83% of Agile practitioners use AI, but most spend 10% or less of their time with it because they do not know where it fits. Our survey of 289 Agile practitioners identifies the real adoption barriers and shows where AI creates value you can act on. Learn more by downloading the free AI4Agile Practitioners Report 2026.
Welcome to the 530th edition of the Food for Agile Thought newsletter, shared with 35,729 peers. This week, Dwarkesh Patel and John Collison press Elon Musk on orbital data centers, power limits, space solar for cheaper AI within three years, while Itamar Gilad urges experiments without hype, since discovery, constraints, maintenance, and outcomes still rule. Roman Pichler outlines a product operating model with long-lived products and empowered teams. Also, Deb Liu shares charts on adoption, costs, and uneven gains, and Tom Geraghty ties status to busyness rather than impact.
Next, Stephanie Leue shows how a “clear” strategy still burns teams out when priorities multiply, and she urges explicit trade-offs, surfaced hidden work, and ranked outcomes with visible de-scoping. Aakash Gupta describes Mike Bal’s AI native PM system using Cursor or Claude Desktop, MCP tools, and vetted research, and Arvind Narayanan challenges Moravec’s Paradox and calls for a diffusion-minded policy. Additionally, Benedict Brady automates AI-generated feedback into PRs, and Teresa Torres and Petra Wille reframe hiring as discovery.
Then, Scott Alexander reports on Moltbook’s first weekend, asking whether AI posts cause real effects, then maps influencers, spam, crypto manipulation, micro religions, builders, and fragile self-moderation. Greg Satell debunks change myths and urges committed minorities plus resistance planning, while Maarten Dalmijn warns that post-failure rules kill competence and trust. Also, Jayshree Seth and Amy C. Edmondson frame AI adoption as team learning with reviews and overrides. Lastly, Victor Yocco refocuses UX on trust, consent, and accountability.
AI initiatives fail for the same reasons Agile transformations did: The majority of failures result from people, culture, and processes, not technology. This article gives you a diagnostic checklist of 10 AI transformation anti-patterns to spot where your organization’s initiatives are coming off track.
Welcome to the 529th edition of the Food for Agile Thought newsletter, shared with 35,753 peers. This week, Jing Hu and Klaas Ardinois unpack OpenClaw/Clawdbot and the real tradeoffs and risks of always-on self-hosted agents, while Stephanie Leue shows how AI exposes broken product operating models and why builder teams beat bolt-on AI. Maarten Dalmijn reframes roadmaps as a choice between Red predictability and Blue adaptability, and Dario Amodei sketches near-term AI risks and safeguards. John Cutler calls out metrics theater and pushes outcome signals.
Next, Ant Murphy suggests product-tech teams can drop roles like BAs and Scrum Masters by pulling engineers into discovery, reducing dependencies, and shipping small batches with decoupled deploy and release. Wes Bush frames product-led growth as table stakes for AI software, with fast time-to-value, agents as users, and per-task pricing. Zvi Mowshowitz reviews Claude’s Constitution and its values-first stance, and Ethan Mollick treats management as the most critical AI skill. Also, Casey Newton repeats a crucial truth: AI creates work slop, so measure outcomes.
Then, Federico Viticci shows OpenClaw/Clawdbot, an LLM-based agent on a Mac mini that chats via Telegram, stores Markdown memory, adds MCP skills, and runs shell tasks, while raising app store policy questions. Mike Fisher links speed to focus, trust, and psychological safety, not pressure; Sean Goedecke treats estimates as political and replaces dates with options and risks, and Aakash Gupta, interviewing Sachin Rekhi, pushes AI prototyping to validate problem solution pairs fast. Lastly, Kieran Klaassen suggests that AI coding fails when planning disappears.
The paradigm shift is here. Andrej Karpathy, former Tesla AI director and OpenAI co-founder, recently admitted he has never felt this far behind as a programmer. If Karpathy feels overwhelmed, how should the rest of us feel?
This article maps the shift across three levels: strategic, product, and individual. Each level demands different responses, while “good enough Agile” no longer provides an income or perspective. The question is where you are on the journey.
Welcome to the 528th edition of the Food for Agile Thought newsletter, shared with 35,758 peers. This week, John Cutler presents a five-act narrative exploring how organizations evolve from delivery-focused work toward product-centricity through a messy, iterative value journey, while Stephanie Leue identifies an “alignment tax” that slows growing organizations and requires systemic redesign. Turning to AI, Christina Wodtke explores how Claude Code enables “(product) discovery coding” through conversation. Addy Osmani offers guidance on writing AI agent specs, and Allan Kelly warns that AI coding will unleash shadow IT alongside security risks.
Next, Rich Mironov suggests executives ask whether products earn their keep by trusting engineering judgment over ticket-level ROI math. Bandan Singh proposes filters before reacting to competitors. Also, Lenny Rachitsky interviews Zevi Arnovitz on how nontechnical PMs ship products using Cursor, while Jing Hu warns that AI assistants may weaken desire by removing anticipation-building friction. Additionally, Mike Cohn proposes that soft skills persist and compound, unlike technical skills, which have a shrinking half-life.
Then, Teresa Torres demystifies how large language models work, explaining tokenization, embeddings, and attention mechanisms. David Burkus suggests effective delegation requires giving ownership rather than tasks, while Anthropic releases Claude’s new constitution, prioritizing safety, ethics, and helpfulness. Lastly, PwC’s 29th Global CEO Survey reveals 56% of CEOs report no AI financial return yet, and Simon P. Couch estimates Claude Code sessions consume 138 times more energy than typical queries, calling for transparency from frontier labs.
