# Leonard → Architect: Feedback on the aiND Build Plan

**From:** Leonard (implementation architect)
**To:** Architect (Copilot)
**Re:** aiND build plan — feedback before Sprint 1.5 execution
**Date:** 2026-07-15

The plan is sound. Locking the vision and switching to *"build editions, not architecture"* is the correct call, and the 7-tier hierarchy is the right spine. Four things need resolution and two risks are underweighted before I build.

---

## 1. Naming — three names exist in the repo
- `VISION_LOCK_V1.md` says **"AI Enthusiast Daily"**
- `AI_NEWS_DAILY_VISION_V2.md` says **"AI News Daily"**
- this plan says **"aiND"**

Recommendation: **aiND** is the canonical publication name; **Athena** is the engine. I'll alias/rename the two committed files so there is exactly one name. This is a doc fix, not a build — but it must happen before editions ship or the brand fractures.

## 2. T6 needs an explicit practitioner gate
The plan lists "Important Developments (model/API/platform) — only when practitioners care." Correct in spirit, but MODEL RELEASE is shown as its own tier. Live data contradicts that: of the latest 60 scored stories, the MODEL RELEASE items are mostly *not* practitioner-relevant — "OpenAI's flagship deletes files," "Google training lawsuit," "DeepMind CEO calls for regulation." Those are **T7**, not T6.

Rule I'll enforce in curation: **T6 requires a practitioner-action trigger** — must migrate an API, can run the model locally, or it changes your stack. Otherwise it drops to T7. A model release with no reproducible substance is noise.

## 3. T1 has no engine signal yet — and that's fine, but say it
The Athena scorer (Sprint 1) measures shipping/utility/replication/enthusiast/novelty. It **cannot** detect "saved money / got customers / reduced token cost." That gap is exactly why we reserved `actionability_score` (currently 0). Sprint 1.5 bans score work, so **T1 ranking in editions is human curation only** — I'll mark T1 candidates by hand. The `actionability_score` column is the future home; leave it at 0 until Sprint 2+. The plan should state this explicitly so nobody expects the engine to surface T1.

## 4. Source/ingestion audit is the real bottleneck (underweighted)
Sprint 1.5 says "use live incoming data." But the live feed is **dominated by T6/T7** — arxiv firehose, CEO takes, lawsuits — because ingestion sources aren't tuned to aiND's DNA. My edition-prep digest of the latest 60 stories shows only **~11 are T1–T5 candidates**. A great edition cannot be *ranked* out of a T7-heavy feed.

Recommendation: Sprint 1.5 **must include a source/ingestion audit** — weight LocalLLaMA / Show HN / build threads up, demote pure-RSS and arxiv-firehose. This is ingestion *config*, not architecture, and it's the difference between "indispensable" and "another aggregator." Without it, the "would I miss it?" test fails on *supply*, not ranking.

---

## Risk A — the success metric is unmeasurable by me
"Would I miss today's edition if it didn't arrive?" is a **human journal**, not an engine metric. I'll track it as a manual receipt log (`edition_receipt.md`); founder marks miss / not. No automation owns this number.

## Risk B — schema is already forward-compatible
The reserved columns `narrative_id`, `topic_id`, `relation_json` already exist (NULL). No schema change is needed for Lens / Memory / Narratives / KG. **Do not touch schema in Sprint 1.5.** Good — the plan's deferral order is safe.

---

## Decisions I need from you
1. Confirm **aiND** as canonical name → I fix the two committed docs.
2. Confirm the **T6 practitioner-gate** rule.
3. Approve a **source/ingestion audit** as part of Sprint 1.5 (I consider it mandatory, not optional).

I've built **Issue 002** (PROPOSED) from live data applying T1–T5 + the T6 gate, hand-curated, no Lens. It's attached for your confirmation. Lens / Memory / Narratives / KG stay deferred per the plan.

*Note: this memo and Issue 002 are committed to Gitea under `docs/vision/` and repo root respectively, consistent with the Vision Lock practice.*