The Malleable Future of Software: Agency, Craft, and the AI Revolution with Max Schoening

The landscape of software development is undergoing a seismic shift, driven by the rapid advancement of AI. This transformation is not just about new tools but a fundamental redefinition of roles, workflows, and what it means to build and use software. Max Schoening, Head of Product at Notion and a prominent AI-forward product leader, offers profound insights into this evolving world, emphasizing the critical role of "agency" and the emergence of "malleable software."

The Origin Story of Designers Coding at Notion

The idea of roles blurring, with designers and product managers increasingly contributing to code, is not new. Max Schoening has been a proponent of this for a long time, having led design and product at GitHub where designers were significant contributors. The shift at Notion, where designers and PMs are now shipping code, has a specific origin story.

"When I joined Notion, we were building a lot of chat interfaces," Max explains. "We were designing them in Figma, but the static image of a chat is essentially a dead fish. You have to feel the AI to some degree." To bridge this gap, Max and two designers created a "small codebase that is very LLM friendly," a playground optimized for AI tools. This allowed them to move prototyping for chat interfaces into a more dynamic, code-based environment.

This "playground concept" was designed to be less intimidating than the main codebase, which had accumulated a decade of patterns. By making it easy to get started and overcome the "fear of the terminal," they enabled designers and PMs to experiment. As AI models have improved, this has evolved, with designers and PMs now contributing to the production codebase, albeit to a lesser degree. The trend is clear: as model capabilities grow, the amount of work individuals can accomplish will increase exponentially.

How Much Designers and PMs Are Shipping Today

The notion that designers and PMs should be shipping code is gaining traction. Max believes it's "very, very useful for designers to move from manipulating Figma documents into code." He recalls a time at GitHub when designers were among the top contributors.

However, he cautions against a sole focus on shipping to production. "I do have a general sort of maybe issue with vibe coding," he states, referring to the rapid, sometimes superficial, use of AI for coding. "I don't feel like the quality of software has increased all that much in the last 12 months. I think the maybe the amount of software has, but it's very, very hard to find software that is is reliable."

Instead, Max emphasizes the importance of designing in the medium that will ultimately become the real product. For him, the value lies not in whether a designer's code lands in production, but in how thinking in code forces them to "consider the medium." This deep understanding of how agent loops work, for example, is more valuable than someone who can only tweak traditional software styles. The only way to truly grasp these complex interactions is by building them with the material they are made of: code.

The Balance Between Shipping Code and Strategic Work

A common concern is that as engineers become more productive with AI, designers and PMs might be squeezed, struggling to keep up with the rapid pace of development. This raises the question of whether their time is better spent on strategic direction rather than coding.

Max's perspective is nuanced: "I actually don't care at all whether designers write code that lands in production." His focus is on the cognitive benefits of coding. "The reason I like thinking in code is because it forces you to consider the medium." He prioritizes designers and PMs who deeply understand how agent loops work over those who can only make minor UI adjustments. This understanding, he argues, is best achieved by building within the material itself – code.

This approach is more about the prototyping use cases than simply shipping more features. While it often leads to designers and PMs blurring the lines and writing production code, the primary goal is to become a "master of the material," not just a cog in the delivery mechanism.

Why Agency Will Help You Thrive in the AI Era

In this rapidly changing landscape, Max identifies "agency" as the key differentiator for success. "I think before it was very easy to always say, 'Well, I will never be able to do this because insert skill issue.'" Now, even with skills at one's fingertips, thanks to AI, what truly matters is agency.

Agency, Max explains, is the understanding that "the world around them is malleable." Those who embrace this will thrive, while those who cling to rigid definitions of their roles – "What does it mean to be a PM? What does it mean to be a designer?" – will find it increasingly difficult.

Examples of High Agency at Notion

Notion employees, or "Ninos," often exhibit above-average agency. Max highlights Brian Leven, who not only blurs engineering and design but also acts as a top recruiter, proactively identifying organizational needs and finding talent. This demonstrates a desire to "affect change, I don't care how it happens."

Another example is Eric Lou, who transitioned from writing strategy documents to actively building prototypes. His motivation was to develop skills that would make him indispensable, even in the early stages of a startup. This proactive approach to shaping one's role and impact is the essence of high agency.

What We Might Lose as Roles Merge

As roles merge and software becomes more malleable, there's a risk of losing specialists. Max uses a physical metaphor: building a hardware startup involves 3D printing for prototypes, which clearly shows imperfections, and then a long road to mass manufacturing. In software, this "engineering part" – ensuring a product works for millions – is often overlooked in favor of rapid feature iteration.

