17 products. 6 months. One system that compounds. Each product builds on the last — the knowledge base connects everything, the portal surfaces it, and the agents run 24/7.
This isn't a chatbot upgrade. It's a complete AI intelligence department — research, trading, communications, and operations — built for a team of 7.
These systems are running in production right now. The foundation everything else builds on.
Ask anything about your broker research — "What does Goldman say about oil?" — and Mímir searches your private corpus, synthesizes across brokers, and responds with sourced citations.
ChatGPT knows what the internet knows. Mímir knows what your broker reports say. Every claim links back to the specific report, page, and date.
Every morning, Mímir clusters the day's documents by topic, ranks themes by broker consensus, and generates a structured brief prioritized for Viska's portfolio.
Why it matters: 238 PDFs arrive daily. The brief means the important insight on page 12 of a Deutsche Bank note reaches the people who need it — before the market moves.
Type /report after any Mímir conversation. The system extracts key findings, preserves citations, structures for the chosen audience, and generates a permanent link.
Make the existing system bulletproof and add the interfaces that multiply access. Your corpus becomes available everywhere — not just Slack.
Your team already uses Claude or other AI tools. MCP means you can pull Mímir's research into any conversation — "What do our brokers say about semiconductor supply chains?" works from Slack, Claude Desktop, or an IDE.
MCP is an open protocol that lets AI assistants connect to external data sources. Mímir exposes its search as an MCP server. Any compatible tool can query it.
Impact: Cross-references become possible — "Goldman says inflation is cooling, but today's CPI print contradicts their view." The system connects broker opinion to primary data.
From one bot answering questions to specialized agents running 24/7 surveillance — backed by institutional memory that connects everything.
Mímir does vector search — find text similar to your question. The knowledge base adds temporal reasoning ("what did we decide before the rebalancing?"), relationship traversal ("all LPs who attended meetings about Fund X"), and decision archaeology ("what was rejected, and why?").
This is the difference between a search engine and institutional memory.
On-chain metrics, exchange flows, funding rates, liquidation data. Reports anomalies — unusual accumulation patterns, whale movements. Writes findings to the knowledge base.
Cross-asset radar — FX, rates, commodities, equity indices. Daily conviction scoring. Flags divergences when data contradicts broker consensus.
Regulatory filings, sanctions, elections, policy shifts. Monitors jurisdictions relevant to your portfolio. Flags risk events before they become consensus trades.
Feed any transcript, get structured output: key metrics, guidance changes, management tone shifts. Compare across quarters — "How has Meta's capex guidance changed over 4 quarters?"
Track how a broker's view evolves — "How has Goldman's oil price target changed in 3 months?" with specific quotes and dates. Narrative drift detection — when language shifts from "cautiously optimistic" to "neutral," the system flags it.
Move from research to action. Live market intelligence, technical analysis, trade journaling, and mechanical risk discipline. Most of the automation layer is already built — Month 3 is deployment.
Real-time social sentiment from X via Grok — catches narrative shifts before they move price. Paired with Claude for deeper analysis.
4 agent repos, 7/8 n8n automations live, Alpaca brokerage integration, FRED economic data, QuantOracle (63 quant tools), risk framework approved, dashboard designed.
Turn the intelligence system outward — LP reporting, thought leadership, brand building. The bottleneck isn't performance, it's communication bandwidth.
Drafted from portfolio data + market context. Performance attribution auto-generated. Consistent voice across every letter.
Viska's edge is macro thinking. The content engine solves the production bottleneck without diluting intellectual quality — AI drafts from your corpus, your team reviews.
Feeds the meeting briefing system, the content engine (personalized outreach), and the knowledge base (relationship history as institutional memory).
Unique capabilities no off-the-shelf tool provides. The system starts thinking ahead — modeling scenarios, watching partners, identifying blind spots.
Claude's extended thinking mode processes your full research corpus, market data, and historical analogies overnight. Output: scenario description, probability assessment, portfolio impact, recommended actions, supporting evidence.
Weekly automated scan. Turns partner due diligence from a one-time exercise into continuous surveillance.
The system runs without me. You own everything — GitHub repo, all infrastructure, all documentation. 100% handoff.
Train a lightweight model on your team's research output, IC minutes, and historical positions. The model learns your analytical framework — what factors matter, how you weight them, what patterns you act on.
What it is NOT: A black box that trades for you. It's a tool that translates your thinking into systematic signals. Every signal requires human interpretation and approval.
After handoff: Optional maintenance retainer (ISK 250,000/month). The system runs independently — you can bring in any developer, or continue with support.