AI
The Best Knowledge Agent Mostly Stays Quiet
I built an agent for a Teams help channel that answers from the wiki, fills documentation gaps with human approval, and otherwise says nothing.
Writing
Automation
AI automations need receipts when they produce nothing
AI automations need to distinguish an actionable result, an evidenced no-op and an unknown run state—and leave a receipt when they produce nothing.
AI
How to build a Codex memory system you can actually trust
Built-in Codex memory is useful for background context. For operational lessons, I want evidence, scope, curation, and a repeatable project setup.
AI
Agentic loops are coming for knowledge work
Agentic loops are useful when knowledge work has something to check against: sources, criteria, retries, and a clear stop point.
Anthropic
Claude Fable 5 Is Live: What the Mythos Connection Actually Means
Anthropic launched Claude Fable 5 today. It's the most powerful model they've made publicly available, and it carries a designation that's worth understanding before you just start using it.
Cowork
Codex Mobile, Dispatch, and the single-thread AI
Two AI products from two different companies found the same interaction pattern independently. That usually means it's right.
M365
Your Tenant has Three AI Labs in It Now. Here's What That Breaks.
Three frontier AI labs now have first-party integrations inside a standard Microsoft 365 tenant. The architectural implications for identity, data residency, and policy enforcement aren't being covered at the right depth.
Claude Cowork vs OpenAI Codex: the business case for each
I've been using Claude Cowork as a daily driver since the research preview. Scheduled tasks, skills, vault memory, and n8n integrations baked in. It replaced a lot of the glue I used to build by hand.
Enterprise
What happens when your AI agent fails silently
Three weeks ago a research pipeline I run finished overnight. Clean exit. No errors, no timeouts, nothing in the logs that looked wrong. The next morning I found three new entries in the records database with scores, bri
Copilot
Managing Copilot Cowork Skills: Design Patterns for Enterprise
The subfolder pattern works in Copilot Cowork, so the design question actually matters. How to structure SKILL.md as a dispatch router, where organisational content belongs, and why the growing community marketplaces are a risk, not a shortcut.
Microsoft
Copilot Cowork Custom Skills, One Week On: Microsoft Quietly Rewrote the Model
A week after my first Copilot Cowork custom skills post, Microsoft has quietly changed three of the behaviours I documented. The install flow, the Skills panel, and the skill format have all moved. None of it is in Microsoft Learn yet.
Microsoft
Copilot Cowork Custom Skills: Here's What I Found
I tested Copilot Cowork custom skills for a week. The docs make setup look simple, but there's undocumented behaviour, built-in skill conflicts, and a persistence gap vs Claude Cowork.
Claude Cowork is the prosumer's AI agent. Here's what it actually does.
I started using Claude Cowork about three weeks ago. Most of the coverage is by devs, written for other devs. I'm not one. I'm a non-developer who builds personal AI infrastructure (automations, skills, pipelines
Context Snapshots: How to Manage Claude's Context Window Across Long Sessions
Claude just got memory. Free for everyone, from March 2026. It remembers your name, your preferences, your writing style, and your recurring projects. If you've spent the last two years re-explaining yourself at the
I Built a Personal AI Stack Without Writing a Line of Code. Here's How.
I'm not a developer. I used to be a chef. About four years ago I made a fairly random career change into IT and ended up as a team leader doing Microsoft 365 architecture. I mention this because it matters for what f