Mark Levison

Stable teams sound like a nice-to-have until you look at the fields where an error kills someone. Most of the academic work deals with familiarity, not stability. Roughly: "we've worked together enough in the past to understand each other as individuals". Cardiac surgeons who worked across multiple hospitals but maintained stable teams had a lower rate of mortality, even with more surgeries performed. Aviation gives the same signal. 73 percent of the accidents for which information was available occurred on the first day the captain and first officer had flown together. A NASA study found that fatigued crews who had a history of working together made about half as many errors as crews of rested pilots who had not flown together before. Tired and familiar beat rested and unfamiliar. Back down to earth: at Wipro increased familiarity reduced defects by 19%, and Larry Maccherone showed stable teams had 40% better predictability. #TeamWork #Agile https://agilepainrelief.com/blog/in-agile-where-change-is-valued-why-is-a-stable-team-so-important/?utm_source=mastodon&utm_campaign=archive-reshare

@gedeonm@mastodon.social nice Canada thanks you.

Put GenAI usage into someone's performance review and you'll get exactly what you asked for. Some organizations now incorporate AI usage into performance reviews. What could go wrong? One team member generates multiple new features with 1000+ line pull requests, then drowns the rest of the team in review overload. Measuring individual productivity has always been a bad idea. Measuring it in an organization using AI is worse. If GenAI is an amplifier, it amplifies both the good and the bad. Measure velocity or individual productivity, and it gets amplified. Measure AI usage, and people will find a way to trigger it: you'll get a large bill from Anthropic or OpenAI to celebrate your "success," while drowning in AI-generated technical debt and security flaws. Two things from my Reading Rabbit Hole this issue: Bryan Finster argues AI is a "High Pass Filter" for software delivery, amplifying strong practices and exposing weak ones (https://bryanfinster.substack.com/p/ai-is-a-high-pass-filter-for-software) I think even the evidence for "amplifies strong practices" is thin, everyone claims it, nobody shows the data. And "Potemkin Understanding in Large Language Models" (https://arxiv.org/html/2506.21521v2) shows these tools don't understand what they're doing. They're very good mimics. Focus on the product and the team, not the individual. Don't measure AI usage or velocity. Not anti-AI. Studying where it breaks isn't anti-AI, it's the precondition for using it well. If your organization is setting GenAI adoption targets, what behaviour is it actually incentivizing? Every issue of my newsletter includes a Reading Rabbit Hole like this one. #AIinSoftwareDevelopment #Leadership https://agilepainrelief.com/newsletter/?utm_source=mastodon&utm_campaign=archive-reshare

@raganwald@social.bau-ha.us love it

@maxleibman@beige.party +1000. 10x more of the wrong code isn’t called winning.

@quinn@social.circl.lu HOA’s seem strange to me as a Canadian. Like shouldn’t exist strange.

@WTL@mastodon.social thanks for organizing. It was fun to see the fediverse in the real world

@briankrebs@infosec.exchange you have me beat. I’m only at 10%. Congrats on being plagiarized more often

@movq@tilde.zone i was trying light of the original post suggesting you should have used 8” inch place 5.25. Consider it a failure to transmit humour over the internet

@gedeonm@mastodon.social I checked with the Prime Minister and he has confirmed my thank you so it's official.

@zachleat@fediverse.zachleat.com We have not one but two RSS feeds: https://agilepainrelief.com/feed.xml and https://agilepainrelief.com/glossary/feed.xml

@jessie@mastodon.social who knew that there were Welsh speaking sharks.

@movq@tilde.zone aren’t those a bit small? Weren’t tall floppies a bit larger?