Tell me what you’re trying to build, fix, automate, secure, or launch. I’ll point you to the useful path—and if a real conversation would help, you can reserve it here.
FIELD NOTES / VERIFIED 2026-07-30
What works. What breaks.
Deep implementation analysis for teams making production decisions: current capabilities, practical architecture, real limits, evidence, security, operation, and the points where human judgment must remain.
EIGHT LAUNCH PILLARS
The production decision library.
These are not model-release summaries. Each article gives the reader a decision frame, architecture, failure modes, evidence standard, implementation checklist, and a restrained delivery path.
0114 min
Agentic systems
Agentic AI in production: a 2026 guide to bounded, observable systems
A practical guide to deciding where agents belong, designing their authority, evaluating complete workflows, and operating them safely in production.
- CTOs and engineering leaders
- Product and AI leaders
- Operations leaders
AI evaluation
The AI agent evaluation playbook: from demonstrations to release evidence
How to build task suites, trace graders, human calibration, adversarial cases, and release gates that measure an agent as an operating system.
- AI and ML engineering teams
- Product and quality leaders
- Security and risk teams
AI operations and security
AI production readiness: security, observability, and operational control
A production-readiness model for action-taking AI systems, covering trust boundaries, least agency, traces, sensitive data, incident response, and rollback.
- CISOs and security teams
- Platform and SRE teams
- CTOs and engineering leaders
AI governance and compliance
AI compliance readiness in 2026: current EU dates and an evidence-first operating model
A current, practical guide to the EU AI Act timeline, system classification, NIST AI RMF, and the operational evidence teams need to maintain.
- Executives and boards
- Legal, privacy, compliance, and risk teams
- CIOs and engineering leaders
Software engineering
AI-assisted software development: an operating model beyond code generation
What current DORA and METR evidence actually says, where AI-assisted development fails, and how to redesign the engineering system around evidence.
- CTOs and VPs of Engineering
- Engineering managers and platform teams
- Software developers
Multimodal AI
A production multimodal content pipeline for GPT Image 2, Nano Banana 2, and Veo 3.1 Lite
How to route image and video work across current models, preserve generated assets, validate brand and policy, and keep humans responsible for publication.
- Marketing and creative operations leaders
- Product and engineering teams
- Brand and communications teams
AI strategy
Build, buy, automate, or stop: an AI decision framework
A disciplined method for comparing policy change, conventional software, automation, vendor AI, custom AI, agents, and deferral before spending heavily.
- Founders and executives
- CIOs and CTOs
- Product and operations leaders
MCP and tool security
MCP security for production integration: identity, tools, approvals, and evidence
A production security model for connecting agents to MCP servers and other tool providers without turning model context into unchecked authority.
- Platform and AI engineering teams
- Security architects
- CTOs and technical leaders
Apply the field notes
The value is in the operating system.
Bring a use case, architecture, prototype, control gap, or model decision. We can turn the general guidance into a representative evaluation, bounded proof, implementation, or remediation plan.
Please do not send secrets, credentials, regulated data, or confidential customer material through an initial inquiry.