NUKE in trouble? How about Dagger

I am heavily invested in NUKE build, however its author has been having a bad time with nasty community engagement.

I hope he will consider taking it commercial, as the product brings a lot of value.

That did, however, prompt me to consider alternatives. The core issue is the well known difficulty of developing and testing GitHub Action workflows locally, which leads to a long stream of error correcting commits when you work with CI.

NUKE mostly solves this, by capturing the build script in strongly-typed, imperative .NET/C# (no more YAML debugging), and allows you to easily run the pipeline locally. Dagger offers an alternative solution for the same problem; here an underlying execution engine is used to orchestrate functions written in Go/Typescript (or untyped alternatives) that chain together.

It allows you to take input (e.g. code in repo) to output (e.g. versioned deployment artifact, such as a container image) and can potentially also be used for code-based CD of those artifacts.

I made a small spike of building a basic .NET solution (with heavy aid from Copilot)

In principle, this sounds great, but lack of .NET support and lack of Microsoft engagement in the Daggerverse for common tasks curbs my enthusiasm a bit. Staying with NUKE for now, but keeping an eye on Dagger.

CI/CD with Dagger and Azure

Fizzy - a fresh take on Kanban

The BaseCamp folks made a Kanban board / task tracker, free for up to 1000 cards. Also the code is open source (like they did with Campfire), which I would think is really valueable for upcoming Ruby programmers. At least I always wished for access to commercially-viable codebases to learn from early in my career, but that was unfortunately not commonly available for .NET in the early 2000s…

Fizzy

How to write a great agents.md

Nice, actionable guidance.

How to write a great agents.md: Lessons from over 2,500 repositories

The Engineers Who Can’t Use AI Agents Don’t Have a Tools Problem

Postulates that to use AI (coding) agents efficiently, you have to be able to

  • understand what you are building
  • externalize that understanding in writing

That is defintely true, but not the whole picture (as in: more is needed, context, experience etc.). It made me think of Kent Becks 90% of My Skills Are Now Worth $0…but the other 10% are worth 1000x

EDIT: See also Adam Tornhills take I tend to delete 80% of the generated code. The remaining 20% I refactor

The Engineers Who Can’t Use AI Agents Don’t Have a Tools Problem

Azure CDK progress

Previously-mentioned Azure CDK seem to be maturing nicely. They are not pushing it very hard, so I guess the intended usage is more via the higher-level deployment story of Aspire?

Azure CDK