Joined September 2011
- Tweets 38,084
- Following 678
- Followers 5,999
- Likes 42,606
3,209 Photos and videos
Pinned Tweet
My latest book is live and available now. This one is all about navigating the harsh realities of enterprise software projects, which claim to be "agile," but not really.
2
2
384
This experiment completed successfully. Qwen 3.5:9b loaded inside @claudeai read my entire manuscript, using skills I had provided, and gave me a thorough analysis of the book.
It took 11 hours 18 minutes to finish, but it worked. 😄
Currently letting Claude Code do an editorial review on a novel I've been working on. Everything is running locally, through Ollama and Qwen 3.5:9b on a 16GB M1 MacBook Air. I expect it'll take hours, but I'm cool with that.
2
289
Long-range estimates are often partly theater, partly a date already set, partly a real planning need. Refuse fake precision, not the conversation.
74
Is there any way to turn off ESC to interrupt in @claudeai ? I can't tell you how many times I have accidentally interrupted a long running process because I thought my focus was another window. I've started minimizing the Claude window now, but then I can't watch its progress.
114
Currently letting Claude Code do an editorial review on a novel I've been working on. Everything is running locally, through Ollama and Qwen 3.5:9b on a 16GB M1 MacBook Air. I expect it'll take hours, but I'm cool with that.
463
Developers speak complexity and risk. Leadership speaks timelines and budgets. Same work, two languages. It's up to the tech lead to manage that translation.
103
When asked for a date, try: what we know, what we do not, a range if nothing changes, what we will flag early if it shifts. Early uncertainty beats surprise.
39
What is your response if you are asked to estimate eight sprints before stories even exist?
3
138
This might be my new favorite bad review of one of my books.
1
108
It’s been running for about ten minutes now. What’s the likelihood of it finishing? M1 MacBook Air with 16GB RAM. I also have a small fan blowing across the laptop vents.
128
"When will it be done?" is a fair question that needs an answer. Budgets and releases need dates. Honest and precise rarely coexist under uncertainty. Managing up is translation.
57
When ceremonies survive budget cuts and test automation does not, you get long releases with a manual QA pile at the end. The math is predictable.
90
AI for learning a new framework: net positive or net negative? I am net positive only if I explain the generated code out loud afterward. Which side are you on?
109
Half the team on incidents for two days? Say it in review before someone asks why velocity dropped. Context reduces anxious polling.
1
85
I'm embracing conventional commits. Commit messages are status for everyone else: "auth: handle expired refresh tokens; fixes PROD-4421" beats "fix login bug" when someone traces the sprint at 3pm.
1
1
95
I do not have imposter syndrome anymore. I have did I actually read that diff syndrome.
2
1
99
Someone with Jira read access asks why nothing is closed mid-sprint. What is your calm, factual go-to?
186
My AI renamed everything for clarity. The clarity is debatable. The merge conflict is not.
1
100
New habit before pushing AI-generated code: asking whether there is a simpler way to accomplish the same thing or whether we touched more code than necessary.
151
The bug found two weeks after you mentally closed a feature is usually feedback-loop length. By then context is expensive. That cost was predictable.
1
1
121
Copilot-era habits still shape how people think: autocomplete in a box. Modern tools can run multi-step work across a repo. Update the mental model. When did yours last change?
89