Agent loops, graphs, and factories: what each one controls
Choose the structure by the decisions you need to control.
Writing
AI, agents, and enterprise software.
This is my independent, unsponsored blog. I share personal views and opinions informed by my professional experience. They do not represent my current or former employers, who are not involved in this site.
Pacing arrivals gives a busy gateway room to recover. The API is overloaded. What should your client do?
Decide when to retry, spread the attempts out, and put a limit on the extra work.
How to tell whether an agent skill helps
Test selection, task results, and the cost of getting there.
Deploying Salesforce metadata with GitHub Actions
Keep the reviewed revision, target org, and deployment evidence together.
Structure skills so agents find the right instruction
Keep the entry point small, move detailed material into focused resources, and make every path and command real.
Write a skill an agent can find and follow
Give a repeated task a clear trigger, a short workflow, an output contract, and a boundary.
What building real docs with Blume took
Exploring Blume by rebuilding a real Salesforce open-source package
Recursive Language Models for messy business investigations
A mobile-first explainer of the Recursive Language Models paper through the lens of a Salesforce consultant investigation.
Claude Code in Action — A Practitioner's Take
I took Anthropic's official Claude Code course. Here's what changed how I work, where I disagree with the course, and the gaps you'll fill on your own.
Track Code Changes with Ticket IDs in Git Commit Messages
Use a simple Git commit-msg hook to enforce ticket IDs like PROJ-1234 in every commit. Stop maintaining file header changelogs and build traceable Git history.