Data Summit Warsaw talk

Workshop for preparing a 35-minute practitioner talk for Data Summit Warsaw: two observed Commonplace cases that earn one core claim and four engineering lessons, with the deeper KB theory used only where it explains a design decision.

Core claim of the talk:

LLMs make it cheap to develop a knowledge base continuously, not merely search or summarize it. Reliability comes from a hybrid design: generated knowledge stays reviewable, stable and checkable rules move into deterministic machinery, and retrieval loads only what the current task needs.

The talk should leave the audience thinking "I can apply these patterns to my own LLM knowledge system next week", while the theory stays in the KB for those who follow the rabbit hole. It is deliberately not a tour of Commonplace's vocabulary — the material-selection table in outline.md records what is included, compressed, or cut, so later sessions don't re-litigate those cuts.

State

  • outline.md — the working 35-minute outline, the vibe-noting and tag-README cases, the four practitioner lessons, the material triage against the KB, the visual spine, and the deliberate exclusions.
  • development-implications.md — the provisional Commonplace roadmap implied by the talk: test an observable loop from raw observation through candidate disposition and cold-session reuse before adding more infrastructure.

What would close this workshop

The talk is delivered (or the submission is withdrawn). On closing: extract anything durable — most likely the "write for reuse; load for the task" formulation and any slide-tested framings that turn out stronger than their KB counterparts — into kb/notes/ or kb/articles/, then delete the workshop.

Boundaries

  • Slides and speaker notes are the deliverable; they may live here while in flight.
  • Presentation-level syntheses (e.g. "write for reuse; load for the task") are allowed to be looser than library claims — they are not KB theorems until promoted through normal review.
  • Timing target for authored material is ~29–30 minutes; the concrete examples are the talk and will create rehearsal variance.

Complete file listing (generated at build time)