Navigating the GenAI Landscape with Coder

on March 27th, 2024

Updated on October 16th, 2024

6 min read

A 5% improvement in developer productivity lowers your software engineering costs by 1%.

If you employ 10,000 developers, that’s an $18M savings based on insights from our latest Cloud Development Environment Adoption Report.

Put simply: Making enterprise developers fractionally more productive is a big deal. It’s why developer productivity is a perennial top priority for leaders.

It’s also why generative AI (GenAI) is a hot topic among such leaders. GenAI can make a developer 20% more productive!

Coder is a developer productivity company. So it’s only natural that we embrace GenAI in all its glory, right?

Nope.

And here’s why.

GenAI in Software Development Requires Choice

There is no doubt GenAI is transformational. I don’t use that word lightly or with a hint of hyperbole. I believe it will be as fundamental a technology shift – and as ubiquitous – as smartphones or cloud computing.

But as with any new technology wave, it will be bumpy. Disruptive technologies always are. It would be easy to hold off adopting GenAI in our core product given this volatility. We’ve seen Google’s Gemini get “too big too soon,” Microsoft’s Copilot hallucinated election results, and ChatGPT falsely labeled students as cheaters.

But immaturity is not the reason Coder isn’t building GenAI into our products.

The reality is our customers have asked us to not bake GenAI technology directly into the platform. Why? Because Coder is, at its heart, a provisioner. We set up and maintain development environments. Coder’s founders, Kyle and Ammar, ensured Coder was unopinionated in its approach. A core value is getting developers safe access to the right tools so they can stay in flow, writing code.

The moment we take that decision out of the hands of our customers by picking a GenAI tool, we become opinionated. That’s counter to why customers choose Coder over competing cloud development environments.

GenAI requires choice. Choice of which tools to use. Choice of which individuals get access to those tools. And even choice for individuals to apply different tools to different projects.

The challenge of GenAI is one of governance, not code completion.

Three Ways Coder Makes GenAI Better

Coder is a GenAI governance solution. In fact, it drives about 40% of our business today. But it’s far too nuanced to simply say Coder doesn’t make the GenAI tools you buy, we make GenAI tools you buy better (to paraphrase the infamous BASF).

Let me be more specific. There are three ways Coder makes the GenAI tools you buy better:

Lowering the cost of developing your own GenAI tools
Accelerating the rollout and governance of GenAI tools
Incorporating GenAI into tools other than coding assistance

Using Coder For GenAI: Lowering GenAI Development Costs

One of the first development environments to shift to the cloud was machine learning. Data scientists and developers incorporating AI into their core applications need:

Access to large amounts of compute
Ability to leverage GPUs for training LLMs
Proximity to the large data sets underpinning the LLMs

This is significantly more complex to do on local laptops. Even the most powerful MacBook Pros struggle to provide the compute, and loading large data sets directly onto developer workstations is cumbersome if not impossible.

Don’t bring AI/ML resources to the developer. Bring the developer to your AI/ML resources.

Coder is the only enterprise CDE optimized for this use case. We provide a template abstraction so you can right-size compute for AI/ML developers. These Templates are powered by Terraform to provision exotic infrastructure elements like GPUs. And we’re self-hosted, so the developer workspace coresides in your cloud next to your data sets.

Skydio documented their journey with Coder, noting:

"The machine learning team can develop in a workspace with cloud GPU and storage, with no need to configure and connect to these resources each time they’re needed."

Skydio is not alone. Data science and AI/ML teams are the initial users of Coder in nearly two-thirds of our customers.

Using Coder With GenAI: GenAI Provisioning and Governance

Our fastest growing use case is configuring Coder to provision and govern GenAI. We’re detailing this in our blog series on deploying GenAI at scale in traditional enterprises, including how we can deploy GitHub Copilot with a single code block.

Enabling developers to deploy GenAI tools directly on their laptop is an operational and compliance nightmare. With Coder, they’re centrally provisioned, using automation and role-based access controls. It’s simple to invoke, revoke, and audit the usage. Think GenAI boosts productivity 20%? Prove it with Coder. Worried about GenAI exfiltrating data? Prove – or disprove it – with Coder.

One large investment bank is using Coder to deploy Copilot to 2,500 of its 7,000 developers. If you want access to GenAI, then you need to go to the cloud. It’s a carrot to abandon local development and embrace cloud development. It’s a better developer experience, and ensures the bank has a safe and controlled rollout of an emerging technology.

A large auto manufacturer is experiencing similar benefits, but they’re providing their developers with access to GitHub Copilot, Google Gemini, or Amazon CodeWhisperer depending on their preferred cloud. They empower developers with choice on a project-by-project basis.

Using Coder As GenAI: Using GenAI Beyond Code Assistance

Future products from Coder will include GenAI as part of the core functionality, it just won’t be code assistance tools.

We are building a constellation of tools, each tackling a different source of developer friction. CDEs are great for eliminating the developer tax of onboarding, build times, and compliance. But what about similar points of friction around issues management, requirements management, and other operational tasks that we continue to “shift left” on the developer?

Stay tuned as we continue to innovate. Coder will be a multi-product company within a year. Keep an eye on – or contribute to! – new projects in our GitHub organization.

Our Mission: Make Developers 10x More Productive

GenAI is one aspect of the Coder strategy. Overall, we’re focusing on several ways to improve developer productivity, including:

Enhancements to our CDE product. We will radically improve Coder OSS and Enterprise, starting with a new tool to deepen insights. Stay tuned!
The launch of new open source projects based on technologies we pioneered in the CDE space. We have a lot to say about CLIs, testing, and ways to improve Go.
Incorporating GenAI thoughtfully and inline with our mission. Coder’s use of GenAI will enhance developer choice and productivity, not limit it.

Today, customers report a 5x gain in time their developers spend writing code.

We can do better. We can do 10x.