Algolia-Central

Messaging-Positioning-Pivot.md

Messaging & Positioning Pivot — Algolia Central

Vault file purpose: holds directional commitments, research gate, workstream roster, pilot targets, production-readiness requirements, and critical open questions. Cross-references the repo scratch files for full depth.

Live scratch (verbatim Arijit, unedited): rc3-phoenix/docs/superpowers/plans/2026-04-22-messaging-positioning-pivot-scratchpad.md

Live structured plan (full detail, §1–§18): rc3-phoenix/docs/superpowers/plans/2026-04-22-messaging-positioning-pivot.md


1. Forcing Function

Adobe Summit 2026, Apr 20–22. Pilots on the table (Adobe Brand Concierge want-in, MasterCard very interested, direct customers + SI partners curious). Story resonates ("nobody said no, nobody said a lot"). Pilot package must exist to convert interest → pilots → productized offering.

Arijit framing: "That will be my legacy here. POC is done. The story is resonating. How do we take this to market and productize it into an Algolia product offering?"


2. Directional Commitments (declared)

These are operating commitments — not fully specified decisions, but the direction is locked.

  1. POC → Product transition. rc2-algolia + Maverick/Elena/Bruno = Phase 0 Bootstrap Baseline. The product is built from research-grounded layers, not incremental v2.
  2. Native Algolia as the refactor target. Anything rc2-algolia reinvents that Algolia provides natively — delete.
  3. Reference implementation = the product. Prescriptive, opinionated, per-use-case. Not a toolkit.
  4. Neural-search configuration is proprietary IP. Internal tuning expertise is the moat.
  5. Category expansion beyond "product search." "Content search" was rejected by Arijit as insufficient framing. New category name TBD (R9).
  6. Agent Factory before product. Build the army of specialized agents; the factory produces Algolia Central.
  7. Live refactor, no big-bang. Availability preserved throughout. Per-layer cutover.
  8. Production-readiness is architectural, not QA. Guardrails, security, brand, zero-hallucination baked into every layer.
  9. Shared-responsibility delivery model. Algolia ships data + guardrails; customer owns front-end + audit.
  10. Research is the gate. R1–R12 closes before agents are re-specified.

3. Research Gate (R1–R12)

# Topic Owner Deliverable Status
R1 Algolia Connector inventory TBD Source × availability × maturity × gap matrix open
R2 Normalization into index (schema + weights) TBD Canonical schema + weight rubric open
R3 Enrichment (must/can/should per field) TBD Enrichment matrix open
R4 Crawler strategy (web/wiki/Confluence) TBD Crawler playbook + source-type coverage open
R5 Transformation engine deep-dive TBD Transformer feature-truth + decision tree open
R6 Enrichment engine deep-dive TBD Fetch feature-truth open
R7 Canonical data structure + per-source normalization TBD Canonical schema + per-source rules open
R8 Neural-search configuration mastery TBD Tuning cookbook + baseline config open
R9 Category framing / positioning naming TBD Named category + JTBD + 5 tested phrases open
R10 Use case spec (FS/MasterCard + BC-feed) TBD One brief per pilot open
R11 Production readiness (guardrails/sec/brand/zero-hallucination/legal) TBD Per-layer spec + contract templates open
R12 Bootstrap defect inventory (Maverick, handoff) TBD Defect inventory with severity + fix-when open

Full detail on each R-item: rc3-phoenix/docs/superpowers/plans/2026-04-22-messaging-positioning-pivot.md#13-research--feasibility-markers


4. Workstream Roster (WS1–WS9)

# Workstream Depends on Enables Lead
WS8 Agent Factory / Dev Engine all other WS (parallelism) TBD
WS1 Architecture R1–R11 WS2, WS3, WS4, WS7 TBD
WS2 Design WS1 WS3 TBD
WS3 Refactoring (humongous) WS2 Pilot-readiness TBD
WS4 Use Case Definition WS1 + partial R Pilots TBD
WS5 Story / Story Building (parallel — no blockers) WS6 TBD
WS6 Presentation (landing, decks, campaign) WS5 Pilot outreach TBD
WS7 Production Readiness WS1, WS2 Pilot trust TBD
WS9 Bootstrap Defect Remediation R12 Demo stability TBD

Dependency graph + agent assignments: rc3-phoenix/docs/superpowers/plans/2026-04-22-messaging-positioning-pivot.md#15-workstreams

Assignment principle: Each workstream gets a human owner (Arijit or delegate) + a specialized agent (per §17 factory). Start WS8 (factory) first — all others accelerate once it's online.


