The interface pattern is, in my opinion, the single most important revolution that software has gone through to date. While interfaces are technical by nature, they shape companies and even the world we live in.

In my word, an interface is a piece of a system, software or not, responsible of connecting system together, enabling them to connect, exchange data and services.

The interface revolution began to grow with an internal memo by Jeff Bezos made to all Amazon development team in 2002.

In 2002, Amazon was still a young company, but it was already a big player on the web. Amazon was scaling, in all direction and this growth lead to technicals and managerial challenges. Jeff Bezos solved this scale issue in one very famous memo.

The famous memo was sent in 2002 and stated :

This led Amazon to shift from a monolithic system (or at least a distributed monolith) to a well-designed Service Oriented Architecture (SOA). Under this new architecture, applications could no longer communicate freely gone were the days of unrestricted interprocess communication, direct linking, and shared memory. While these old communication methods may seem messy by today's standards, they were common practice at the time. Since software rarely needed to be distributed or replicated, direct data sharing was a viable approach.

With this mandate, Jeff Bezos compelled Amazon to think of their services as individual Lego blocks that could be used by any team within Amazon or even any team in the world.

This initiative led to the development of one of the most complex software systems in the world: AWS. AWS's success largely stems from its simple and reliable interfaces.

Interfaces should be treated like contracts: while having a bug in software is not a major issue, breaching a contract will have serious consequences.

When interacting with a production-grade API like Amazon S3 (one of the first systems built by AWS over 20 years ago), you expect the service to respond correctly, securely, and promptly. This confidence comes not only from Amazon's strong track record but also from the formal contract that you sign into when using their APIs : the SLA.

Interfaces can only work if you establish a trust relationship with them. Without trust, you cannot build upon them. Just like a house with unstable foundations will collapse, an unreliable interface will fail. That's why each interface should be treated like a contract.

The brilliance of Bezos's memo extends beyond its technical aspects. It also laid the groundwork for designing a new organizational structure.

Having strong interfaces within an organization (both engineering and non-engineering) enables the creation of small, autonomous teams. These teams maintain complete responsibility for their managed software, with all communication between teams' systems occurring through APIs.

Teams communicate their data and expertise using a common language through clear interfaces. For example, team A can consume team B's services autonomously—team B doesn't even need to know team A is using their service. This is exactly how AWS operates: countless teams use S3, but the S3 team doesn't interact with each consumer.

This model of independent teams consuming each other's APIs has become a renowned organizational approach. Spotify refined this model to perfection, and their organizational structure has become one of the most widely adopted templates for engineering teams.

The interface revolution is not only technical. Interface reduces big system complexity, abstracting their internal working and providing a simple way of using them.

Nowadays, complexity is everywhere, so interface are there as well. They have become common, take for example cars, 50 years ago the interface to the engine was much more complexe than today : engine was manual, you needed to select which gear you want to get in, in order to match the wanted speed of the driver. Today, most of the cars are built with automatic gear transmission, meaning that you dont need to think about which gear the car is in. As a result the old interface was the gas pedal and the gear, the output was speed, now you only have to think about the pedal.

Another great example of interface simplification is the Global Positioning System, also known as GPS. Before GPS, navigation required extensive knowledge of maps, street signs, landmarks, and cardinal directions. Users needed to plan their routes in advance, keep track of turning points, and constantly monitor their surroundings to stay on course. Even experienced navigators could get lost in unfamiliar cities or during poor weather conditions.

With GPS, this complexity has been hidden behind a simple interface. Users simply input their destination, and the GPS handles all the complex calculations—determining current position via satellite triangulation, finding optimal routes through real-time traffic data, and providing turn-by-turn directions. The system even automatically recalculates routes when users make wrong turns or encounter road closures. This transformation has made navigation accessible to everyone, regardless of their geographical knowledge or orientation skills.

Deleting friction is what most big business do today. Most of the GAFAM are based over this principal (not really, but their core concept is). When a new interface is created, and if this interface solves a real problem, user will adopt it.

• Google aimed to create a simpler interface between users and the world's internet knowledge
• Amazon created a new way of buying goods, instead of having to go to a physical store you use only the brand new interface : the marketplace website
• Uber created a new interface between the user and the taxi through their app.

Simplifying someone's life by addressing a real pain point is the foundation of product-market fit. These companies created excellent interfaces for tasks that were previously tedious.

Logistics giants like DHL and FedEx can be seen as simple interface. In fact they all have REST API you can interact with (Amazon must be interacting with them at some point). They are complexe system which act as interfaces between businesses and global shipping networks. You dont want to recreate a shipping network, you would prefere to use one already in place.

In the software world, companies like Stripe and Twilio have built billion-dollar businesses by providing specialized interfaces. Stripe simplifies connections to payment systems, while Twilio bridges applications with communication networks. Rather than creating entirely new services, these companies take complex systems and transform them into clean, standardized interfaces that businesses can easily integrate and depend on.

A new category of internet services has emerged, with players like n8n, Zapier, and IFTTT whose sole purpose is to act as universal interfaces that can connect different interfaces together.

Creating new interface will be a big part of the next decades.

AI and LLM have already revolutionised the interface world. They introduced a new communication method with machine, with natural language. The big problem now in LLM AI is to find the way to interface them with all of the other interface already available. This problem might be solved shortly using the AI Agent, which are quite popular nowdays. Once this problem is solved, a once size fit all interface will have been created. This will be the mother of all interface, the one that can interface anything with everything.

Complexe software have a lot of interface chained together. For example when you make a request to a webservice, a big number of interface are triggered : screen API, the application, the network API, the kernel, the communication network, the data center, the server (and all of its interfaces…). This big chain is very complexe, quite reliable but not perfect. Adding interfaces over interface mean that we must rely a lot on the small interface that are wildly use.

When one of the small interface fell, it the all chain that is broken.