Build without limits - SDLC Briefing, #4

Hi! I am Ricardo Sueiras, leading SDLC transformation at Indicium AI. That means I get to work with customers and help them understand how to use agentic AI to build software. New tools, practices and opinions turn up every week, and Build without Limits cuts through the noise. In each edition we put together a short, curated briefing on what's worth reading, watching and worth trying. You can skim it in five minutes and dive deeper wherever something catches your eye.

Top of Mind

“Tell me how you measure me and I’ll tell you how I will behave.”

― Eliyahu M. Goldratt, Critical Chain

Hat tip to my former colleague Ray Krueger for that gem. Measurement is top if mind, and keeps coming up in conversations. James Shore has put together some of the best thinking in this space, so check out his series Assessing AI Impact: Notes for Your CFO

Must reads this week

These were my favorite posts this week.

The State of Context Engineering is a blog post come speaker notes from Drew Breunig's session at AgentCon. It was an easy read and I learned a few new things. I have given similar talks in the past (check out my talk at Prompt Engineering Conference, Tips and Tricks for better prompting)

Hello concurrent-cd is the new home that Patrick Debois has put together to share research, experiments, and findings on the confluence of people and agents.

What's next for DORA I was pretty shocked to find out that there are going to be no more independent studies on software engineering practices and productivity. If you care about developer experience, this is going to be a miss. However, the methodology and questions are there for anyone to see - who will step in and carry on the work?

What I Learned Building an AI Software Factory is a nice walk through of one persons journey and learnings in creating a software factory. Well worth a read.

Watch this

I am still messing around with Claude Code mods, and love seeing what folks are up to. AICodeKing shared a few that look genuinely interesting. I know a few folk who always need to see the files being created during a Claude Code session, so like to run it within an editor. One of the mods shown here might just present a cool solution for those folk.

Tools to try out

Third-party tools links are provided as-is. They belong to their owners, follow their own licenses, and are used at your own risk.

command centre is an amazing tool from my colleague Lee James. It provides an easy way for you to quickly extract insights across your Claude Code deployment. From finding out and managing token spend, discovering use cases, to tracking compliance and usage habbits. Give it a try.

paparats-mcp is a relly neat tool that gives Claude Code or any MCP client semantic code search across many repositories: by meaning, with who calls what, and what a change will touch. It runs on your machine or sandboxed in a container. Why might you want to try this out? This approach will help you consume way less tokens, and be much more responsive. The README of the project has some data on that so give it a go.

shunt is an interesting tool from the folks at Spotify that helps drastically reduce token used by implementing a technique that tries to intercept operations that consume large reads. I heard about it after reading the blog post, Portal by Spotify cut my Claude Code token usage by 90%

From the field

One of the best parts of the Forward Deployed Engineering job is that you get to see up close how engineers use agentic engineering tools like Claude Code. There is always stuff coming up that you can learn from, and this week, this issue surfaced up. TLDR; in your Claude Code sessions, a simple feature that helps you signpost where you currently are, can quickly consume tokens over time.

What is interesting is that this issue has two dimensions. The first, the recap message you see during your sessions, is one that you can easily disable via /config (turning recap from the default of true, to false to disable it). All good, except that this doesn't fix the issue for subagents. The issue comments go on to show how you can work around this (effectively disabling the spawning of agents). This gist provides a hacky solution which I would not implement myself, but am sharing as its interesting to dive deeper into this issue.

If you enjoyed this, then remember to subscribe to the IndiciumAI blog so that you can keep up to date with future editions of this newsletter, as well as other great content that technologists at IndiciumAI will publish.