Skip to content

Releases: mwsp-code/IC-Copilot

IC Copilot v0.1.0 — Evidence before eloquence

Choose a tag to compare

@mwsp-code mwsp-code released this 20 Jul 00:28

The first public release of an evidence-first Investment Committee Copilot for US-listed equities and ADRs.

Watch the workflow

Watch the 91-second narrated AAPL walkthrough

Open the live Streamlit app

Highlights

  • Story-first IC one-pager with verdict, causal bridge, counter-thesis, and next action.
  • Frozen no-key demos for AAPL, NVDA, BABA, TSLA, GS, and SPCX.
  • SEC filing comparison, XBRL normalization, management-source intelligence, and peer metric read-through.
  • Candidate, Research-Ready, and High-Conviction evidence gates.
  • Company model, model-derived scenarios, reverse DCF, and market-implied expectations.
  • Price attribution, price-only market capture, and optional point-in-time consensus history.
  • Optional guardrailed LLM synthesis and secondary critique.
  • OS-keychain credential storage and provider-health diagnostics.
  • No-network research-integrity benchmark and 377 collected regression tests.

Known limitations

  • Live provider coverage depends on user keys, network access, and vendor entitlement.
  • Historical analyst revisions require snapshots captured at the time or licensed/manual imports.
  • Global official-source adapters remain progressive rather than universal.
  • Scenario probabilities are illustrative until sufficient resolved outcomes exist.

Integrity promise

Missing evidence remains Unknown. LLM output cannot independently create citations, targets, calibrated probabilities, or High-Conviction promotion.

IC Copilot produces research hypotheses, not financial advice, and does not execute trades.