Lenny's PodcastMax Schoening: Why agency beats skills as AI flattens craft
Through Notion's prototyping playground, designers ship code in the terminal; the first 10% of every project is now free, exposing who has agency.
CHAPTERS
- 1:15 – 3:10
Why roles are collapsing: AI changes how product teams build
Lenny sets up the core theme: AI is collapsing the traditional Venn diagram of designer/PM/engineer roles, and Max has effectively lived this future for years. Max frames the conversation around how AI is reshaping workflows, prototyping, and what teams will value next.
- •AI accelerates building and blurs role boundaries on product teams
- •Max’s background spans PM, design leadership, and engineering
- •The episode focus: what changes in the next months/years as AI integrates into work
- 3:10 – 6:22
Origin story: why Notion designers started prototyping in code (not Figma)
Max explains how Notion’s early AI/chat interface work exposed the limits of static design mockups. The team built an LLM-friendly “playground” codebase so designers could prototype interactive AI behavior directly, making iteration faster and more realistic.
- •Static Figma mocks are a ‘dead fish’ for designing AI interactions
- •A small, LLM-friendly playground lowered fear of the terminal
- •Designers moved chat-interface prototyping from Figma into code
- •The playground served as a ramp toward contributing to production code
- 6:22 – 8:25
How much designers/PMs ship now—and why “vibe coding” isn’t the goal
Max describes a spectrum: designers and PMs can handle small UI tweaks and prototypes, and some contribute to production. But he’s skeptical that AI tools have automatically improved software quality, and he emphasizes designing in the medium of the final product over shipping code for its own sake.
- •Designers at Notion increasingly prototype and sometimes ship code
- •Workflow friction: teams sometimes reverse-engineer code prototypes back into Figma
- •Small changes (styling tweaks) are easy; reliability and quality remain hard
- •The real value: designing in the same medium engineering will ship
- 8:25 – 10:33
Shipping vs strategy: why learning the “material” beats landing PRs
Lenny raises the concern that faster engineering might squeeze PM/design time for strategy and cohesion. Max argues he doesn’t care whether non-engineers ship to production; he cares that they understand the substrate—especially agent loops—because that’s where product differentiation is moving.
- •Coding is useful primarily as a way to understand the medium
- •Understanding agent loops matters more than tweaking UI styles
- •Tooling is evolving into “operating systems” for agentic work (coding harnesses)
- •Prototyping in code helps teams design what will actually be built
- 10:33 – 11:50
The AI-era differentiator: cultivating agency (not job titles or “skills”)
Max names agency as the trait separating people who thrive from those who struggle as AI makes more skills instantly accessible. He predicts rigid attachment to role definitions (“what it means to be a PM/designer/engineer”) will become a liability.
- •AI reduces skill barriers; agency becomes the main bottleneck
- •Agency is unevenly distributed and historically mapped to “founder energy”
- •Seeing the world as malleable becomes a career advantage
- •Role identity rigidity will slow people down
- 11:50 – 13:53
High-agency examples at Notion: recruiting, redefining roles, doing the work
Max shares concrete examples of agency from Notion teammates who operate beyond their formal job descriptions. The through-line is taking responsibility for outcomes—whether that means recruiting, prototyping, or reshaping how one contributes.
- •Brian Lovin as an example: blurs design/engineering and drives recruiting needs
- •Eric Liu’s shift: from strategy docs → Figma → building prototypes directly
- •“Drive Notion like it’s stolen” as a mindset for non-founders
- •Agency shows up as changing the system, not just executing tasks
- 13:53 – 15:57
What we risk losing as roles merge: specialization, craft, and real engineering
Max cautions that role collapse can erase specialists at the edges—both in engineering rigor and design craft. He uses a hardware metaphor: prototypes are easy, but scaling to reliable mass production is the real engineering challenge, which software discourse often underweights.
- •Risk: losing specialists as everyone becomes a generalist
- •Hardware analogy: prototype vs manufacturing-scale reliability
- •Critique: token/feature obsession distracts from scale engineering
- •Design craft and “delight” can get diluted by off-the-shelf patterns
- 15:57 – 17:44
How to develop agency: tinkering, making, and realizing the world is changeable
Max’s advice for building agency is surprisingly practical: make things. Tinkering creates a feedback loop that proves you can change your environment, and it shifts agency from a corporate “workaround” mindset to a creative, human one.
- •Agency grows through making—tools, meals, small projects, experiments
- •Steve Jobs idea: the world is built by people no smarter than you
- •Avoid framing agency as merely outmaneuvering organizational constraints
- •Making things builds confidence that change is possible
- 17:44 – 21:15
Malleable software: reclaiming ownership over your computing life
Max defines malleable software as software that serves users’ interests and adapts to them, rather than locking them into the vendor’s worldview. AI makes personal tool-building easier, but Max argues real malleability requires platforms that support shared, communal workflows—not just one-off scripts.
- •Apps often bundle UI + data ownership into rigid “little squares”
- •Extreme alternative (full DIY/Linux) is powerful but too costly for most people
- •AI unlocks personal tool-building, but platform design determines scalability
- •Ink & Switch/Jeffrey Litt influence: collaboration + malleability without losing modern UX/security
- 21:15 – 23:56
Design philosophy lens: Dieter Rams, usefulness first, and “buildings that learn”
A pinned Dieter Rams clip becomes a jumping-off point for Max’s design beliefs: usefulness precedes beauty. They connect this to long-term adaptability—homes (and software) that evolve over time often outperform pristine top-down designs.
