BOAS.DEV / VISKA
Status: In Development
‹ §05 Iceland AI Infra
§06 · Roadmap

Six milestones. Any phase boundary is a reprioritization gate.

No rigid forward commitment beyond the current milestone. What ships is what was still the right thing to ship when the gate opened.

§06.1The Governing Principle
Rule 1
Phase-boundary reprioritization. Between M(n) and M(n+1), Viska + Boas re-scope against what was learned. No milestone executes because it was written down six months ago.
Rule 2
Shipping artifact is the milestone, not the meeting. M(n) is done when Viska can use what shipped in their workflow — not when a demo was given.
Rule 3
Dependencies are visible upstream. Every milestone names what it needs from Viska before work begins. Missing dependencies = slip, no silent retries.
§06.2Milestones M1 – M6
Milestone
M1
Macro Research Intelligence
Shipped
Ships
  • Pantheon research corpus queryable
  • Daily macro brief pipeline (FRED + CB RSS + broker research)
  • MCP workspace access for Claude Desktop
Viska sees / uses
  • Daily brief in Slack + email
  • Natural-language interrogation of 30k+ chunks, 29 broker sources, citations
  • MCP tool for ad-hoc questions
Risk
  • Broker-source licensing review (per source)
  • Corpus scope update as Viska universe evolves
Viska dependencies
  • Team onboarding to Claude Desktop
  • Priority research sources list
Milestone
M2
Multi-Agent Surveillance
Queued Next
Ships
  • 3 agents 24/7 on Viska infra — Crypto (on-chain + exchange flows), Macro sentiment (CB speech/minutes + G4 curves), Geo-risk (OSINT on jurisdictions/energy/supply)
  • Alert pipeline · L1/L2/L3 escalation
Viska sees / uses
  • L2 alerts in market hours · L3 alerts off-hours
  • Alert dashboard (CF Access) — live list, dismissal, false-positive rate
Risk
  • Alert fatigue — requires 30-day calibration window
  • OSINT source licensing review
Viska dependencies
  • Watchlist definition (tickers, jurisdictions, counterparties)
  • Alert-channel preferences (Slack · email · SMS · ntfy)
  • L2/L3 threshold policy
Milestone
M3
Trading War Room
Queued · Partial Built
Ships
  • Paper-trading desk on Alpaca
  • 4 Claude Code agents + 7 n8n pipelines
  • 8-rule institutional risk
  • 7-page React dashboard
  • 90-day paper track record before live discussion
Viska sees / uses
  • Dashboard at private subdomain (CF Access)
  • Daily pre-market brief + alerts + EOS review
  • PM Decision Card per signal
  • Trade Audit Log from signal to fill
Risk
  • Alpaca + data-source rate limits under stress
  • 90-day paper-only = discipline, not limitation
  • Credential rotation (Hades)
Viska dependencies
  • Watchlist (from M2)
  • Risk framework accept — 1/5/20/8 · −5/−10 · enum-locked
  • Session-time commitment (pre/mid/EOS)
Milestone
M4
Investor Communications
Queued
Ships
  • LP-letter pipeline (quarterly + ad-hoc · performance attribution + commentary · draft → review → send)
  • Thought-leadership (2–3/week IS + EN)
  • Meeting intelligence (pre-brief for every external meeting)
Viska sees / uses
  • Draft LP letter in Viska brand within 24h of quarter close
  • Thought-leadership calendar tied to build milestones
  • Meeting pre-briefs landing 2h before meeting
Risk
  • Voice drift — must capture Viska editorial voice, no generic flattening
  • Requires voice-capture pass (3–5 pieces + style notes)
Viska dependencies
  • Past LP letters (last 4 quarters) for voice calibration
  • Counterparty CRM (spreadsheet acceptable)
  • Editorial approval workflow
Milestone
M5
Scenario Modeling & Signals
Queued
Ships
  • Overnight scenario lab (5k Monte Carlo, regime-aware, 10/50/90 drawdown)
  • Proprietary signal generator on Viska universe + risk framework + historical fills
  • Signal registry with historical reads
  • LP-ready performance export (CSV/PDF/XLSX)
Viska sees / uses
  • Morning arrival: overnight scenario results per position + universe
  • Signal registry with current read + hit rate
  • LP attribution export on demand
Risk
  • Compute budget — Viska-hosted (Atvinnustefna-aligned) vs external cloud (cost-metered)
  • Overfitting — mitigated by walk-forward + out-of-sample
Viska dependencies
  • Compute decision (local · Iceland DC · external)
  • Historical fill data (last 2 years)
  • Universe confirmation
Milestone
M6
Distribution Intelligence & Handoff
Queued
Ships
  • Partner surveillance pipeline (distribution partners, allocators, introducers)
  • Codebase handoff to Viska team or designated maintainer
  • Full ops documentation — architecture, incident playbook, credential recovery, renewal cadence
  • Post-engagement maintenance option contract
Viska sees / uses
  • Partner activity digest
  • Full codebase access (already Viska's)
  • 3–5 day training with Viska technical lead
  • Decision gate: internal maintain vs contract
Risk
  • Handoff quality depends on who Viska designates as technical owner
  • No "leave + hope" allowed — if no owner, contract is only safe path
Viska dependencies
  • Technical owner designation (or contract decision)
  • Doc review cadence
  • Credential rotation ownership
§06.3Cross-Milestone Dependencies
Watchlist
Defined in M2 consumed by M3M5. One definition, propagated.
Risk framework
Accepted in M3 used by M5 (signal thresholds) · referenced in M4 (LP commentary).
Voice / brand
Captured in M4 reused in M6 (handoff doc style, partner digest style).
Credential rotation
Touched in M3M5M6. Single owner at Viska prevents drift.
Iceland compute
Optional for M5 but becomes preferred if Atvinnustefna grant participation pursued. Decision gate at M4/M5 boundary.

One reason phase-boundary reprioritization matters — a change in one milestone can reshape the dependencies of the next three.

§06.4What This Plan Is Not
Not a fixed scope. Every boundary is a renegotiation gate.
Not a timeline commitment in weeks. Milestones are gated on acceptance, not calendar.
Not a total budget. Each milestone's deliverables + dependencies are the commitment. Line-item changes are a conversation, not a failure.

The commitment is the method, not the roadmap.