5. Pilot Targets (stated so far)

  • Adobe Brand Concierge — direct pilot, want-in confirmed at Summit.
  • MasterCard — very interested (verify spelling — dictation "MasterGard"). Financial services vertical.
  • 1–2 direct customer pilots — TBD (curiosity expressed at Summit).
  • 1 Adobe-partnered pilot — 2 customers inside, Adobe as distribution partner.

Objective: prove it out → build the package.


6. Production Readiness — Four Pillars + Two Vulnerability Axes

Pillar What it covers
Guardrails Input/output filters, prompt injection defense, policy enforcement, content safety
Security Auth/authz, tenant isolation, encryption, audit logging
Brand Protection Voice/tone, forbidden topics, competitor handling, crisis patterns
Zero Hallucination Retrieval grounding, confidence thresholds, "I don't know", citation discipline

Vulnerability axes: Legal (contracts, liability, audit-trail) + Technical (exploit paths, SLOs, incident response).

Shared-responsibility matrix (Algolia ↔ Customer): see ...plan.md#16-production-readiness — scribe-drafted, pending legal validation.


7. Critical Open Questions

These must resolve before heavy execution on dependent workstreams.

Q Question Blocks
Q3 Current pitch verbatim — needs capture WS5
Q8 Mode A vs Mode B narrative weight in pitch WS5, WS6
Q11 Does Maverick/Elena/Bruno roster survive refactor? WS1, WS2
Q13 What replaces "content search" framing? WS5, WS6, WS4
Q14 FS/MasterCard use case shape WS4, R10a
Q15 Brand Concierge data architecture + integration points WS4, R10b
Q16 What survives from current bootstrap data layer? WS3
Q18 Agent factory starting roster WS8
Q19 Live refactor cutover strategy WS3
Q20 Fix Maverick-skips-follow-up now or in refactor? WS9
Q21 Shared-responsibility boundaries per risk class WS7
Q22 How many parallel Claude instances is realistic? WS8

Full list: ...plan.md#4-open-questions


8. Dictation Decodes (confirm when convenient)

  • "MasterGard" → MasterCard ?
  • "AAP" → AEP (Adobe Experience Platform) ? Or AAM?
  • "AP market oh" → AEP, Marketo ?
  • "Alina", "Bruna" → Elena, Bruno ? (per agent roster)
  • "Leah central" → Algolia Central ✓ (confirmed contextually)
  • "content concentrates" → "content concentrations" / "content silos" / "content sources" ?
  • "three slider or two slider" → 3-slide / 2-slide pitch decks ?

9. Paused Work

  • Pipeline audit Phase 3rc2-algolia/docs/superpowers/plans/2026-04-16-pipeline-audit-and-cleanup.md. Browser trace capture paused. Folds under WS3 + WS9 going forward.

10. Cross-References

  • Live scratch (verbatim): rc3-phoenix/docs/superpowers/plans/2026-04-22-messaging-positioning-pivot-scratchpad.md
  • Live structured plan (§1–§18): rc3-phoenix/docs/superpowers/plans/2026-04-22-messaging-positioning-pivot.md
  • Paused pipeline audit plan: rc3-phoenix/docs/superpowers/plans/2026-04-16-pipeline-audit-and-cleanup.md
  • Project TODO: Projects/Algolia-Central/TODO.md

11. Vault Update Protocol (for this file)

  • Frequency: After each major capture checkpoint in the plan doc, or when Arijit confirms a new directional commitment.
  • Content rule: directional commitments + research gate + workstream roster + critical open questions only. No scratch, no raw dumps — those live in the repo.
  • Update trigger phrases: "commit this", "lock it in", "add to vault", or scribe-surfaced at end of session with Arijit's ack.