Feature: Investment Thesis Drift Detection

[coderleeon]

Problem

Investment research is often revisited as new earnings, news, and filings become available. While AI Berkshire produces structured reports, it is difficult to identify exactly how the investment thesis has changed between research sessions.

Users currently need to manually compare reports to determine whether changes are due to updated fundamentals, valuation shifts, or changes in qualitative assumptions.

Proposed Solution

Add a "Thesis Drift" comparison mode that compares two research reports for the same company.

Example output:

• Valuation: Improved
• Management Quality: Unchanged
• Competitive Moat: Weakened
• Capital Allocation: Improved
• Overall Recommendation: Changed from "Watch" to "Buy"

The report could also highlight which new evidence caused each change.

Alternatives Considered

Manual report comparison is possible, but becomes increasingly difficult as research is updated over months or years.

Additional Context

This would make long-term investment tracking significantly easier and reinforce disciplined decision-making by clearly separating changes in evidence from changes in market price.

[xbtlin]

Thanks for the thoughtful proposal — I really like this idea, and it fits the spirit of the tool well. Tracking how a thesis evolved over time is exactly the kind of discipline this should reinforce.

It's not something I can prioritize in the near term, so I can't put a timeline on it yet — but I'd like to keep it on the roadmap. If anyone (including you) wants to take a stab sooner, I'd gladly review a PR; otherwise I may come back to it down the line.

One thing worth flagging for whoever picks it up: the hard part isn't the diff itself, but distinguishing a real thesis change from LLM phrasing variance — two reports with identical fundamentals might word "moat" differently. The comparison would need to reason over the underlying evidence, which likely means reports should emit a consistent set of structured dimensions first.