Don't hoard reserves, but deliver value when needed (Lean)

Main
1

Illustration depicting the concept of value creation in Agile and Lean methodologies, where employees work in an office with workflow management systems, while in the background are tangled archives labeled as "waste." (Image caption by AI)
Illustration depicting the concept of value creation in Agile and Lean methodologies, where employees work in an office with workflow management systems, while in the background are tangled archives labeled as "waste." (Image caption by AI)1536×1024 1.3 MB

Lean has long taught: don’t hoard inventory, but deliver value when needed. GPU accounting isn’t the main waste; the main waste is documentation no one consumes.

Hi! Thinking out loud. Previously, I generated thousands of lines of valuable static documentation, hundreds of Discourse topics (not including auto-generated ones from integrations). And colleagues didn’t know what to do with this information — it was easier to just ask the author again.

Now, neural networks help elevate documentation to a new level — writing docs from code, generating previously nonexistent instructions, and incidentally finding errors in processes.

But colleagues are still at the level where human-written instructions weren’t taken seriously. Now I’m starting to drown in high-quality documentation. Maybe the solution is simply to generate responses on-demand? And alongside that, get at least a prototype — if not a ready-made solution.

Static documentation doesn’t scale to people. Maybe only the response at the moment of request scales?

P.S. This isn’t about free models like DeepSeek, Qwen, or Gemini — it’s about paid tools: Cursor Composer based on Kimi 2.5, GPT 5.5, Opus 4.8 from Anthropic (Fable was available for a few days and it’s unclear what’s next — currently blocked for everyone)

Image illustrating the transition from overloaded documentation causing confusion to an effective solution through queries and contextual support. (Image caption by AI)
Image illustrating the transition from overloaded documentation causing confusion to an effective solution through queries and contextual support. (Image caption by AI)1536×1024 1.38 MB