Emergent Organization

Software Teaches Us How to Design Organizations

Conway’s Law reveals a fascinating truth about software companies: their products inevitably mirror their organizational structure. In the New Hackers Dictionary, Eric Raymond summarized this law as “If you have four groups working on a compiler, you’ll get a 4-pass compiler.”

But this relationship goes both ways: new software design patterns inspire how we design software teams. For example, microservice architectures have led to the rise of small “two-pizza” teams — small, autonomous groups that can be fed with just two pizzas.

This reciprocal influence between software design and organizational structure isn’t just a coincidence. Both organizations and software systems grapple with similar challenges, like coordinating multiple components to produce desired outcomes, improving throughput, eliminating bottlenecks, optimizing efficiency, and reducing errors.

When breakthrough insights emerge in software design, organizations naturally take notice and adapt. Just as software systems must evolve to meet new challenges, organizations must also remain flexible, learning from the most successful patterns in complex systems.

There’s a bigger question in all this — what software patterns might reshape the companies of tomorrow? As we journey into the age of AI and distributed systems, the answers to this question could fundamentally transform how we think about organizational structure. 

Organizational Structure and Software Architecture Have Developed in Parallel

The intertwined evolution of software design and organizational design provides deep insights into the structure of both code and companies. 

For example, in a pre-cloud world, software was designed with grand architecture and deployed ready-made. Every decision regarding scale and communication was rigid and pre-established. This type of software was developed by organizations with a prescriptive “waterfall” organizational structure with strict hierarchy and division of power.

As software became cloud-based, asynchronous, and service-oriented, organizations changed: teams became more independent, flexible, and agile. Remote, asynchronous work became possible. As software has become dominated by real-time metrics and dashboards, organizations have adopted company dashboards, OKRs, and real-time feedback systems.

This parallel evolution extends beyond just technology. In their attempt to constantly improve, organizations will look to all complex systems for inspiration. In the book “The New Spirit of Capitalism,” Botanski and Chiapello say that organizations have always adopted lessons not just from technology but also from the brain.

Today, computer science and neuroscience are intertwined. How can recent advances in both fields help us rethink human organizations?
Understanding this parallel evolution isn’t just an academic exercise. We can use it to predict where organizations might head next. As software evolves toward more distributed and self-organizing systems, we might expect organizations to follow a similar path. 

The Bitter Lesson

The most significant shift in computing today results from the success of deep neural networks. At the heart of this success is a crucial yet counter-intuitive software design strategy called “The Bitter Lesson.”

Required reading for every aspiring AI researcher, “The Bitter Lesson” is an article written by computer scientist Rich Sutton. Sutton explains that, over time, AI researchers keep making the same mistake of over-architecting systems prescriptively to solve specific problems. He says researchers keep finding that “building in how we think we think does not work in the long run.” The big successes we see in AI come from resisting the temptation to tell the computer what to do and designing systems that have the structure, data, and scale to learn on their own.

Instead of “programming” how systems like large language models should solve complex problems, we give large networks the scale, power, and data to figure out how to solve problems themselves while aligning their incentives to encourage them to self-organize in patterns that may seem illogical and incomprehensible to the original designer.

To build successful AI, we have to stop “designing” software. More importantly, we must teach ourselves to stop thinking we can design software.

What if a Company Looked Like a Brain?

If we are fundamentally rethinking how software is designed, should we not also encourage a parallel rethinking of software organizations? There’s an intriguing paradox: organizations like OpenAI have achieved breakthrough success by rejecting traditional software design principles. However, they organize themselves using conventional corporate structures that would look familiar a decade ago.

What could a radical rethinking of organizational structure inspired by modern AI look like?

