TL; DR: Quo Vadis, AI? — Food for Agile Thought #522
Welcome to the 522nd edition of the Food for Agile Thought newsletter, shared with 35,812 peers. This week, Gary Marcus questions the viability of large language models, citing unresolved flaws such as hallucinations and weak reasoning — Quo Vadis AI? Teresa Torres and Petra Wille reflect on AI’s role in user interviews, warning that tech cannot replace skilled human insight, while Peter Yang sees the PM role in AI-native firms shifting toward builders who align quickly. Also, OpenRouter’s massive token data analysis reveals rising agentic use and open models, and Willem-Jan Ageling dissects why many corporate transformations still stall.
Next, Dave Baines urges product managers to stop hiding technical work and instead align it through transparency and joint prioritization. Chris Jones and Marty Cagan remind stakeholders that their job is not to demand features but to enable outcome-driven problem solving. Saffron Huang’s team finds Claude changes engineering workflows at Anthropic and raises career questions. Also, Teresa Torres shares 50 AI use cases from her own process, and Eric Barker recommends strategy over emotion when dealing with difficult coworkers.
Then, Jing Hu questions the true ROI of AI investments, citing McKinsey data showing limited gains and weak agent adoption. Mike Brock calls AGI from LLMs a flawed idea dressed as progress. Mike Fisher warns that chasing velocity without slack leads to chaos, Martin Eriksson urges teams to simplify decisions through focus and consent, and Andi Roberts challenges Tuckman’s model. Finally, John Cutler explains why tools fail when they don’t tackle underlying behavioral constraints.