make 80% of the functionality, as well is it changing the worth of Proof-of-Concept implementations.
Building enterprise software is expensive. Over the years, many approaches have attempted to alleviate this with varying success, and the latest is to use AI to aid in software development. Here, I go over a project we did with an intern to evaluate the benefits that.
I wanted an image gallery, and ended up picking Vernissage. The documentation comes with a description of how to host is using Docker Compose, but I wanted to use Kubernetes instead, so here's how I did that.
I am moving away from WordPress, which I used to use to share my photos, and now I need a replacement. This post summarizes my wishes and some of the applications that nearly satisfy them. TL;DR: I picked Vernissage.
One of the hardest things to do, psychologically, is admitting you were wrong. It hurts core beliefs and our sense of self. Deep canvasing might be the only way to counter that.
A simple guide to when using AI to solve a problem is appropriate and a comparison to Eisenhower's matrix for task categorization.
Here, I'm surveying personal blogging software, and how I set up my software of choice, Ghost.
All you need to know about Britney, computer science, and psychology
The internet is all up in arms about the Honey browser extension, calling it a scam. People claiming that are extremely naïve.
Algorithms allow computers to perform tasks. To many, AI and algorithms seem like magic and get lumped together. Simultaneously, algorithms get a bad name due to algorithmic social media feeds. I find it important nurture the ancient art of algorithm development, and here I explain why.
In this post I report on some further progress on integrating Cerbos authorization with Spring Security and Spring Data JPA. I made a library we're currently using, and here I outline how the code can be used and how it works.
I've previously talked about AI's uncanny valley problem, that a little AI is worse than no AI, and an experiment showing that people accept any reason for performing small favors is a great example of why that is.
The uncanny valley is an effect describing how we like a picture more if it resembles a person, but dropping significantly in the transition from great cartoon to badly rendered real person. Here, I posit the same is true about the usefulness of AI.
WordPress, quite possible to most popular CMS, is no longer a viable solution. It is no longer free software and the ecosystem is likely to fragment to the detriment of users.
Dr Doom is humorously lauded as having predicted 10 of the past 1 crises, and prediction markets aim to make better predictions than that. They are founded in genuine research, but have flaws. Here, I present 3 reasons one of the biggest prediction markets, Polymarket, is not really that useful.
Authentication is the process of checking if a user is really who they claim to be. Authorisation on the other hand is determining whether an already authenticated user should have access to a given resource. For a project, I've been playing with Cerbos, and here I'll summarise my experiences.
AI is a fancy word for statistics. Two methods are in common usage today: image generators and language models. They rely on different methods, and I'd argue the image generators are smarter than the language models. Image generators basically work by removing random noise from an image. They
Computing had peaked. Nothing was happening. Operating systems peaked 30-40 years ago. First, there was nothing, then there was command lines, then was GUIs, and then GUIs with working multi-tasking. Operating systems had peaked and nothing of value was ever subsequently added. Except, every year or two, the
As a kid, in the 80s/90s, "we" had a SodaStream machine. It produced barely passable off-brand colorful fizzy drinks, but was more expensive to use than buying real pop. About a month ago, my stash of soda was empty. I normally buy in bulk and only
Record companies used to front the costs of record production, financing the unsuccessful artists with the proceeds of the successful ones. They provided a service justifying their cost. Today's artists come with built-in fan bases and albums worth of hits. No record company took a risk on
