I’m back after the holiday weekend and some shenanigans at the San Diego Padres baseball game yesterday. Today, in between real work and building a Chrome plugin to test out an idea, I also read a number of fascinating items.
[blog] Coding Agents vs. Workflows vs. Orchestration vs. Platforms: How to Architect the AI Dev Stack. You might not need them all in every situation, but each of these components in the stack has a useful role.
[article] Inside a Software Factory. I’m not sure if the whole “software factory” as currently described will survive. We’ll see. This post does a fine job outlining where humans play a role in these factories, along with design considerations.
[article] What is the future of Software Engineering when nobody needs to write code? If you want to read one article that sums up much of the current discussion about AI and software engineering, read this one.
[blog] I Replaced 15 Go Packages With Nothing But the Standard Library — Here’s What It Actually Took. It’s hard to do this with most languages. External packages have a place, of course, but when you don’t HAVE to use them to get something done, you’re at an advantage.
[blog] What happens when AI burns down the backlog? Will the backlog run out? That was a silly question to ask even 6 months ago. But now? Are you identifying useful things to work on at a fast enough rate? Maybe not.
[blog] AlphaGenome Atlas: A predictive map of every possible DNA letter change in the human genome. There aren’t many frontier labs consistently doing work like this. Special stuff!
[article] The quality paradox of AI-generated code. Why are code maintainability and change confidence moving in opposite directions all of a sudden? Interesting analysis here.
[article] Google Cloud races to catch up in the AI deployment wars with Accenture deal. If you’re looking to significantly change how and what you build, there’s a lot of value in getting outside help. More here.
[article] “Twenty years of brand building simply froze in time”: How coding agents select their tools of choice. This article is about this research. My team has been doing similar research. It’s fascinating to see what LLMs suggest, and what they don’t. A model’s bias for jumping to web search (over relying on what’s frozen into the weights) has a big impact too.
[blog] Not All LLM Workloads Are Equal: Benchmarking TPU Performance on Classification vs. Generation. Different categories of actions scale in different ways. This is useful info as you set up AI infrastructure.
[blog] How to write a definition of done for an AI feature. Good point that we shouldn’t mistake evals for a requirements spec. PMs need to keep understanding user need AND definition of what “done” looks like.
[article] What it took to triple our software engineering output in 18 months. Here’s a solid playbook. Do these things, and you’ll be in good shape.
Want to get this update sent to you every day? Subscribe to my RSS feed or subscribe via email below:
Richard Seroter's Architecture Musings 
Leave a comment