2 · Early-warning signals — catch thesis-relevant shifts before consensus, prioritised by the Strategic Time Grid's forward intensity ramps.
3 · Contradiction / thesis-test — adversarially hunt evidence that refutes active theses.
Cadence: hybrid — scheduled baseline sweeps + on-demand surge for events (FOMC, earnings, a named thesis).
No new infrastructure. Every play composes existing capabilities (research-sweep, youtube-intel/youtube-search, x-sweep, firecrawl, research-orchestrate, background-research, score-findings, vault search-notes/browse-notes) around one new asset: the Strategic Time Grid (research/strategic-time-grid.md).
Architecture
Two collection engines feed one synthesis/delivery layer. Enhancements plug into the engines.
[Strategic Time Grid] --ramp deltas--> (A) Grid-Ramp Radar ---\
>--> (C) Committee Brief --> Slack #feed
[Active theses register] --------------> (B) Contradiction Tracker --/ + Viska-Wiki
+ on-demand brief
(D) Voice divergence --- source-lane into A & B
(E) Deep-theme dossier --- surge mode of A (research-orchestrate)
Collection layer
A
Grid-Ramp Radar
scheduled baseline · goal 2
Input: the grid's per-theme quarterly intensity. Compute a forward ramp delta per theme over the next 1–2 quarters (Ekkert→Lágt, Lágt→Sterkt = positive ramp).
Themes with a positive upcoming ramp get research priority now — collect signal before the theme intensifies. Run filtered sweeps (RSS/youtube/X/firecrawl) scoped to ramping themes.
Score V×R×E (score-findings) with a timeliness/novelty boost for ramping themes.
Output: ranked finding blocks tagged by theme + ramp state.
Dependency: ramp computation needs the full 2027–2029 matrix (pending source). Runs on 2026 deltas in the interim; richer once locked.
B
Contradiction Tracker
scheduled re-checks · goal 3
Input: register of active theses — TIER-1 Sterkt grid themes + fund position/thesis cards (funds-vault positions/).
For each thesis, run adversarial sweeps seeking disconfirming evidence. Maintain a contradiction ledger: open contradictions tracked until rebutted or escalated (model case: the unrebutted CF long-end-supply loop).
Enhancements (fold into engines; build after core)
D
Voice Divergence Monitor
Track the 6 fund voices (Alden, Gromen, Howell, Wang, Werner, Hayes) for new grid-aligned theses and divergence between them (disagreement = signal). Source-lane into A & B.
E
Deep-Theme Dossier
research-orchestrate multi-lane deep dive on the single highest-ramping theme per cycle. Surge mode of A.
Data flow
Grid → ramp deltas → A's theme priority list (refreshed per cycle).
A + B sweeps → raw findings → score-findings (V×R×E + boosts) → scored, theme-tagged blocks.
Scored blocks + contradiction ledger → C synthesis → brief.
Brief → Slack #feed + wiki note. Findings worth persisting → wiki research/ (and corpus iff the gate decision ViskaRes-003 lands an ingest path).
Build sequence (incremental)
C skeleton — grid-structured committee brief over the existing morning-report flow. Visible product fast.
A radar — wire the grid-ramp priority list as C's primary feeder (2026 deltas now).
B tracker — contradiction ledger + adversarial sweeps; seed with the CF loop.
D + E — voice lane + deep dossier surge mode.
Scheduling — move baseline sweeps to scheduled/cloud-triggered runs (hybrid).
Dependencies & open questions
Item
Status
Full grid matrix (2027–29) for richer ramp deltas
pending source file / funds-vault themes/
Active-theses register for B
derive from grid TIER-1 + funds-vault positions/ (read path TBD)
Scheduling surface
cron/cloud-trigger pattern (cf. armed Warsh re-sweep); confirm attribution flag for committing runs
Scoring weights
V×R×E boost magnitude for ramp/timeliness; tune after first cycle
Corpus path
whether briefs/findings also embed depends on ViskaRes-003 (Option 4 inverted gate)
Out of scope
Rebuilding any sweep tool (all exist).
The morning-report mechanics (mimir-post Slack).
The corpus-gate redesign itself (separate decision, ViskaRes-003).