By Teresa Torres · producttalk.org · @ttorres on X · LinkedIn
Teresa Torres reflects on 2025 and lays out her 2026 roadmap, framed around the idea of 'just now possible' enabled by generative AI. She describes building AI coaches and tools, sunsetting her cohort-based Deep Dive courses in favor of on-demand courses and a Discovery Habits Toolbox for organizations. She explains a dual strategy of AI teaching tools that build discovery skills and AI automation tools that do discovery work, while acknowledging the trade-off of outsourcing learning. She also discusses experimenting with a media business model via subscriptions and sponsored ads, and closes by stressing the need to retire declining offerings to make room for new ones. The piece matters as a candid case study of how an expert educator is adapting to AI disruption.
01Key takeaways
- Generative AI can deliver personalized, detailed feedback at scale, enabling on-demand deliberate practice that was previously only possible in expert-led cohorts.
- Training alone rarely changes how an organization works; leaders must participate, reinforce the new habits, and coach their teams.
- Use AI for some discovery tasks and do some yourself, since the act of doing the work carries learning value that outsourcing removes.
- Competing with free AI alternatives requires making your offering clearly better and lower-effort for the customer to justify its cost.
- Regularly review your portfolio and retire products that no longer grow, even when it's emotionally hard, to create room for new bets.
- Treat the business as a set of experiments, diversifying revenue sources rather than depending on one model.
02Key sections
- Reflecting on a year of surprises
- The author recaps personal and business highlights from 2025, including new podcasts, first AI products, and an injury-driven shift toward building with AI.
- Vision for AI that augments people
- Torres states a hopeful vision where AI helps teams connect with customers and amplify more voices, while openly naming risks like job loss and replacing human contact.
- Moving from cohorts to on-demand learning
- Deep Dive cohort courses are being split into smaller on-demand courses, with AI coaches providing personalized feedback alongside Slack, office hours, and challenges.
- Augmenting versus automating discovery
- The author splits effort between teaching tools that build discovery habits and automation tools that perform discovery tasks, arguing that doing the work yourself retains real value.
- Business experiments: subscriptions and ads
- Facing pressure from cheap AI alternatives, she tests paid subscriptions and sponsored ads, promising disclosure and editorial independence while admitting discomfort with advertising.
- Sunsetting to make room for what's next
- She frames retiring declining products as a necessary, if painful, part of the lifecycle that frees resources for new experiments.
03From the post
“2025 was full of many surprises. Some good ones—I launched two new podcasts, I built my first AI products, I completely changed the way I work thanks to Claude Code. And some tough ones—I'm still struggling to fully recover from a severe ankle break in March, I”
04Frameworks mentioned
Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.