2026 is the year that we are asking everyone to get comfortable with using LLMs in their daily workflows. I want to talk about why that matters, what it looks like at Sentry, and how we're thinking about adoption.
If you haven't been paying attention LLMs have gotten to the point where they're actually quite powerful for a lot of tasks. They're by no means a cheap solution to many problems, but in a lot of situations we're optimizing for getting more done vs doing things the cheapest way possible. Think of this as Venture vs Private Equity: we're on the Venture side of the world, and we invest in growth. While my experience is primarily in using LLMs for engineering tasks, the same principles apply to other domains. I will focus on that for the sake of this conversation, but the message rings true for every department and every role.
You may know this already, but my role at Sentry is not contributing code; I do it because I enjoy it and it's really a core reason I'm even in this industry. It's the reason I built our MCP server, and recently I've even been able to start contributing to the core of Sentry again. What you may not know is I have not personally written code in more than six months. I've fully adopted agentic coding, and have been able to ship production changes to Sentry as well as other software with it. I've identified production security vulnerabilities, shipped internal tools, and brought a product to market fit - all by kicking off work in-between meetings and a few do-not-schedule blocks on my calendar. It's not always the most effective way to get the job done, but it's allowed me to multitask in a way that's never been possible before, and it's created productivity gains like I've never seen. It's not without its problems, but it's here.
Now is the time to get comfortable with using these tools yourselves. Many of you are, and some of you are in as deep as I am. Others may still be skeptical or just haven't taken the plunge yet. I want you to make the time to get comfortable with the tools, talk to your peers using them, and become experts. This is now a necessary skill for you to have in your career in order to stay competitive. It doesn't mean it's going to replace the skills you've already developed, but it is a huge boon to remove monotonous tasks, to improve your quality of life, to free up time for more interesting work. It builds on the domain knowledge you've already developed and makes you much more capable in your role.
At Sentry we've been forward thinking here in the engineering org. We've unlocked budget, relaxed IP restrictions, and opened up the most cutting edge tools to our engineers. We've also been very intentional about adoption, but so far we haven't seen the organic growth we'd like in the company. About half of engineering is using AI tools in their daily workflows, but only a small subset of engineers have gotten truly comfortable with the state of the art. I want everyone to get comfortable with using the tools throughout your day - you don't have to go as far as I have, but you need to develop expertise in using them. It is quickly becoming a required skill.
Our focus in 2026 is full adoption across the company. I will be focusing on engineering, but the expectation is across the board adoption. We're going to be looking at how trends are going, trying to understand which tools are and are not working, and look for ways to accelerate education and adoption. This will rely on everyone at the company leaning in, both in making the time to learn the technology, going into it with an open mind, and being willing to help your peers. There's a lot of things happening and they're changing constantly, which means it's a continuous learning process.
One example of investment here is a tool I built over the last two days called Abacus. It's intended to help us understand adoption within engineering. I had to build this tool because nothing existed, and I was able to do it in two days because I am well versed at using Claude Code in combination with my engineering domain experience. We'll be using this to help gauge adoption - we want to see more commits touched by AI, more "average usage" of tools within the org. This project is a great testament of why the tools are valuable, but also I'm hoping folks within engineering and adjacent orgs get some value out of the visibility as well.
If you're in engineering and have ideas for what we can do here to grow adoption, to improve how we're using the tools today, reach out to PERSON or OTHER_PERSON as they're good points of contact on this. There's also the #SLACK_CHANNEL for ad-hoc discussions. Again, this is not something that we can solve top-down, we will need folks to lean in and we are going to expect it.
If you're in another department, take the lead or work with your peers to learn how you can best adopt the tooling. I know many of you are, but I'm sure there's just as many that aren't certain what would help them in the day to day. The easiest place to start? Use that ChatGPT subscription we give you. If you've got something you're confused about, drop a question in there and see if it helps you. Maybe you're working on a new project that you're unfamiliar with? Have ChatGPT help you come up with a project plan. You'll be amazed at how fast the tools are progressing, especially if you tried them in the past and were not satisfied. The agents are a great peer.
Lastly, you may see big numbers when it comes to how much money we're spending on some of these tools. Make no mistake the numbers are not comfortable, and there will be a point where we need to address that. For now we're simply looking to stay ahead of the curve and understand what is working and what isn't, so we're being more relaxed with budgetary spending. This is especially true within engineering where some of us are spending more than $100/day on these tools. Yes, it's a lot. Yes, it's worth it. Yes, we will fix it at some point.
Pardon the typos, I've got more things to ship ;)