Similarly, while design systems enable quick UI creation, the "delight in craft" can be lost. Max emphasizes the need to ensure that in this merging of roles, we don't lose the specialists who bring deep expertise and a focus on quality and nuance.

Advice for Developing Agency

To cultivate agency, Max draws inspiration from Steve Jobs' quote: "One day you wake up and you realize the world is made up by people no smarter than you." The key, he believes, is "making." By tinkering and creating things, individuals embark on a cycle of creation and learning. This process awakens the understanding that change is possible.

"It's almost like training a model," Max says, comparing it to developing taste. The more you "do reps," the better you become at predicting user reactions. This iterative process, combined with a willingness to experiment and learn, is fundamental to building agency.

Malleable Software Explained

Malleable software, in Max's view, is software that "works closer to the interest of the people that use it than the interest of the corporation that makes it." He contrasts this with the current app-centric world, where layers are rigidly glued together, limiting user customization. While the alternative of running a full Linux distribution offers malleability, it's often impractical.

The ideal is a platform or operating system that encourages this malleability, allowing users to have ownership over their computing lives. Notion, he suggests, is moving in this direction, acting more like an operating system where users can build their own tools and workflows. This resonates with the idea that "you can just change things."

The Dieter Rams Video and Design Philosophy

Max's pinned Twitter video of Dieter Rams critiquing furniture design highlights a core philosophy: "design should be first useful and then beautiful." Rams' critique of overly artistic but impractical pieces underscores the importance of functionality. Max connects this to malleable software, suggesting that a good way to assess usefulness is by a product's ability to be changed and tweaked. This echoes the idea that homes that adapt over time to their inhabitants' lives are often the best, even if they require costly modifications.

The SaaS Apocalypse Debate

The idea that SaaS tools will become obsolete, replaced by custom-built solutions, is a topic Max addresses with a nuanced hot take. He argues that much of 2010s SaaS was a "fancy form around a spreadsheet," offering guidance but less malleability. The "as a service" component, however, remains crucial. Most people don't want to maintain the entire software stack.

Instead, Max predicts that tools will become more general, akin to the word processors and spreadsheets of the '90s, but delivered as a service. The true value of SaaS lies in the maintenance, the ongoing effort of specialists solving problems. He believes the "SaaS apocalypse" is exaggerated, and while things will change, the need for specialized tools and services will persist.

How Product Building Has Changed in the Past Two Years

Max describes the current state of product building as "the first 10% of every project are now free." The effort required to build the initial version of a startup or explore new ideas has drastically reduced. This allows for more experimentation and exploration of different paths.

The shift from "memos to demos" is now amplified. Instead of writing lengthy product requirement documents, teams can quickly create functional prototypes. This "builds in iteration into the product much earlier." The challenge now lies in the last 10% of development, which still represents 90% of the effort.

What’s Next in How We Build Products

Max is conflicted about the future of product building. He believes in the durability of "plain text" and code but questions whether we'll solely rely on chatting with AI. The future of tools like Figma is uncertain, as AI agents might take over direct manipulation.

He also points to the "automation versus augmentation fork." The speed of AI inference is a critical factor. If inference is nearly instant, will we revert to direct manipulation, molding code like clay? Or will multitasking remain the norm? Max suggests that for many cognitive tasks, there might be a point of "good enough" intelligence, after which faster, cheaper, and more localized models will be preferred over the absolute frontier.

Token Spend and ROI Conversations

The conversation around AI token spend is evolving. While initially, companies encouraged exploration with unlimited budgets, Max anticipates a shift towards ROI conversations within six to twelve months. He believes that the delta between proprietary and open-weight models will influence this. If the gap widens, proprietary models might dictate the future. However, if the gap remains narrow, we'll see a diffusion of models, with companies becoming comfortable running their own, smaller, cheaper models for specific tasks.

Max cautions against using token spend as a boastful metric, comparing it to bragging about lines of code written. The focus should be on the value generated, not the consumption of resources.

Getting People to Change How They Work

Changing ingrained behaviors is challenging. Max notes that roles further from engineering often embrace AI more readily due to the immediate "superpowers" it provides. Engineers, on the other hand, may need to be guided to see manual interventions as a sign of something potentially wrong in their "software factory."

The key to driving change, he suggests, is to make the new way of working so compelling that it feels like a bug to revert to the old. For engineers, this means ensuring that every human intervention feels like a deviation from an optimized, agent-driven process.

Max’s AI Stack

While Figma's role might be shifting, Max doesn't see it as necessarily trending down. He acknowledges the rise of "vibe coding" but emphasizes the surprising utility of the terminal. He encourages PMs to use terminals over GUIs, believing it will lead to a deeper understanding of computing substrates. Tools like Conductor, which resemble developer tools, are also gaining traction.