- •Rams’ critique: museum-worthy objects can fail at real usability
- •Usefulness as a north star; beauty as a secondary outcome
- •Stuart Brand connection: environments improve by adapting over time
- •Malleability helps reveal whether something is truly useful
- 23:56 – 28:26
The “SaaS apocalypse” debate: why services and maintenance still matter
Max argues the SaaS apocalypse is exaggerated: people don’t want to maintain full stacks indefinitely. AI will likely push software back toward more general, flexible tools (90s-style), while “as-a-service” persists because upkeep, expertise, and scale reliability remain valuable.
- •Many 2010s SaaS products were “guided spreadsheets” with less malleability
- •Core SaaS value: maintenance, specialists, and tending the “software garden”
- •Prediction: more general-purpose tools resurface, still delivered as services
- •Notion AI helps people finally learn/operate malleable tools via built-in tutoring
- 28:26 – 30:28
How product building changed: the first 10% is free, iterations are cheap
Max explains the biggest shift: AI makes the earliest phase of projects nearly effortless, enabling many parallel explorations and faster “something to react to.” PRDs matter less when you can demo quickly, but the last-mile quality work remains brutally hard.
- •AI makes the first version/prototype dramatically cheaper and faster
- •“Demos not memos”: show the thing instead of describing it
- •Teams can explore multiple directions with agents in parallel
- •The last 10% (polish, reliability, adoption) still dominates effort
- 30:28 – 34:17
What’s next: modality shifts, inference speed, and “good enough” intelligence
Max is unsure whether the future is constant chat, direct manipulation, or something hybrid—much depends on speed and cost. He also challenges the assumption that everyone always wants the smartest model; many workflows will prefer “good enough” models that are faster, cheaper, and local.
- •Open question: does direct manipulation return via near-instant inference?
- •Speed changes workflows: queued jobs vs continuous interactive “clay molding”
- •Skepticism that frontier intelligence is always the winning product strategy
- •Focus on exoskeleton-like augmentation vs “god in a box” centralization
- 34:17 – 48:39
Token spend → ROI: adoption incentives now, cost scrutiny later
They discuss how Notion currently treats token spend as a low-priority constraint to maximize exploration. Max expects ROI conversations to become unavoidable soon, with outcomes shaped by whether frontier labs maintain a large gap over open-weight models and whether businesses demand flexibility and vendor choice.
- •Unlimited/loose token policies can accelerate learning and behavior change
- •Token spend is a vanity metric (like lines of code) if not tied to outcomes
- •Leaderboards can help force behavior change in large orgs (Meta example)
- •ROI future depends on lab gap vs open models, and on avoiding provider lock-in
- 48:39 – 51:34
Notion AI’s advantage: context-rich workspace + permission-aware enterprise search
Max credits Notion’s long investment in AI and the connected-workspace model: agents perform better when they can operate within rich, permissioned context. He frames Notion as closer to an operating system, enabling agents to “roam” across company knowledge in a way siloed apps can’t.
- •Notion AI predates ChatGPT-era hype; early intuition from leadership
- •Agents need context; connected workspaces reduce “narrow orifice” integration pain
- •Permission handling and enterprise search are differentiators
- •Notion as an OS metaphor aligns with how coding agents thrive in Unix-like environments
- 51:34 – 56:41
Shipping faster without lowering the bar: “shots on goal” + “obviously good” quality
Max explains the cultural levers for speed: reduce preciousness, increase attempts, and rely on iteration—while maintaining a high standard for what’s worth shipping. He emphasizes consolidation work (simplifying too-many primitives) as the hidden cost of moving fast.
- •Culture tends to get precious after strong product-market fit—must fight it
- •“Shots on goal” increases experimentation and learning velocity
- •Quality bar: “only make obviously good stuff,” validated by iteration
- •Speed creates complexity debt—must reconcile multiple primitives into a simple core
- 56:41 – 1:05:07
Taste, great products, and the tiny core: iteration builds intuition
Max defines taste as an internal simulator that predicts what a specific “in-group” will value, built through repetitions and feedback. He argues successful products typically hinge on a tiny superpower—an exceptionally strong core—while feature accumulation rarely creates greatness.
- •Taste = a mental “virtual machine” trained via reps and feedback
- •Designers build taste via side projects and constant exposure to new tools
- •Great products have a tiny core superpower (PRs, git push deploy, blocks/slash commands)
- •Beware the death spiral: “one more feature will make it great”
- 1:05:07 – 1:07:29
User-centered thinking: jobs-to-be-done, honest communication, and avoiding “marketing brain”
Max uses jobs-to-be-done less as dogma and more as a forcing function: what is the user truly hiring the product for? He also highlights how teams often lose clarity when they switch into landing-page or internal-company mode, drifting away from how they’d explain the product to a friend.
- •JTBD as a reminder to zoom out and adopt the user’s perspective
- •Distinguish what users want vs what the company wants users to want
- •Landing pages often trigger vague, clever-sounding marketing language
- •Communication test: explain it at a whiteboard, then compare to your actual messaging
- 1:07:29 – 1:27:22
Hot takes and closing corners: UBI, AGI plans, exclusivity, failures, and advice for young builders
Max delivers a set of philosophical takes: “knowledge work is already UBI,” AGI wouldn’t change what he wants to do (he’d still tinker), and small groups drive outsized impact. He closes with failures (polishing the wrong product core) and advice to avoid frenzy, cultivate agency, and focus on what you genuinely care about.
- •UBI hot take: knowledge work already resembles a privileged baseline income
- •With AGI, Max would keep building—reduce meetings, increase tinkering
- •Contrarian view: exclusivity can be good; small groups run the world
- •Failure lessons: over-polishing the wrong thing; feature creep when the core is weak
- •Advice: work hard, zoom out, read history, and don’t let fear drive career choices
