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
How to approach reviewing code has been top of mind this week, and came up in conversations during the Applied AI Symposium that Indicium AI held in London this week. My timeline also had some interesting posts on this topic. First up was Spotify sharing how they tackle this in their post AI Changed How Spotify Builds. What We Learned (and Fixed) About Quality at Higher Velocity. Anthropic also dropped perhaps one of the more interested pieces on this topic, looking at how they approach doing review at scale in How the Claude Code team does code review, and one of the authors also pencilled another in this topic that is also worth reading, Agentic Verification. Anthropic also published a recipe in their Claude Cookbook that walks you through creating a scheduled reviewer for your repo, and it is well worth implementing this recipe. Go check it out (and other really cool recipes the folks are sharing), Building a scheduled repository reviewer.
Must reads this week
- Escalate hard decisions with the advisor tool is an experimental feature within Claude Code that I will admit I did not know about it until I heard about it during a meetup I attended this week (remember folks, meetups are super valuable for learning these sorts of things!). This feature allows you to use cheaper models, and then when needed, advisor will kick in and consult stronger models when needed (for example, if its bumping into repeating errors, or before, task is completed)
- Redesigning source control at Anthropic is an interesting post that looks at some of the engineering choices that Anthropic are making when it comes to how to make source control work at agentic engineering speed. Scion is their bet, and its unlikely that many will find themselves in need of implementing their own, but we are seeing folks already working on improving the scaling capabilities of git, so who knows.
Watch this
Zach Lloyd, founder and CEO of Warp did a talk at the AI Engineer conference in New York last week, talking about the software factory they built to help them manage the open sourcing of their Warp terminal (agent control plane). There is a lot of good stuff in this, including the detailed workflow and sample repo you can try for yourself. At Indicium AI we also help our customers implement software factories, and the architecture in this session had a lot of similarities to the approach we have taken, so lots of nodding from me.
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.
claude-code-templates - I was drawn to this repo because Daniel San put together a really cool Claude Code mod that lets you inspect whats inside your context window at every turn, but then got lost in all the other cool stuff Daniel has been putting together. Defo a repo to spend some time exploring.
diff-tour - this is a really neat idea, that follows a thread of similar other tools in this space. How can you make it easier to reviews to walk through your diffs during PR reviews. diff-tour attempts to solve this by creating guided walk throughs the way a senior engineer might do. Works with a number of coding agents, including Claude Code.
archify - is a tool that I am looking forward to trying out, especially in some of the code modernisation and onboarding use cases I am helping customers with. The tool uses agents to create beautiful interactive experiences that help you navigate and understand how a project or codebase works. It works with multiple coding harnesses, including Claude Code, and is one of the better tools in this space.
mixture-of-models-claude-mod - is a neat tool that allows you to provide dynamic routing of requests to your Anthropic models or cheaper, locally hosted ones. Similar to how Kiro's Auto mode works, but this one allows you to do this from within your Claude sessions. Requires a bit of fiddling, but if you like that sort of thing, give it a go.
From the field
This week, everyone has been talking about reverse engineering with agents. Whether that is using tools like rea, or one-shotting using the latest generation models, my socials are full of folks beginning to share what they are doing with it (including this cool one that looks at how gamers are using it). I am interested as I can see some practical use cases for this in code modernisation use cases where perhaps source code is not available or the confidence is low. Have you checked it out yet?
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.
