Published on

Our Cultural Principles

Written by
Codeium Team
Our Cultural Principles.

What is it like to work at Codeium? The culture of a company is critical, but is often reduced to simple platitudes that don’t have actionable corollaries. Yes, we believe in treating each other with respect and bringing our best to work, but these statements do not help make the unique aspects of the Codeium experience concrete to anyone, ourselves or folks interested in working here. And yes, every company needs to be customer obsessed and build a great product, because otherwise they will die, so those statements are not meaningful either.

Instead, we focused on cultural principles that have direct corollaries that reflect how we truly operate as a team. Many of these aren’t core to other companies, and therefore the resulting actions aren’t what you would see executed at every company. However, that is exactly why we believe these are helpful to understand what makes Codeium different.


We solve problems, not build ideas

We stay close to reality and operate in a data-driven manner, which sounds simple, but, given the number of AI startups without a product or userbase, is clearly not a widespread priority.

We understand that ideas are cheap, while effort and execution are not, and in an industry where there are no existing playbooks, we ship fast and seek feedback often so that we do not develop in a vacuum. We have a lot of well-informed theses on what could work, but do not pretend like we definitely know the answers. We are not a research lab and are not tied to methods, features, or products if they do not drive value, and are obsessed and honest with ourselves on what that value is, unafraid to roll back effort. We are far less concerned about being right on an original thesis than eventually reaching a result that works.

This is why one of the earliest bets as a company was to make our individual product completely free. All other companies charge individual developers (or create an incredibly nerfed free tier), but we greatly value having a large user community that will rapidly give us feedback on our cutting edge technology. This way we know with statistical significance that we are solving real problems, so that long-term we are simply the best product.

We maintain healthy realism

We question our decisions and position in this rapidly changing industry, but do not get carried away into paranoia.

There are dangers in both ignoring the progress of competitors and overreacting to every announcement made, especially since it is no secret that there is a lot of hype in this rapidly evolving AI industry. We try to keep realism both internally and externally, by moving fast to out-execute competitions while reacting to industry changes in a measured manner, but also being honest with customers about the realities of the technology today. We seek out and welcome multiple opinions and build towards diversity in thought so that we do not misinterpret external changes and internal positions, but at the same time do not paralyze ourselves by playing devil’s advocate for the sake of it. We should expect that there could be tectonic shifts in the underlying industry and technology, so while we make sure to celebrate wins, we do not view progress with rose-tinted glasses and assume we deserve the success.

There are many examples of when we react when appropriate and stay calm when there is hype. When ChatGPT was launched in November 2022, we assessed the technical feasibility and end value of such a modality for Codeium. Until then, our plans were focused on shipping domain specific models that could document or fix bugs and had none for a general purpose chat. We changed directions and shipped Codeium Chat in April 2023, seven months earlier than GitHub Copilot. This was us taking swift action. On the flip side, when GitHub Copilot announced PR summarization in March 2023, we actually tried the beta functionality and realized that the accuracy bar was so far off that it wouldn’t be valuable for quite a while. So, instead of rushing to just copy a feature for the sake of it, we focused our efforts in improving real value-driving capabilities. This feature from Copilot is still not generally accessible.

We run lean

We see smallness as an advantage in that it forces us to prioritize properly, increases decision making speed and quickly de-risks larger investments.

As any developer at a large organization can relate to, headcount bloat is a real problem that slows progress down, and as a startup in a fast moving space, we cannot allow that to happen. Instead, we optimize people’s time across the entire organization, automating aggressively to avoid work that is “linear” in headcount if manual. We have high expectations for the quality and quantity of each other’s work, but at the same time, no one is too important to perform a task or validate a hypothesis. We strongly believe that pedigree and work experience is never the only reason to hire someone, and hire for specialties only when the need has outgrown the people working on it.

At Codeium, we have exceptional people with backgrounds ranging from autonomous vehicles to quantitative finance to security, to bring together fresh approaches and perspectives to the most difficult machine learning and systems problems. However, there is a common thread of people who can learn very fast and creatively work hard towards focused goals. Anyone who interviews at Codeium knows that our focus is to assess creativity and real coding ability, with a strong foundation of algorithms and systems as needed for the role.

We seek unconventional advantages

In short, we actively seek out the blind spots in industry thinking when building and selling.

This is a recognition that in order to win in a crowded market, especially as an early stage company, incremental gains and conventional thinking will be insufficient. The default state of a startup statistically is failure, so we are flexible in process and work hard to create opportunities out of possibilities. Technically, we analyze critical problems from first principles, taking into account the immaturity of our industry, and allow ourselves to take scary technical bets that are balanced against reasonable future expectations. We cannot operate to a customer’s expectations because that means we are running too safe; we have to surprise customers with solutions that they often do not believe are possible.

A lot of our technical decisions are grounded in unconventional advantages. One of the biggest ones was launching a self-hosted enterprise product far before we launched a SaaS one. We realized that the most surefire way we could bootstrap learning from real enterprises was to identify enterprises that could not use existing tools. Specifically, in a world of SaaS-only tools, we would be the only solution for regulated industries that could not send their code outside of their tenant (think defense, finance, healthcare, etc).

We give away our Legos

At the core, this means we share the work and optimize for company success rather than individual glory, knowing that will come as well.

This phrase was not coined by us! Molly Graham, a major force in defining Meta’s culture during its scaling, is often cited as a proponent of this principle. In such a cutting-edge field, we constantly share knowledge and insights to improve everyone and foster unexpected new ideas. Tactically, where possible, we avoid titles and fixed teams because this promotes an isolationist mindset that stifles such sharing. This principle is also a recognition that there is an infinite space of things that we can be working on, and so it is intuitive to tackle more and more ambitious problems. However, this means handing off existing work to make time for the new problems, and so we trust that credit will be given for being part of a solution, no matter the stage. It also means that to trust the next person with our existing work, we maintain a high bar on hiring and a high degree of accountability day-to-day so there is a default trust in others to continue on the work at the same high level.

It is very common for engineers on the team to have built systems and tools all across the product stack, to have done everything from frontend and model training. It is also not uncommon for a member of our GTM team to transition a customer relationship to another, even before an agreement has been reached.


As our company grows, our cultural principles will hopefully evolve to reflect who we must be in the moment, but we believe that today, we have a team, a determination, and an attitude to be industry-defining. If these principles resonate with you, please apply to one of our positions on our careers page.

Explore more

Jun 11, 2024enterprise7 min

A deeper dive into recent enterprise readiness capabilities: indexing access controls, subteam analy...

Apr 30, 2024enterprise1 min

WWT uses Codeium to accelerate their developers.

Apr 18, 2024company1 min

Codeium is the only AI code assistant in the 2024 Forbes AI 50 List.

Stay up to date on the latest Codeium & AI dev tool news

Please enter a valid email address