April 18, 2026
I'm very excited about v2.1.0 of §Blawx, and the corresponding v0.2.0b of the Blawx-MCP server, live now, because this is the first time §Blawx users have had access to real frontier strength agentic AI assistance in generating §Blawx code.
I want to try about a million things with it, but I can't yet, because there are a few more low-hanging fruit to grab first. Soon, the MCP server is going to be expanded so it will work across all your projects simultaneously, without needing to restart the MCP server to change projects.
Also, I'm adding write, edit, and delete capabilities to the MCP toolkit for rules, and their individual text parts. Once that's done, it should theoretically be possible to have your agent take any law on the web and convert it into the structured representation that §Blawx uses, and then start encoding it for you.
When those changes are made, there will be a new version of Blawx-MCP to make them available.
Later, the primary objective will be to deploy a version of Blawx-MCP live as a remote MCP server. That's going to take a little while, but once it's done it will mean that anyone with a paid §Blawx account will be able to add §Blawx to the toolkit of any AI Agent they own that supports remote MCP. So that will include web-based versions of ChatGPT, Claude, and others, spreading the utility of §Blawx to tools that legal experts are most likely to already be using.
I have recently changed strategy with regard to AI and §Blawx. For some time I have anticipated that the way to most effectively deploy AI with §Blawx would be to build custom workflows and expose them as tools through the §Blawx user interface.
I have recently come to a change of heart on that. Essentially, it seems to me that anyone who is creating software that is for people, and not for AI agents, risks wasting time and effort. Five applications, each with an agent built in, are nowhere near as valuable as one agent with the ability to use all five of those applications.
There is a huge network effect in exposing tools to agents, and far less of a benefit in adding agents to tools.
The models and techniques involved in agentic AI are also advancing so quickly that it is truly hopeless for anyone (much less a single guy spending a few hours a week) to expect to be able to build anything optimized to the current models, and keep it up to date. Or to generate a user experience on par with expectations that are driven by what the frontier model companies are delivering.
From a very practical standpoint, letting people use their own AI with §Blawx also eliminates a lot of complicated pricing problems with regard to token usage.
So going forward, I will no longer be adding built-in AI features like Chats, or the previous AI Code Suggestion feature. Instead, users will be encouraged to use §Blawx alongside agentic AI systems that they obtain elsewhere. I think that's going to allow me to focus on the things that right now only §Blawx can do, and to really accelerate exploration of Rules as Code.