Tinderbox Meetup 31MAY26: H1 2026 Reflections AI-Assisted Thinking, Reader View, Writing for AI
| Level | Intermediate |
| Published Date | 6/2/26 |
| Type | Meetup |
| Tags | AI-Assisted Thinking, H1 2026 Reflections, Reader View, Writing for AI, 5CKM, 5Cs of Knowledge Management, Eastgate, Identity Praxis, Inc., Mark Berstein, Michael Becker, Tinderbox |
| Video Length | 01:27:35 |
| Video URL | https://youtu.be/Bm6ZZFtonOk |
| Example File | TBX Meetup 31MAY26_Chat.txt (9.0 KB) |
| TBX Version | 11.5 |
| Instructor | Michael Becker |
In this Tinderbox meetup, we reflected on what the community has learned during the first half of 2026 and began thinking about what we want to explore in the second half of the year. Much of the conversation centered on the rapidly evolving relationship between Tinderbox, AI, programming, documentation, and knowledge work.
Mark Bernstein opened by describing recent progress with Tinderbox components in JavaScript and the emerging Tinderbox web reader. This led to a broader discussion about AI-assisted programming, “vibe coding,” and the tension between quick, disposable code and long-term maintainable software architecture. Participants explored how AI can help generate, inspect, refactor, and explain code, while also noting the risks of undocumented, brittle, or poorly understood outputs.
A major theme was how to communicate effectively with AI. Mark Anderson described his work on creating an AI-oriented primer for the Tinderbox TBX/XML format. He argued that AI does not read like a human and that, when we want it to understand complex systems, we should write deliberately for the AI rather than assuming human-facing documentation is enough. This included discussion of XML, HTML, definition lists, inheritance, prototypes, and the importance of concise, structured explanations.
The group also discussed the Tinderbox reader as a possible new publishing, training, and storytelling environment. We considered whether it could support learning tools such as flashcards, read-state tracking, adaptive navigation, credentialing, and provenance of authorship. This led into a larger conversation about adult learning, emphasizing relevance, problem-solving, case studies, group work, and helping learners overcome fear by giving them hands-on experience with tools.
Several technical possibilities were raised, including using JavaScript libraries and npm packages to extend Tinderbox functionality, speeding up runCommand, connecting Tinderbox with local AI models, improving MCP integration, and using the web reader to make Tinderbox documents accessible from browsers, iPads, and other devices. Participants also revisited RSS, web standards, and earlier experiments in web-based Tinderbox access.
The meetup concluded by returning to the core practice of Tinderbox work: incremental formalization. We discussed the value of noticing when a piece of information may need to become metadata, trusting that impulse, and capturing it early in attributes, prototypes, rules, or edicts. The group framed this as a shared pattern across many different kinds of work: using Tinderbox to think, structure, analyze, and eventually transform knowledge into useful outputs.
Looking ahead, we identified several possible themes for the second half of 2026: AI-aware documentation, Tinderbox as a publishing and learning environment, provenance of human and AI-authored work, adult learning and training workflows, and the challenge of helping individuals and organizations turn personal knowledge work into repeatable business processes.