2 karma · joined June 12, 2026
Will the multi-file diff just apply differences separately across files, or will it detect moved code too?
These tools require a certain approach to do pair-programming, but I’m OK with not actually sitting at the same computer alongside someone, or sharing some video conference. More a case of being a navigator with occasional driving where I do the coding locally and submit it back to the LLM.
If you can pay, you can use the locally-running tools to get a more interactive session going.
The usual caveats about using AI apply, of course, but that’s the fun part of working with others, be they human or not. You never know quite how it will go!
I am oversimplifying, and treating the black hole as already collapsed inside the event horizon from the point of view of somebody well away from it; that’s not accurate. Perhaps thinking of that mass as being ‘just about to collapse into the event horizon’ is the best way to think about it, because that’s what that somebody would perceived from the outside who is not in free-fall into it.
Need a proper physicist to follow up on this, my training stopped at special relativity and Newtonian gravity…
(PS: studying tensor calculus and general relativity is on my bucket list - thanks for reminding me to get a move on with it!)
All of these have a couple of thousand commits that I'd done myself over the years, so I don't have that sense of "what has been built and what's my contribution?".
What I find very fulfilling is to finally get to have drudge-work tickets that I'd put off for ages done. It's nice to be able to pair-program with the LLM as well, I rather enjoy the back-and-forth of ideas - I'm sometimes pleasantly surprised by how insightful an LLM can be. Pair-programming / reviewing keeps that sense of involvement going, in fact it accentuates it somewhat, because I have to keep on top of what the LLM is doing.
The novelty of conversing with a synthetic quasi-intelligence hasn't worn off yet, either.
It's not always plain-sailing, though. Sometimes they go off the rails. :-)
I’ll keep an eye on how this will integrate with PBT, that’s an interest of mine.
1. Kinetic Merge - it's a CLI tool for merging in Git repositories. What it gives over and above the core Git merge is the ability to merge through the motion of code from one location to another. Yes, Git can track file-renames and merge through them, but it doesn't follow code moving around inside a file, or a file being split into two or more pieces, or a new file accumulating code that used to live in several files. Essentially, the sort of thing you get with a lot of refactorings or automated code tidy-ups.
I've been at this since 2023; it was a long-standing itch I had to scratch since back in the early 2010s, finally got some time to work on it. Now in its 57th release, GitHub tells me!
It's written in Scala and packaged up as a standalone executable - needs a JVM >= 17 to run it on Linux / Windows / OSX.
(https://github.com/sageserpent-open/kineticMerge)
2. Americium - this is a mixed Java / Scala framework for folk who want to write property-based or generally parameterised tests, but: a) want the failing tests to shrink automatically without writing custom shrinkage logic and b) don't want to have the test structure and expectations DSL dictated to them by the framework.
Its job is to serve up test data into your flavour of Java or Scala test, look for an exception from the test and get shrinking. The rest is up to you.
There is a separately shipped artifact that plugs into JUnit5 / JUnit6 for a tighter integration, including test replay via IntelliJ or Visual Studio or whatnot, but that's an optional extra.
I've worked on this since 2020, now hitting its 52nd release. I use it for my other projects, and I know of a couple of corporate teams that use it because I evangelised it to them, but ironically, I'm only just getting round to using it for the problem that motivated me to write it in the first place.