Food for Agile Thought #524: Irreplaceable Skills, Prototypes over PRD, Context Before Code, Deflation by AI

Max Berry: How To Not Be Replaced by AI

Max Berry argues AI is shrinking entry-level jobs as leaders automate file-based work. He uses Moravec’s and Polanyi’s paradoxes to show why physical presence and tacit judgment resist automation. His advice: become an accountable decision maker, learn to orchestrate and verify AI outputs, and show proof of work.

(via BBC Future): A 'toaster with a lens': The story behind the first handheld digital camera

Stephen Dowling recounts how Kodak engineer Steve Sasson built a handheld digital camera in 1975 from CCD parts and a cassette deck, stunning managers yet arriving too early for consumers, foreshadowing Kodak’s digital shift.

John Cutler: I see a lot of companies confused about product operating models and operating systems. What’s the difference? Why does it matter?

John Cutler explains why a product operating model is a useful abstract pattern for thinking, while an operating system is the real, designed machinery of how work runs. Do not confuse adopting “patterns” with building mechanisms.

(via First Round Capital): Context Before Code: How Notion Put an AI Engineer on the Sales Floor to Discover What Actually Needed Building

Theo Bleier embedded with Notion’s sales team for a month to understand real workflow pain before building anything. He first removed copy-paste friction with a lightweight tool, then focused on the deeper leverage point: account prioritization using product signals, supported by constrained research and draft messaging with humans in the loop.

Shreyas Doshi: The Product Builder's True Journey

Shreyas Doshi argues that product builders mature when they see that every feature creates attachment and constraints. Mantras from exceptional leaders mislead average teams. Outcomes require situational judgment across decisions, not slogans.

(via Every): I Talked to More Than 100 Companies About AI—Here's What's Actually Working

Natalia Quintero says AI adoption stalls due to a lack of clarity, not tools. Teams document workflows, start with champions, and build automations. Define success like training interns, then scale across teams.

Tomasz Tunguz: A Flash of Deflation

Tomasz Tunguz argues Gemini 3 Flash marks deflation: 70-79% cheaper than frontier peers, while within 9% of the best benchmark scores. He says this shift unlocks uneconomic AI product use cases.

Michael Wall: One Year with ChatGPT Pro as a First Hire

Michael Wall describes treating ChatGPT Pro as a first hire in his solo music licensing business, using chats, Deep Research, and Codex for coding, research, planning, and experiments, cutting costs and freeing time to compose.

Ianemmanuel Crueldad (via Medium): The Agentic AI Bubble: When Simple Automation Would Work Better

Ianemmanuel Crueldad argues that hype around agentic AI often confuses capability with necessity. Many workflows need predictable, deterministic automation, not autonomous decisions. Adding “intelligence” can increase complexity, reduce reliability, and create failure points.

The most valuable question we can ask isn’t “can we make this autonomous?” but rather “should we?”

Austin Tedesco (via Every): How AI Can Cut Your Planning Cycle From Two Weeks to Two Days

Austin Tedesco outlines how teams can use AI to compress quarterly and annual planning by combining a shared knowledge hub, meeting notes, and speech-to-text. He stresses clarifying questions, iterative drafting, and cross-team integration to surface conflicts early.

Esther Derby: Speaking Up

Ester Derby argues hierarchy makes people stay silent until problems explode, hurting decisions and engagement. Leaders can change this by normalizing problems, exploring concerns and hunches without interrogation, and actively amplifying others’ voices.

Simon Powers: Fields and Awareness: Why change only happens when we work with what’s already in the room

Simon Powers argues that change fails when the field stays unchanged: tone, power dynamics, and unspoken rules in the room. Build awareness, name what is present, and shift context before frameworks.

(via InfoQ): 📺 Empowering Teams: Decentralizing Architectural Decision-Making

Peter Hunter and Elena Stojmilova explain how Open GI moved from a legacy monolith to a cloud SaaS architecture by decentralizing architecture through Team Topologies, a DDD context map, an advice process with principles, ADRs, and an open weekly advisory forum.

(via Figma): Prototypes Are the New PRDs

Emma Webster argues that product managers gain clarity more quickly through prototyping than through PRDs. High-fidelity, interactive prototypes help explore options, validate with users, align stakeholders, and clarify behavior details through concrete artifacts rather than static documents.

Chris Matts: DORA metrics considered harmful.

Chris Matts argues that DORA metrics help teams improve delivery, but become harmful when used by executives. Outside the team, they incentivize production over customer value, enable blame, and distort behavior in cross-team environments.

Lizzie Matusov: How do daily stand-ups boost team performance?

Lizzie Matusov reviews two studies showing that daily stand-ups increase psychological safety, which in turn boosts satisfaction, learning, and performance. Stand-ups help when leaders model vulnerability and focus on blockers.