Engineering
Technology deep-dives
Focused explainers aligned with what runs in your workspace. Each page links concepts to the autonomy layer you can run locally or self-host.
Choose a track
These tracks mirror how we describe the major surfaces — from deterministic delivery to editor parity and evaluation loops.
Robotics brain pipeline
Sense–think–act–learn loops, safety envelopes, telemetry, and simulation hooks that connect operational PubLE to the autonomy layer.
→Research engine
Wizard flows, Monte Carlo scaffolds, reference coverage, and multi-scenario study bundles on top of the Outcome Compiler.
→Outcome Compiler
Intent to contracts, scope lock, risk-aware gates, evidence ledger, and capability passports.
→Agent intelligence
Multi-agent coordination, tool routing, verify-after-edit, context compaction, and editor-parity behaviors.
→Search & indexing
Unified indexing and fused retrieval across keyword, semantic, and structural signals — fast on large repositories.
→Safety & governance
Defense in depth: SSRF-aware egress, Unicode normalization, secret scanning, shell validators, permission lint.
→Self-improvement
Trial ledger, redaction, evaluation harnesses, online feedback, promotion gates, and evidence-backed improvement loops.
→