Finding the right location matters. When your team is arriving from Africa, Asia, and across Americas, easy air connections, reasonable costs, and minimal visa friction are all prerequisites. Curaçao checks every box: a compact, walkable island with direct flights from multiple continents, no visa required for many of our team members’ passports, and surprisingly, a cost that doesn’t require a corporate travel budget to manage. The turquoise water and year-round warmth don’t hurt either. So, this year we returned to Curaçao a second time.
Two forward-facing themes shaped everything we discussed, planned, and built. The first: Getting to “Done!” faster, for all the developers who use Code on the Go (CoGo™), and for our own internal team. The second: Using AI more effectively, both as a development tool for our users and as a force multiplier for our own work. In our products, we’re committed to AI that works without a live internet connection. That’s not just a philosophical position; it’s a practical requirement for the communities we serve, where reliable broadband is often the exception instead of the rule. Local AI also has a welcome side effect: it doesn’t leak private data or depend on massive, energy-hungry data centers.
Before turning to what’s next, it’s worth reflecting on our past. Some of the most interesting features in our first release—agentic AI integration (Gemini) and a robust plugin system—grew directly out of 2025’s week in Curaçao, proof that what we work on during our in-person all-hands retreat actually ships.
The 2026 Hackathon: Two days, five proofs-of-concept
Like last year, two days were dedicated to our hackathon. The instructions: Do something related to CoGo. Be amazing. Here’s what the team created.
CPU and Memory Profiler
A mobile device has tight constraints on power and heat: a small battery and less memory than computers on your desk or lap. Until now, there were few phone-native tools to find and fix performance bottlenecks and memory bloat. That changes here. The CPU and Memory Profiler lets you track your app’s real-time CPU usage and use flame graphs to pinpoint what’s consuming the most processing time. You can freeze the program at key points to inspect memory usage without needing a second computer. Performance tuning will no longer be restricted to those with high-end hardware. You’ll do it with the device you already have and you’ll get to Done! faster.
Go Components
No one builds an Android app entirely from scratch; you build on reusable components created by others. These components are powerful but configuring them means wading through endless options in fragmented, text-based formats—a tedious process that drains your productivity.
Go Components offers a better way forward. Instead of digging through configuration files, you select a component, see a clear graphical display of its available options and defaults, make your choices, and press Go. Within a second, the configured component is inserted into your code. If something isn’t right, you can tweak it and try again instantly. Because the cost of making a mistake is practically zero, you can iterate rapidly, deliver a better app to your users, and get to Done! faster.
Speech to Code
Speech-to-text has been around for years, but speech-to-code is a much harder problem. Code is structurally brittle; in languages like Python or Fortran, a single misplaced space completely changes what a program does. This makes voice-control for an IDE genuinely difficult.
To solve this, our project uses multiple layers of AI working cooperatively on the device. Each layer generates a different interpretation of your voice, and the pipeline then selects the option that best matches your intent. You talk, it codes. It means less thumb typing, more fun, and a faster path to Done!. (And, it captures our AI theme for the week.)
Collaborative Peer Programming
Research shows that pair programming accelerates problem-solving and produces cleaner code. You can’t easily share a small phone screen… but what if you could share two?
This project allows two or more people on the same local network to share their screens and code cooperatively. The connection is password-protected and entirely local, an architectural choice to guarantee privacy and safety without requiring an internet connection. If the host shares control, any connected device can edit the code, creating seamless opportunities for mobile pair programming, mentorship, and collaborative troubleshooting.
If the host restricts control, connected devices can follow along in a “look but don’t touch” mode. A teacher could project their live coding environment directly to a classroom of students’ phones, with no projector or darkened room required.
Semantic documentation search
CoGo contains an immense amount of documentation, including the Android Developer website, Java and Kotlin documentation, and more. But finding the right answers across tens of thousands of pages is challenging. Traditional keyword searches often generate too many irrelevant results while skipping related materials that lack the exact search string.
Semantic search solves this by searching by concepts rather than keywords. Your query is mapped into a multi-dimensional space where the nearest results share related meaning, delivering a measurable boost in precision. It’s worth noting this project runs the entire semantic pipeline directly on-device, honoring our commitment to AI that doesn’t require a data center, respects your privacy, and barely touches the battery. For newer phones with more storage, users can also download a localized AI model that adds automated summaries on top of the search results to save even more time.
What’s next
Our two themes for the coming year—getting to Done! faster and using AI more effectively—are at the heart of nearly everything you’ve just read about. These are proof-of-concept projects and not ready to be released, even as experimental features. We’ll let you know when they are.
CoGo is free and open source. We don’t sell any part of it, and we’re not seeking to profit from the communities we serve. Our goal is simply to build the best possible phone-native IDE, and then get out of the way so you can build great apps.
Tell us what you think. Which projects genuinely excite you? Which ones would make your life better? Email us. (We read everything.) And if you want to support our work, please sponsor us. Happy coding!