Which Roles AI Will Transform Next

Max's hot take is that while models are rapidly improving at coding, progress in other domains like writing has been less impressive. He believes that as the cost of creating software approaches zero, "software engineering will go into all the other domains." This means that rather than AI automating HR tasks by writing code, HR professionals will use AI to encode business practices directly into software. The acceleration of "software eating the world" is the primary transformation.

When Companies Will Start Caring About ROI

Max predicts that within six months, companies will begin scrutinizing AI costs. He hopes for a diffusion of models rather than a widening gap between proprietary and open-weight options, to avoid centralization of power. This diffusion will likely lead to ROI calculations based on whether it's cheaper to use smaller, specialized models. He also sees parallels with the "cloud wars," where layers commoditize and businesses seek choice, avoiding lock-in with single providers.

Why Notion AI is So Successful

Notion AI's success, Max believes, stems from its ability to leverage context. Agents need access to information, and Notion's connected workspace provides this naturally. He likens Notion to an operating system, creating an environment similar to Unix for coding agents. This, combined with a willingness to tackle hard problems like enterprise search with automatic permission handling, contributes to its effectiveness.

How to Ship More Quickly While Maintaining Quality

Max emphasizes "shots on goal" – increasing the number of experiments and iterations. With AI making experimentation easier, this strategy is more viable than ever. However, he stresses that feature count is a "silly metric." The focus should be on fewer, exceptionally good features that offer combinatorial power.

The challenge of software quality remains. Max advocates for "incremental correctness," iterating rapidly while striving for "obviously good" products. This means increasing shots on goal but also having the discipline to consolidate and refine core ideas, even if it means delaying the next release.

Building Taste Through Iterations

Taste, Max explains, is the ability to "run a virtual machine in your head where given an idea, you can predict for a certain in-group whether they're going to like it or not." Developing taste requires "reps" – iterative practice with feedback. This is akin to training a model.

He notes that designers with high taste often have side projects and are constantly exploring new tools, exposing themselves to diverse ideas and surrounding themselves with tasteful things that inspire them to improve.

What Matters Most in Building Successful Products

The most critical factor in building successful products, Max asserts, is the team. He also highlights the importance of a "tiny core that is so exceptionally good." This core superpower, whether it's multi-touch on the iPhone, the pull request on GitHub, or blocks and slash commands in Notion, is what makes a product truly great. He cautions against the pitfall of believing that adding "just one more thing" will make a product finally great.

Using the Jobs-to-Be-Done Framework

Max uses the Jobs-to-Be-Done framework as a reminder to "think holistically about what the user wants to hire your product for." It encourages product teams to step out of their internal perspective and consider the user's actual needs, rather than what the company wants the user to want. This "zoom out" perspective helps avoid getting lost in the details of creation and ensures the product truly serves its intended purpose.

Hot Take on Universal Basic Income

Max's hot take on Universal Basic Income (UBI) is that "we already have universal basic income. It's called knowledge work." He suggests that the needs for contentment are less than we often assume, and the hierarchy of jobs we've built is not strictly necessary. He believes humans are inventive and will always find ways to insert themselves into the loop, even with advanced AI.

What Max Would Do with AGI

If AGI were to arrive and work became optional, Max would continue doing what he does now: tinkering, building, and trying to make the world more malleable. He codes not just for utility but as an intellectual challenge, akin to playing chess or Go. He would explore new areas, like the intersection of AI and robotics, driven by curiosity and a desire to create.

Contrarian Corner: The Power of Exclusivity

Max's contrarian view is that "inclusivity isn't always all that great." He believes the world is often run by small groups and that sometimes being exclusive, focusing on a specific, high-value segment of users, can lead to building truly exceptional products. This doesn't apply to essential services but to areas where catering to a niche can foster deep engagement and innovation.

Failure Corner: Working Diligently on the Wrong Thing

A significant failure for Max was starting a competitor to Notion in 2014. They spent years polishing the editing experience, implementing features that are now common in tools like Obsidian. However, they underestimated Notion's core innovation: the block-based system, which was initially a "terrible" editor. This taught him the hard lesson of "working diligently on the wrong thing for way too long" and the importance of identifying the true core value proposition.

Advice for Young People in Silicon Valley

Max advises young people in Silicon Valley to "just don't let the rush or the frenzy sort of distract you from the things that you actually care about and are passionate in life." He encourages hard work, especially in the early years, but warns against the anxiety of missing out or being left behind. He suggests zooming out, reading history, and trusting that focusing on genuine interests will lead to a fulfilling path.

Key Takeaways