
Every software organization is bleeding from decisions made years ago by engineers who are no longer around. A constraint baked into a 2019 architecture silently shapes every system built in 2026. Nobody documented it. The person who understood it left. One day it detonates a multimillion-dollar outage blamed on “architectural debt” without revealing where the next failure hides. Nexus solves this using IBM Bob as its foundation. Bob is the only AI capable of reading an entire repository simultaneously full git history, PRs, commit messages, and architecture docs together. That full-repository semantic context enables causal tracing across components built years apart by different teams. Bob extracts Decision Provenance Records: structured captures of architectural choices, rejected alternatives, assumptions, constraints, and involved engineers. These become nodes in a Neo4j Causal Temporal Graph, where edges encode relationships between decisions across time. Bob queries this graph in plain English, surfacing answers no search tool can find. Nexus continuously runs an Assumption Decay Monitor, evaluating whether assumptions still hold and ranking risks by blast radius how many downstream decisions fail if an assumption breaks. Demonstrated on PostgreSQL: Bob analyzed the full repository, extracted 15 Decision Provenance Records, built a graph of 162 nodes and 202 relationships, and identified four assumptions actively breaking today. Knowledge concentration analysis exposed engineers whose departure creates critical single points of failure. The result is a living organizational risk forecast: assumption decay trajectories, knowledge concentration scores, and prioritized remediation. Every AI tool in 2026 competes on velocity. Nexus competes on comprehension.
17 May 2026