Another CI video is out. How @InputOutputHK does CI using @EarthlyTech, including auto-discovery, Multiplexing, post build processing and a bunch more. youtube.com/watch?v=vgELhSmT…

Programming With Expressions in Rust, Ruby, Kotlin and Scala

Problem: I needed group documents by relatedness on the @EarthlyTech blog. Doing so led me on a journey to learn about word vectors, the OpenAI text embeddings API, and cosine similarity. It turns out it's all less complicated than it seemed. At first, I tried it by hand, making a 3-dimensional word vector based on the dimensions (pets, footwear, construction materials). It's a silly grouping, but it gave me a good intuition for what's happening. You can then calculate the angles between these vectors and see how relatedness is arrived at with word vectors. It's just some geometry. After that, it's easy to extend to using a word2vect implementation and ultimately to OpenAI's text embedding API with whole document vectors in 1536 dimensions instead of 3. The math remained the same, but by the end, I had a relatedness measure based on a whole document's semantic meaning, across the 400+ articles on the Earthly Blog. Then, I just added it to the blog sidebar. Check out my write-up if you'd like to learn more or see it in action. What you can do these days with a few lines of Python blows my mind. earthly.dev/blog/cosine_simi…

Video is up here: And in it Dr. Alexander Mikhalev shows how he builds firecracker VMs with Earthly. It's the first in a Earthly Show and Tell series I'm making. youtube.com/watch?v=AyUh6-N2…

Hell yeah! I'm already using XL for one of my builds. It runs in less minutes now, so the cost change isn't bad.

I wanted to get this video in under 5 minutes but 8 minutes for the basics of JQ is still pretty good I think. youtube.com/watch?v=plFObIaJ…

AWK is a great tool but I realized that all my AWK usage is the simplest of stuff, so I've now made a 5 minute AWK tutorial. This is what you need to know: docker ps -a | awk '/rust-example/{ print $1 }' youtube.com/watch?v=15DvGiWV… Take some command line command, and grab a specific value from a specific row and column. I explain and go through a bunch of examples in the video if you have 5 minutes to spare. It will save you more than 5 minutes if you aren't using AWK now.

𝐁𝐮𝐢𝐥𝐝 𝐚 𝐂𝐨𝐝𝐞 𝐓𝐫𝐚𝐧𝐬𝐥𝐚𝐭𝐨𝐫 𝐔𝐬𝐢𝐧𝐠 𝐋𝐋𝐌𝐬 - 𝐈 𝐝𝐢𝐝 𝐢𝐭. A bit ago, the whole @EarthlyTech team was gathered in a board room at the Hilton Garden Inn in Minnesota. After a fun day of boating and then curling, we were getting down to planning. @VladAIonescu was saying that one of the biggest adoption hurdles with Earthly is the time investment customers need to spend to change over their builds. Ideas were kicked around for easing adoption. The most ambitious was a GitHub action (GHA) to Earthfile converter. Well, I built it using GPT-4, and it's out. Also, I wrote up tips for others building on top of LLMs. I came out of the experience pretty bullish on the potential of large language models to do code translation work. Blog Post: earthly.dev/blog/build-trans…

As a software developer, I've seen how incentives can drive the development of enterprise software and dev tools in ways that aren't always user-friendly.