The 6 Levels of CodeGen Automation

The lines between human skill and artificial intelligence are getting thinner, entering us into a new age of development tools. 

Generative AI Code generation (codegen) stands out, offering varying degrees of autonomy that range from basic assistance to full-scale automation. This article seeks to define and explore these levels of autonomy, offering a lens for tech visionaries to better understand the future of coding. 

Using the SAE Levels of Driving Automation™ as both inspiration, and as a guide, we’ve developed a framework to determine the level of autonomy for codegen tools that are on the market. We’re calling this framework, simply, Stride’s Levels of Code Automation (though maybe we’ll create a society of software developers to oversee this one day!).

Read on for a detailed view into what each level entails along with examples for each.

Level 0


Level 0 environments are hard to come by these days, but it’s worth using this as a benchmark to measure just how far we’ve come in such a short period of time. In Level 0 environments, you are coding and you must constantly supervise what is being created. 

As a frame of reference; level 0 cars have super basic functionality like blind spot and lane departure warnings or automatic emergency braking.

All coding tasks, including writing, testing, and debugging, are performed manually by developers at Level 0. At this level, the developer is writing code entirely manually using basic text editors without any assistance from the tool beyond syntax highlighting.

AWS CodeWhisperer Screenshot
dropdown arrow

Level 1


Transitioning to Level 1 introduces a collaborative partner into the coding process—the basic coding assistant. While you are still coding, you have syntax highlighting, code suggestions, and simple error detection at your fingertips.

Level 1 automation assists with minor tasks but requires significant manual input for code creation and problem-solving.

To make this more real, Level 1 cars have lane keep assist or adaptive cruise control, not both. Only having one of these features makes it so that you still have to do the majority of the work. It’s a little luxury and shares similar features with Level 2.

This level of assistance marks the beginning of a partnership between human intellect and machine precision, simplifying repetitive tasks and allowing developers to channel their efforts into more complex and creative challenges. It's a subtle yet significant step towards greater productivity and higher code quality.

dropdown arrow

Level 2


Level 2 automation offers a more robust form of collaboration: semi-Autonomous code generation. This is far and beyond growing to be the most popular, and available, level of autonomy offered to the market today. 

Keeping with the car analogies, Level 2 cars have both lane keep assist and adaptive cruise control. The pairing of these technologies is like peanut butter and jelly - these can take the dread out of long drives and morning commutes. It’s almost magic and once you start using these tools it’s hard to imagine living without them. This is where things start feeling luxurious.

Level 2 allows the automation of more intricate tasks, such as code formatting, refactoring, and basic testing, comes into play. These tools are not mere extensions but collaborators that ensure consistency and enhance error reduction. Developers still wield the primary creative force, crafting the bulk of the code. However, they now have the accelerative power of AI - safeguarding their time from the mundane. 

This balance between human ingenuity and automated accuracy is pivotal—it ensures the craft remains in the hands of the creator while imbuing the creation process with efficiency.

dropdown arrow

Level 3


Level 3 Autonomy is on the other side of the great “Autonomy'' divide - in Level 3 you are not coding. This is where the paradigm shifts from assistance to autonomy. The tools at this level are not just participants in the coding process; they become stewards of it. 

Beep beep, car comparison incoming! Level 3 was only recently achieved in automotive, as of November 2023 by Mercedes-Benz. Imagine, within the limit of certain speeds and parameters, your car is able to drive and maneuver through traffic on its own. That’s level 3.

Significant portions of coding, including the generation of boilerplate code and routine tasks, are managed by automation. Developers pivot to a supervisory role, focusing on high-level design and complex problem-solving. As a developer, you’re still able to inspect the output of the autonomous engine and make changes and updates as needed.

This autonomy revolutionizes coding workflows, accelerates development timelines, and allows human talent to venture into uncharted territories of innovation. It's a powerful testament to the synergy between developer intuition and algorithmic precision.

dropdown arrow

Level 4

Highly Autonomous

Level 4 is highly autonomous -  the art of coding begins to transcend human boundaries. In automotive, Level 4 means a driver is not required, but driving is only available if very specific conditions are met.

At this level you’re able to generate entire modules, or applications, based on high-level requirements and specifications. Automated testing, deployment, and integration are standard.

Here, the role of the developer morphs from creator to curator, overseeing the process and focusing on strategy rather than syntax. Developers are primarily involved in specifying requirements, reviewing generated code, and handling exceptions or edge cases that automation cannot resolve.

This is the first level where full-stack code generation, integration with deployment pipelines and automated security testing are standard.This level of autonomy is not just about efficiency; it's about redefining the essence of software development, opening doors to innovation that were previously constrained by human bandwidth.

dropdown arrow

Level 5

Full Automation

Imagine… self-healing software [pause for oohs and aahs]. Level 5 systems are entirely hypothetical (for now) but will inevitably become available. To make the comparison to automotive - this would be the equivalent of a car that doesn’t need a steering wheel or pedals and can navigate all on its own.

Unlike previous levels where AI assists or complements human developers, at Level 5, AI could potentially innovate by creating novel algorithms or solutions. It would adapt to new technologies and integrate them without human input, evolving the software solutions it manages. Level 5 systems continually learn and improve from successes and failures, optimizing over time without external direction.

In this environment the role of a traditional “software developer” would shift to setting high-level goals and constraints, such as business objectives, ethical guidelines, and legal compliance. Developers might also engage in more creative or complex problem-solving that the AI cannot handle.

Full automation raises significant ethical questions, such as accountability for software failures or unethical outcomes, the impact on employment for software developers, and the need for new forms of governance and oversight.

Achieving Level 5 automation would require overcoming substantial technical challenges, including the creation of AI with advanced general intelligence, robustness, and reliability.

dropdown arrow

The journey through the levels of code generation autonomy offers a glimpse into a future where the lines between human developers and AI tools are not just blurred—they are intertwined.

For organizations and professionals, understanding and embracing these levels will be critical in staying at the forefront of technology. As we look ahead, it's clear that the role of the traditional developer is not diminished but elevated, as they become architects of innovation in an age where code writes itself.

These levels of automation can help you better understand, and compare, the tools that are available on the market today. We encourage you to use this framework to help you navigate conversations as you continue your codegen journey.

If you’re interested in a demo of the most advanced code generation tool available today we’d love to show you what we’ve built.

Contact us today to schedule your interactive demo.

dropdown arrow