Every graph engineering post includes a checker node. Every one of them draws it too small. The usual version sits before your synthesis step and catches empty outputs, contradictions, off topic results and broken formatting. Those are real failure modes. A model can spot all of them. The failures that stall enterprise adoption are not like that. Was this agent allowed to touch this repo. Whose permissions did it run under. Who finds out what it did. Open the flow and the answers are on screen before anything runs. The trigger step says when the agent may act. The Use Claude step says which repository it may touch and what to do there. The connection notice says whose permissions it borrows. The Slack step says who finds out. Four steps, four boundaries, all visible while you build it rather than reconstructed from logs afterwards. That is not quality control. It is a decision boundary. A good checker node is not a model grading a model. It is a step that knows what it may touch, whose authority it borrows, and who it has to tell.

Everyone is drawing dependency graphs in a YAML file this week. The graph that decides whether your team ships has been running in your tracker for years. The explainers get the basics right. A node is one unit of work. An edge is a real dependency. Most of the edges you drew are fake, because those two steps only run in order since that is how you typed them. Then they stop. The graph sits in a file, runs on your laptop, and disappears when you close the terminal. At Atlassian we shipped the other half this week. Jira Automation can now call Copilot, Cursor or Claude Code as an action step inside any rule. The pieces are the same. What holds them is different. A node is a work item, not a prompt. An edge is a trigger or a condition, not a depends_on line. The graph outlives the session, because the tracker is the runtime. Speed is not the point here. A node can now be a person, and the rule still runs. Where does your dependency graph live today. A file, or a system?

Focus - The best time to stop a distraction is before it starts. The second best time is now. - boz.com/articles/focus

How You Communicate in product open.substack.com/pub/yuezha…

The Importance of Writing in the Age of AI The Tim Ferriss Show - #695: Shane Parrish on Wisdom from Warren Buffett, Rules for Better Thinking, How to Reduce Blind Spots, The Dangers of Mental Models, and More. Via AI-powered podcast app @snipd_app. share.snipd.com/snip/b794bbf…

The Power of Automatic Rules Snipped from The Tim Ferriss Show - #695: Shane Parrish on Wisdom from Warren Buffett, Rules for Better Thinking, How to Reduce Blind Spots, The Dangers of Mental Models, and More. Via AI-powered podcast app @snipd_app. share.snipd.com/snip/86fc0b6…

In 2023, I used @loom to avoid 811 meetings — making me a Loom All Star 🎊

New update for Product Managers. You can customise what the 'Create' button does in Jira Product Discovery community.atlassian.com/t5/J…

Excited to share I've joined @Atlassian as a Principal Product Manager for @Jira! Loved the welcome pack💙 #newhire #atlassian #Jira #Agile

Most of us will have around 40,000 weeks in our lifetime. What and how to think about making the best of time leebyron.com/4000/

Proud of my product side project 😊. It's been a big year with thousands of puzzles played around the world on #iOS & #Android Try the app sudoku.social/

An interesting line-up of speakers and masterclasses at Something Digital #SD22. From the future of web, pitching, getting to Net Zero and the conundrum of increasing space debris in orbit.

My new article on how to give teams time to learn then deliver (at big tech & #startups) is available now 👀 linkedin.com/pulse/how-give-… #Management #Microsoft #teamwork

4 Lessons I Learned from My Personal #MBA. Practical insights to hire better, protect time, communicate more and have more valuable connections. #altmba #hiring #management linkedin.com/posts/domwillia…