FIELD NOTES / HELIX

Ideas for software that can explain itself.

Essays on engineering graphs, trustworthy AI agents, software memory, and the systems we are building at Helix.

FEATURED
Company 3 min read

Software should explain itself

Why the next generation of engineering systems must preserve the reasoning, evidence, and relationships behind every change.

Engineering IntelligenceCompany
Read the essay
LATEST WRITING

Notes from the work.

Clear thinking about the context, evidence, and relationships modern engineering teams need.

01
Engineering Intelligence 4 min read

OpenClaw 2.0 said the gateway is the trust domain. The graph is the security docs.

We indexed the latest 44 merged PRs on openclaw/openclaw with Helix. The change center isn’t the sandbox — it’s gateway security docs and the install-policy scan.

Engineering IntelligenceOpenClawAgent SecurityOwnership
Continue reading 02
A luminous engineering graph connecting code files, pull requests, decisions, and owners around a central software system.
Engineering Graphs 4 min read

How to build an Engineering Graph from GitHub data

A practical architecture for turning repositories, pull requests, commits, reviews, owners, and incidents into a graph that explains how software actually works.

Engineering GraphsGitHubCode IntelligenceKnowledge Graphs
Continue reading 03
A changed software node sends highlighted paths through services, tests, owners, and production systems to reveal blast radius.
Change Intelligence 4 min read

Software change blast radius: a graph-first guide

Why file search misses the real impact of a code change—and how relationship paths reveal affected services, owners, tests, and production risks.

Change ImpactEngineering GraphsPull RequestsSoftware Architecture
Continue reading 04
Software clusters connect to luminous expertise nodes, with one isolated area revealing concentrated knowledge risk.
Engineering Leadership 3 min read

Code ownership is a graph, not a CODEOWNERS file

A practical model for combining declared ownership, review history, expertise, incidents, and team boundaries without confusing activity with authority.

Code OwnershipKnowledge RiskEngineering GraphsDeveloper Experience
Continue reading 05
AI Assurance 2 min read

Evidence, not another AI review

Autonomous software work needs an independent trust layer that verifies both how an agent behaved and what its code actually achieved.

AI AgentsEvidence Ledger
Continue reading 06
A code change passes through separate execution and outcome verification rings before becoming a chain of verified evidence.
AI Assurance 3 min read

Evidence-backed AI code review: what a trustworthy report needs

A concrete framework for AI code review that separates facts, inferences, requirements, tests, agent behavior, and human judgment.

AI Code ReviewEvidence LedgerPull RequestsSoftware Quality
Continue reading 07
Abstract coding agents pass through permission, sandbox, recording, and verification layers before reaching a protected repository.
Agentic Engineering 3 min read

The missing trust layer for autonomous coding agents

Coding agents are becoming production actors. Here is the architecture teams need to govern execution, verify outcomes, and preserve durable evidence.

Autonomous AgentsAgent SecurityEvidence LedgerDeveloper Tools
Continue reading 08
Engineering Graphs 3 min read

Engineering graphs for agentic software

How graph-shaped context gives coding agents better retrieval, safer plans, and feedback loops that improve with every engineering event.

Engineering GraphsAgentic Systems
Continue reading 09
A timeline of incidents, decisions, commits, and outcomes flows into a persistent luminous software memory graph.
Engineering Leadership 3 min read

Software memory: how to preserve why engineering decisions were made

A durable model for connecting architecture decisions, pull requests, incidents, code, and outcomes so teams stop repeating the same archaeology.

Software MemoryArchitecture DecisionsEngineering GraphsKnowledge Management
Continue reading 10
A question pulse travels through a selected subgraph of code, services, people, and tests before becoming a grounded answer.
Engineering Graphs 3 min read

GraphRAG for codebases: how to answer engineering questions with relationships

A practical retrieval architecture that combines knowledge-graph traversal, code evidence, and language models for better codebase answers.

GraphRAGKnowledge GraphsLLMCodebase Understanding
Continue reading 11
A pull request connects to dependencies, code, tests, owners, and risk evidence around a transparent central signal.
Change Intelligence 3 min read

Pull request risk scoring without the black box

How to build an explainable PR risk model from change surface, dependency impact, test evidence, ownership, incidents, and agent execution.

Pull Request RiskChange ImpactEvidenceEngineering Intelligence
Continue reading 12
Pull requests, incidents, deployments, agents, and tests orbit a central engineering graph in a continuous learning loop.
Engineering Graphs 3 min read

The continuous Engineering Graph: how software understanding compounds

A closed-loop architecture that learns from pull requests, incidents, deployments, agent runs, corrections, and production outcomes.

Continuous IntelligenceEngineering GraphsFeedback LoopsDevOps
Continue reading 13
Repositories, teams, policies, tests, sandboxes, and agent nodes align into a clear autonomous engineering operating system.
Agentic Engineering 4 min read

How to prepare your engineering organization for autonomous coding agents

A practical readiness guide for repository context, permissions, sandboxes, verification, ownership, metrics, and rollout boundaries.

Autonomous AgentsEngineering LeadershipAI GovernanceDeveloper Productivity
Continue reading
KEEP EXPLORING

Build with context.
Ship with confidence.

See how Helix turns code, decisions, incidents, and ownership into a living Engineering Graph.

Explore the Engineering Graph