← Blog
03 — Topic
Artificial Intelligence
17 posts on this topic.
- 3 min
AI Transformation Fails at the Foundation, Not the Model
AI transformation fails at the foundation, not the model: the data it reads, the contract for what it may do and the human who signs off. Three questions.AI · Transformation - 3 min
When a Config Change Is Really an Acquisition
A model router is one line of config and a new external data processor. Treat it as an acquisition, test the premise, and prefer a local fallback.AI · Governance - 3 min
Two AI Sessions, One Repo: What Broke
AI sessions sharing one git working copy collide in five ways, from vanished drafts to a stuck lock file. Here is the rule behind each.AI · Git - 3 min
Absent, Not Disabled: Giving a Small Model Read Tools
A small model can read a live CMDB safely when write tools are absent, not disabled, and every call is scoped, capped, cited and audited.AI · ServiceNow - 3 min
You Don't Need a Bigger Model. You Need Better Retrieval.
Fix retrieval before buying a bigger model. A small CPU-only model went from 200-second waits and off-topic citations to useful answers without growing.AI · RAG - 3 min
Five ServiceNow API Calls That Return Success and Store Nothing
Five ServiceNow Table API calls report success yet store nothing: currency fields, ampersands in references, bad choice values, class filters, partial access.ServiceNow · CMDB - 4 min
A Dry Run Is Not Permission
A clean dry run proves nothing about write access. Safe AI-assisted production writes need a preview, stated intent, a window, a canary and a read-back.ServiceNow · AI - 2 min
Introducing solvent-brain: A Starter for an AI-Maintained Architecture Knowledge Base
solvent-brain is a free, MIT-licensed starter for an AI-maintained architecture knowledge base: an operating contract, resume state, rules and skills.Enterprise Architecture · AI - 2 min
Pushed Is Not Deployed
A push is not a deploy. Ten releases were logged as deployed while the server ran an old version; three rules keep the two apart.DevOps · AI - 3 min
Every App Needs Two Capabilities
Map every app to the Business Capabilities it serves and exactly one technical capability for what it is. That second lens makes rationalization work.Enterprise Architecture · Capability Modeling - 5 min
Two Tiers: Spend the Expensive Model Only on Judgment
Use a strong model only for judgment and schema, and a cheaper model for bounded, repeatable skills. One rule decides which tier gets each task.AI · Claude - 6 min
Sessions Forget: Designing Resume State for AI Agents
AI sessions lose whatever lives only in context. A resume page, a ruled-out list and fixed boot and close rituals keep long-running work intact.AI · Claude - 5 min
Never Let the Generator Grade Its Own Work
AI work becomes trustworthy when a separate reviewer assumes it is wrong, runs the checks and never fixes its own findings.AI · Claude - 5 min
Green Tests, Broken System: 4 Ways a Test Suite Lies
A passing test suite can still hide defects: fixtures built on assumptions, rules that never fire, checks that stop looking and gates that defend the bug.Testing · AI - 6 min
Building an AI Second Brain for Enterprise Architecture
An AI-maintained EA knowledge base works with three layers, fixed resume rituals, a skeptical reviewer and one rule: cite, don't restate.AI · Enterprise Architecture - 3 min
Your Agent Doesn't Run Your Hooks
Guardrails in one AI tool's settings bind only that tool. Rules that must hold belong in git: pre-commit hooks, CI and tests every writer has to pass.AI · AI Agents - 3 min
Face, Voice and a Pass-Phrase: Lessons From a 2018 Hackathon
One short video, three checks: face, voice and the spoken pass-phrase. What a first-place 2018 Microsoft/TAG hackathon taught me about layered authentication.Identity · Biometrics