Building error handling into your Microsoft Flows
How to stop treating workflow failure as an exceptional surprise and start making routing decisions based on success, failure, skips and timeouts.
IDEAS CAPTURED AT THE TIME
Most of my writing starts with a practical problem, a new technology that looks promising, or a question that refuses to leave me alone.
How to stop treating workflow failure as an exceptional surprise and start making routing decisions based on success, failure, skips and timeouts.
A practical user-onboarding problem involving Azure B2C, temporary passwords and two different implementation approaches depending on how much expression pain you fancied.
An intentionally ugly expression for turning a human date string into UTC, published as a challenge to make it shorter and cleaner. The opening premise was basically: this works, but surely we can do better.
Scheduled automation, parallel branches, SharePoint data, tokenised HTML, MailChimp API calls and the sort of integration glue that turns “low-code” into actual application plumbing.
One of those requirements that sounds trivial until the edge cases arrive: active and expired metadata, editability, search refiners and a Power Apps form writing valid SharePoint structures.
The scale of the technology changed. My preference for finding the awkward edge cases did not.
Not a “VR is the future” victory lap. The article asks what immersive enterprise interfaces are actually good for, where they are awkward and whether large datasets might be more interesting than giant virtual Word documents.
Introducing the Semantic Publishing Protocol: structured, agent-readable publishing with intent, attribution, context and consent rather than assuming tomorrow’s software will consume the web like yesterday’s browser.
A model for what happens when agents consume the web without traffic, attribution or incentive: publishers lose value, publishing contracts, and the information ecosystem AI depends on slowly gets poorer.
A longer, deliberately speculative piece about language, consciousness, substrate and the assumptions people smuggle into claims that machines can never understand.
There are older community posts, practical notes and a few stranger pieces. The Medium profile is the least lossy surviving index.