» When designing an experiment to test a hypothesis, ask how might it fail, and what you need to find out

» Luck and uncertainty mean that even good decisions can have a bad outcome (and vice versa)

» Framing our beliefs with a confidence percentage makes us more willing to accept contradictory evidence

» Anyone can access generative AI, so simply using it is not a competitive advantage

» SEO has long sought to game search engine rankings; AI provides a new tool for doing so

» LLMs trained on AI output become increasingly detached from reality: “model collapse”

When sunsetting a product or feature:

» Explain why honestly and be transparent about how it will impact your users

» Be understanding as they may be disappointed or frustrated by the change

» Be responsive and supportive by answering your users’ questions and helping them with the transition

» If possible, offer your users alternative features or solutions

» Context helps to make a service good as opposed to simply existing

» By not consciously designing our services, we instead force our users to link actions together

» Outcomes for users are core user needs that a service helps them to meet

» A starting point can be to ask different teams what ‘good’ looks like to them

» Empowered teams can only succeed if the leadership team is on board

» Leaders can’t scale or have all the answers; empowered teams stand a better chance

» Leadership needs to be open and transparent when issuing a “must-do” edict to an empowered team

» Product leadership is continually striving for coherence of approach and clarity of purpose