Update Dec 31, 2025

§Blawx has been up and running for a couple of weeks, now, and the good news is that it is still up! The bad news is that there are some growing pains to work through, still.

Some Known Issues

• API key creation does not seem to be working properly for paid subscribers

Thankfully, we don't have any paid subscribers, yet, so low pressure on that front.

There are also a lot of errors that are being thrown in the back end that I need to learn better how to monitor and dig into.

What I'm Thinking About - Blawx for Compliance Property Testing

For several years I've been thinking that there may be a use case in software development for using Blawx encodings as a compliance specification and testing tool. In the last couple of weeks I've realized that it might be really easy to demonstrate that capability.

Here's how I imagine it would work:

1. A subject matter expert uses Blawx to generate an encoding that answers a set of questions accurately, the same set of questions that a software project needs to correctly answer.
2. The Blawx user generates an API key and shares questions, gives the API key to their software developers.
3. The software developers put the API key into an open source blawx-mcp server, which uses the Blawx API to ask questions and get answers.
4. The software developers give the Blawx-MCP server as a tool to their coding agents, and instruct the coding agents to use the answers from the Blawx server in order to generate property tests for their software project.
5. The software developer runs the tests, and if their software is generating answers that it should not be, that will show up as a failed test.
6. The software developer can ask their coding agent to send the same facts to the Blawx server, and get an explanation from the Blawx server as to what the output of that test should be, and why.

There are a lot of things that make this a really appealing approach:

• Software development is the primary use case for RaC.
• Compliance is a real challenge with real costs that Blawx could contribute to reducing
• Blawx's anticipated relative weaknesses in computational efficiency don't matter, because it is being used to test the end product, not as a part of the end product.
• "Be a critical part of your organization's software development team" is a decent pitch for subject matter experts to learn to do legal knowledge representation in Blawx.
• Software developers don't need to know the content of the rules in order to generate software that is compliant with them.
• With the increasing sophistication of LLM coding agents, the software developers will often not even need to know what the translation is between their software's data structure and the data structure used in the Blawx encoding.
• When the software developer needs instruction, the LLM coding agents in combination with the Blawx reasoner can provide expert explanations for why a test result is correct or incorrect, on the developer's timeline, day or night.
• It encourages (but does not require) software developers to isolate their legal logic from other parts of their application, making software easier to maintain as rules change.
• The missing pieces (an open source MCP server, and some associated API end points) are relatively painless to create.

So I'm currently working my way toward a demonstration of an MCP server that can be run locally by developers to give their coding agents access to the knowledge in a Blawx encoding, and an illustration of how that knowledge can be used to generate property tests for software.

Improving Automatic Code Gen Is Under Rethink

I really want to get back to improving how the automated code generation features work in the premium version. But my recent experiences with the new generation of coding agent technologies over the past few months have dramatically changed my perspective on what the ideal workflow would be for having the assistance of AI in generating and testing a Blawx encoding.

I haven't worked through all the implications, yet. What is clear to me now is that to maximize the benefit to the user, the user and the coding agent both need to have access to more of the Blawx project simultaneously. The current UI makes too much of the project invisible to the user, making it difficult to let the user see and control what the AI agent is doing. Making the project available to the user through the UI and the agent over the API in a collaborative way also requires changes to the architecture, which is a bigger problem than just changing the UI.

So I need more time to think through how I want to that to work, and what is the minimal viable version that can be slowly improved with time. When I have a better idea of that, I'll come back to it.

Rest assured, using LLMs in the development of RaC encodings is still the best available solution to the most important problem in the space. So I'm not going to forget!

Coming Soon: Rules as Code Europe 2026

I'm grateful to have received an invitation to participate in the second edition of the Rules as Code Europe conference being held in The Hague, in March of 2026. I may be presenting as part of the "Rules as Code - Community" track of presentations (keep your eyes on the list of presentations), but either way I'm very much looking forward to learning from my European colleagues in the space, who have been doing incredible things, and whose experience is not as widely shared as it ought to be.

If you're going to be there and would like to meet for coffee, or would like to learn more about Blawx or teach me about what you're doing, let me know.