# [AIN#41] More software factories and hello .si

- Publication: Loop Labs Lab Notes (https://looplabs.cc/blog/)
- Author: Rich Steinmetz
- Published: 11 Oct 2026
- Canonical HTML: https://looplabs.cc/blog/ain-41-ai-nuggets/

---

I briefly wrote about this a couple of weeks ago and wanted to go into a bit more detail, now that the first batch of features churned through the factory.

Facing roughly 50+ new API features I needed to add to an Enterprise application, I didn't really want to repeat my daily routine and expect different results from doing the same thing over and over. That same thing basically means invoking skills based on the coding agents' results in different stages of the process:

- Like `/grill-me` for the spec;
- `/implement` when `/grill-me` is done;
- `/hands-on-code-review` when `/implement` is done;
- `/implement` when `/hands-on-code-review` found some heavy items that needed decisions;
- `/hands-on-code-review` when `/implement` is done again;
- Open PR for our in-house bot's review;
- `/implement` its feedback and post with `/publish-review-feedback`;
- Request human review;
- And.So.On. You get the gist.

Instead of going through these motions 50+ times, I wanted to have a factory-type thing move items as autonomously as possible. The first version of it was supposed to move items from idea to human-ready PR, with a couple of loops and guardrail checks in between.

![](https://looplabs.cc/assets/ain-41/software-factory-map.png)

Getting to move those items from A to Z was not as easy a feat as you would think. The first batch needed constant adjustment of the Orchestrator, the guardrails, and a review of things the agents decided to stop at, e.g., because they didn't have enough information, didn't know where to move, or thought there were risky decisions to make.

Though overall, I was way less "in the loop" and more "on the loop", watching things progress and making them unstuck where needed. So, it was more fun, and the quality was surprisingly good. I had at most a few comments per PR regarding code quality and maybe a few behavioral nuances.

One learning nugget I wanted to share about this process this week is that for business-sensitive work, from idea to PR is still too YOLO. I knew this would be a problem, but I didn't know how bad a problem it would be. For business-critical applications, you should get a good spec first if you want a more automated process. Otherwise, you will either end up with an incorrect implementation because requirements aren't finalized, or your pipeline will get stuck waiting on you, requiring better acceptance criteria definition.

For the next batch of API features, I'll be A/B testing our in-house coding agent (which now has increasingly more factory capabilities) vs. the next version of the local software factory. Both with well-prepared specs sourced from `/grill-me` (or similar) sessions on each feature idea.

## softwaremakers.ai => softwaremakers.si

"softwaremakers.ai is so 2026" - A Rails.Builders friend told me last Thursday while laughing at my new `.ai` domain. It's truly amazing that we live in a time when the current year is already too old to be hip to live in. Only a few days later, I understood what he meant: `.ai` was superseded by `.si`. So, given the peer pressure, `softwaremakers.SI` is now mine and currently just a redirect, but let's see where it goes.

Speaking of which (what an elegant transition), if you are interested in hacking for a couple of hours together with other Senior+ Rails Engineers on whatever engineering challenge you are working on, I'm trying a new format on Saturday in two weeks:

- [Rails AI Practice Club](https://softwaremakers.ai/cohorts/ai-makers-club-for-engineers-boolean)

Bring your AI engineering project and build alongside other engineers:

- Present what you're gonna work on.
- Work in silence.
- Present what you've built.

It's gonna be fun, help you make progress on your AI engineering improvements and get feedback from other engineers.

## [AIN#41] AI Nuggets Week#41

### 📺 Expert Knowledge YouTube Nuggets

**10 Levels of Jev For Agentic Engineers** (IndyDevDan) - If you want to understand what Jev is about and see it on kinda practical examples in your AI Engineering workflow, this is the video.

### 🔬 Field Specimens

From the ClickFunnels Engineering blog:

> Link: [Grok Team Bots, a protocol for personal agents, and funnels managed from the API](https://funnelcoders.myclickfunnels.com/blog-dev-with-ai-2026-10-09)

> Link: [Clicky: inside the graph](https://funnelcoders.myclickfunnels.com/blog-clicky-inside-the-graph)

### 🎰 Just for Fun links

(I'm having fun so you don't have to)

> https://x.com/mattpocockuk/status/2108827955837239606

[View this post on X](https://x.com/mattpocockuk/status/2108827955837239606)

Open source maintenance, according to Matt Pocock. You have to click the link and see the video. I think this is really funny, but I also have weird humor. 🦎