Highly performant neural networks have defining characteristics that may one day be applied to organizations:

• Self-organizing “plastic” structures rather than prescriptive organizational design
  • Instead of rigid org charts designed from the top down, picture structures that form and reform based on emerging needs.
• Reinforcement learning from clear feedback
  • Neural networks learn through unambiguous feedback about their performance. What if organizations could create similar feedback mechanisms, allowing teams to adapt and learn in real time?
• Overlapping networks versus isolated departments
  • Imagine overlapping spheres of collaboration in which information and expertise flow freely and operational knowledge exists in distributed but interconnected patterns. 
• Lack of a cohesive “center”
  • Picture a company in which organizational wisdom emerges from the interactions of its components rather than concentrating decision-making in a central executive function.  
• Emergent structures that form and dissolve based on needs
  • Organizational structures continuously flex and optimize themselves to address the challenges rather than remain fixed. 

This vision isn’t entirely unprecedented. Take pre-2014 GitHub, for example. It attempted to implement aspects of an organizational model inspired by the very open-source tools it created, designing itself as a flexible, bossless organization. 

At the time, it was a proud example of an “unstructured organization.” At early GitHub, there were no formal titles and top-down defined projects. As early employee Zach Holman described in 2012: “[t]eams are all informal. If you want to work on something, just go work on it. If you need a designer, grab a designer. Boom, a two-person team.”

However, the organization drastically changed in 2014, giving rise to what is sometimes called GitHub 2.0. At that time, GitHub formally announced that it was abandoning its radically flexible organizational culture and establishing VPs, functional leads, and official performance reviews.

The reasons why GitHub had to undertake such a formal change would take pages to discuss and can be approached from many different viewpoints. Indeed, the need to scale a company under the demands of newly raised venture capital contributed to the change (GitHub raised $100M from Andreessen Horowitz in 2012).

We can learn two important lessons from the rise and fall of GitHub’s boss-less organizational structure:

• Moving or removing structure and replacing it with informality is insufficient to transform an organization.
• It is difficult to change companies’ natures if they operate within a larger global environment that rewards hierarchy and structure.

Open Source as the Future?

Does GitHub’s story mean that all non-hierarchical organizations are destined to fail? Are today’s prescriptive structures combined with management metrics the only way to build a scalable organization?

Consider that GitHub began as a closed-source interface wrapped around git, an open-source project created by Linux founder Linus Torvalds. GitHub’s retreat from its experimental culture might be better understood as an inevitable tension between its venture-funded, proprietary nature and its role in the open-source ecosystem.

One may find examples of successful, informally structured software organizations in the true open-source community. Take separate projects, Linux and PostgreSQL, as examples. Linux powers over 90% of the world’s top web servers, and PostgreSQL is one of the most popular databases. Both projects exemplify brain-like organizational structures: thousands of contributors coordinate without rigid hierarchies, forming fluid teams based on interest and necessity rather than top-down directives.

Perhaps the organizational model we’ve been seeking — one that mirrors the human brain’s networked, adaptive nature — has been hiding in plain sight.

The missing piece may not be a new organizational framework but rather a deeper investment in sustainable paths to open-source development. By directing more resources toward open-source projects and tackling increasingly complex challenges, we might finally unlock the full potential of brain-like organizational structures.

The Courage to Be Crazy

Twenty years ago, the idea of neural networks surpassing traditional programming seemed implausible to most experts. Today, as we witness the unprecedented success of AI systems that learn and self-organize, we must ask ourselves: will there be a similar inflection point with organizational design? 

While GitHub’s experiment with unstructured organization may have ended in a return to hierarchy, the thriving open-source community shows us that brain-like organizational structures can work at scale.

What if the primary barrier to brain-like organizations isn’t that they don’t work but that we haven’t invested enough in making them work? The breakthrough moment for neural networks didn’t come from a fundamental change in their architecture — it came from massive increases in scale, data, and computing power.

The future of work might not look like the rigid hierarchies we’re familiar with but rather like the adaptive, self-organizing networks transforming the software world. We just need to be audacious enough to embrace that change.

Further Reading:

• “What should we do with our brain?” Catherine Malabou, 2008
• “The Bitter Lesson” Rich Sutton, 2019
• “GitHub: exploring the space between boss-less and hierarchical forms of organizing” Journal of Organizing Design, Burton et. al. 2017