SG/Sentinel Future Research And The Path To MVP

Summary

The boundary marker: "we have enough design to focus next on building the first MVP" — further design has diminishing returns until implementation surfaces real-world learning. It lists the memos still worth recording (the fast-track reconception first; incident response and the after-block lifecycle; Sentinel's own failure modes), the prior-art deep-dives to run during implementation (OPA, Sigma's compile pipeline, Coraza, deception patterns, BeyondCorp), what is deliberately parked, and a ten-step MVP sequence with its definition of done — including the cost measurement that closes the founding argument's loop.

Key concepts

• The MVP boundary — what to build first, what to defer — the roadmap page follows this discipline
• The parked list — parking is sequencing, not abandoning
• Gaps gate steps — each MVP step lists the tabletop resolutions it depends on

Key ideas

• Fail-open versus fail-closed for Sentinel itself is an open question flagged as existential — the guard's own failure modes need their own memo.
• The concrete cost model is the loop that validates the founding motivation.
• The v0.27.59/60 implementation pair narrowed even this brief's scope further — tiny core won.

On this site

The roadmap page is this brief's discipline applied to the built MVP.

Read the document