The AtmosphereConf talks your skyline missed
docs(spec): network attention display review round 1
- Simplify 'missed' detail text to static 'Your network missed this'
(attentionInverse is always 1.0 for missed state, so percentage is
always 100% and adds no information).
- Use percentage for 'engaged' state instead: 'X% of your network
missed this' — the gradient IS meaningful there.
- Add className='h-full' to LumeCard in code snippet.
- Add note to copy card content JSX from existing page.tsx.
- Fix prose contradiction about unauthenticated fetch behavior.
+10
-7
+10
-7
docs/superpowers/specs/2026-04-10-network-attention-display.md
+10
-7
docs/superpowers/specs/2026-04-10-network-attention-display.md
···
140
140
{scoredTalks.map(({ talk, score }, index) => (
141
141
<Link key={talk.rkey} href={`/talk/${talk.rkey}`}>
142
142
<LumeCard
143
+
className="h-full"
143
144
glowIntensity={score?.intensity ?? 0}
144
145
tileIndex={index}
145
146
score={score}
146
147
>
147
-
{/* Card content: speakers, title, chips — same as current */}
148
+
{/* Card content: speakers, title, room chip, duration chip.
149
+
Copy the existing JSX from talks/page.tsx (the <div className="p-5">
150
+
block with speakers, h2 title, and metadata chips). */}
148
151
</LumeCard>
149
152
</Link>
150
153
))}
···
179
182
180
183
| `score.state` | Text | Color |
181
184
|---|---|---|
182
-
| `"missed"` | `"{X}% of your network missed this"` | `text-primary-fixed` (mint) |
183
-
| `"engaged"` | `"Discussed by {N} of {T} follows"` | `text-on-surface-variant` (muted) |
185
+
| `"missed"` | `"Your network missed this"` | `text-primary-fixed` (mint) |
186
+
| `"engaged"` | `"{X}% of your network missed this"` | `text-on-surface-variant` (muted) |
184
187
| `"unknown"` | (no detail strip) | — |
185
188
| `null` | (no detail strip) | — |
186
189
187
190
Where:
188
-
- `X` = `Math.round(score.layer1.attentionInverse * 100)` — e.g., 97
189
-
- `N` = `score.layer1.uniqueFollows` — e.g., 3
190
-
- `T` = `score.layer1.totalFollows` — e.g., 423
191
+
- `X` = `Math.round(score.layer1.attentionInverse * 100)` — e.g., 83
192
+
193
+
**Why no percentage for "missed":** The `missed` state means `uniqueFollows === 0`, so `attentionInverse` is always exactly `1.0` — the percentage would always read "100%". A static message is clearer and avoids pointless math. For `engaged` talks, the percentage IS meaningful: "83% of your network missed this" conveys the gradient between "almost nobody talked about it" and "most of your follows discussed it."
191
194
192
195
**Interaction:**
193
196
- Desktop: detail appears on `:hover` via CSS transition (opacity 0 → 1, translateY(4px) → 0). No JavaScript state needed.
···
242
245
243
246
## 8. Edge Cases
244
247
245
-
- **Not authenticated:** Grid renders without scores. No `/api/crawl` call is made (the hook fetches, gets 401, sets `mentions=null`). Same as today.
248
+
- **Not authenticated:** Grid renders without scores. The hook fires `fetch("/api/crawl")` which returns 401; the hook treats this as "no data" and sets `mentions=null`. Same visual as today.
246
249
- **Zero follows:** `/api/crawl` returns `followCount: 0`. `rankTalks` produces all `unknown` states. Grid renders without glow. Acceptable — the user needs follows for the scoring to be meaningful.
247
250
- **Crawl timeout (30s):** `/api/crawl` returns 504. Hook receives error. Grid stays in Phase 1.
248
251
- **Partial crawl data:** Some talks have mentions, some don't (out-of-scope talks). `rankTalks` classifies absent talks as `unknown`. They sort to the bottom.