Our OKRs for Q4 2026
Community, recurring workloads, a dependable platform, and a growing team: what we want to achieve this quarter and how we will measure it.

Our ambition: make BoxLite the default home for agents and the applications they build.
We want developers to choose BoxLite whenever an agent needs a computer: to build, run, and deploy useful work. Earning that position means winning developers’ confidence, turning experiments into recurring workloads, and building a community that helps shape the product.
Q4 is about proving we can earn that role. This quarter we focus on developers, because they build the paths everyone else will use. We are putting our goals in public so developers can judge our progress, challenge our priorities, and help us build something they depend on. Airbyte’s public OKR post inspired this format.
How we use OKRs
Our goals should force us to make choices and deliver results developers can feel. We will review progress weekly and give the team room to choose the best way to reach each result.
We will count real usage, accepted contributions, and useful responses to problems. When priorities change, we will explain why and keep the original targets visible. These are company goals; they are not individual performance scores.
Our objectives for Q4 2026
Target date: December 31, 2026
1. Build a community that shapes BoxLite
We want developers to help determine what BoxLite becomes: build with it, challenge its limitations, contribute improvements, and bring other builders with them.
Key results
- Weekly active members on Discord
- GitHub stars
- External developers who open issues
- New and returning external contributors
The key results form a ladder, from joining the conversation to returning with a second contribution. Returning contributors are a strong signal: they found something worth improving and chose to invest more of their time.
A weekly active member posts or replies on Discord in a given week. External means not employed or paid by Polygala, Inc., the company behind BoxLite. A contribution is accepted code, documentation, or a reproducible test case: a bug report with steps we can rerun and confirm.
2. Win recurring workloads
We want teams to put BoxLite on the path of work they depend on, return every week, and choose to pay for the value it delivers. Our job is to turn early interest into lasting use.
Key results
- Active external teams
- Design partners
- Active boxes per active team
A box is active when it runs a real application or agent workflow at least once a day, and a team is active when it has at least one active box. We count BoxLite Cloud teams from usage data, and local teams when they confirm their usage with us. A design partner is a team that agreed to run its intended workflow on BoxLite and give us feedback every week. We want active boxes per team to keep growing: our goal is for teams to run all their agent workloads on BoxLite.
These groups can overlap. Together, they show whether teams keep using BoxLite and get the results they came for.
3. Become a platform developers depend on
Developers should be able to get started quickly, run their work reliably, and get a useful answer when something fails. Every frustrating setup, failed operation, and unanswered report is work for us to address.
Key results
- Success rate when creating, starting, or resuming a Cloud box: at least 99.9%, and 99.99% by the end of 2027.
- Response time to any external message on GitHub or Discord: under one business day.
- Time to resolve high-priority bugs: 1.5 days.
We measure the success rate separately for each operation over the final 28 days of the quarter, with at least 1,000 operations of each type; a per-minute automated check adds operations when real traffic is light. Service failures, including operations the service abandons after its own timeout, count as failures; user errors, such as a missing image or an exceeded quota, do not. A response must address the message; an automated acknowledgment does not count. A high-priority bug blocks a team's work or puts its data at risk.
4. Build the team
We are growing the team to deliver these results and the work that follows. We want every new teammate to own a result from start to finish.
Key results
- 2 A+ engineers
- 1 A+ teammate each in go-to-market, product, and developer relations
If you want to help build BoxLite, email founder@boxlite.ai.
What we will work on next
We will prioritize projects that make BoxLite easier to adopt and more valuable to the teams already building with it. Foundation work prepares later quarters and moves no Q4 key result.
Community and customer efforts
- Help developers ship useful applications through practical examples, tutorials, and direct feedback.
- Turn design-partner feedback into product improvements and make it easier for community members to contribute.
Product efforts
- Bring your own images. Use private images from your own registry and manage the images your boxes run.
- Room to grow. Larger, flexible disks for Cloud boxes that can grow while the box runs.
- The same features everywhere. Bring the SDKs and the REST API to parity, including attach, resumable exec sessions, fine-grained file access, and git helpers.
- Catch failures before developers do. Cloud error-rate monitoring and alerts, plus a per-minute automated check of the full customer flow.
- Reliability in public. A public page showing create, start, and resume success rates for BoxLite Cloud.
- An agent in every box. A path from task to live application. Equip boxes with agents and provide the authentication, database, storage, and filesystem services their applications need.
- Keep your box running. Pay only for what it uses. Scale CPU, memory, disk, and GPU with demand. Base charges on actual consumption rather than reserved capacity: zero CPU consumption should mean zero CPU charge, even while the box is running.
Foundation and long-term work
- Build the foundation for that ambition. Advance our own VMM and the BoxLite Cloud rewrite in Rust through working milestones that demonstrate the improvements.
- Make local and cloud interchangeable. Work toward instant migration in both directions, carrying running state and files so developers can move without starting over.
How we will report back
At quarter end, we will publish the results for every key result, including where we fell short. We will explain what developers taught us and what we will change. Progress should be visible in the product and in the numbers.
If you are building with BoxLite, tell us what you want to run and what is getting in your way. Join us on Discord or open an issue on GitHub.
