AI Policy

Our core philosophies for AI:

Embrace AI, Cautiously

We should be quick to integrate technology that can improve our work. AI is an example of one such technology that has vast potential to improve our work.

We must be careful to avoid being carried away by the hype surrounding AI. Many bold and inflammatory statements are put out by those in the AI industry. We must always remember to take these statements with a large amount of skepticism. These guys have to make such statements, otherwise their venture-backed funding will dry up.

This is why we embrace AI, but cautiously. We will try the new agents. We will try the new models. But we will only settle with those products and tools that prove their value by giving us real return-on-investment in our work. We will avoid vendor lock-in, knowing that any tool’s pricing may suddenly jump if the venture-backed subsidies dry up.

Coding Agents

Coding agents are a great way to save a lot of time. They automate the process of getting ideas from brain to code. Why would we manually type code when the agent can make edits much quicker than we can ever type?

We use coding agents like Amp, Shelley, and OpenCode for a lot of our coding work. We use OpenRouter as our LLM gateway for easy access to many models. OpenRouter API keys with a weekly limit are a powerful way to give tools and team members access to models without accidentally breaking the bank.

Contributor Responsibility

No matter how a developer writes their code, whether through a coding editor or a coding agent, they are responsible for the code they contribute. You can not put responsibility on the coding agent when something goes wrong. The coding agent is only a tool and can not take blame. Do not contribute code you are not comfortable supporting in production.

AI In Our Products

We have yet to integrate AI directly into our products. It seems like it could be a powerful time-saver for our customers if done well. We must do such integration cautiously and carefully. More information will be added here as we determine best practices in this area.