Introducing Dusk Labs

Dusk Labs began in 2021 with three friends and a shared desire to build useful technology together. Our first project was with a privacy-focused financial technology startup, where we developed a stablecoin and on-chain lending product. Through that work, we discovered a shared passion for combining our technical skills with business judgment to turn ambitious ideas into meaningful products.

During this work, we met Polaris Partner Group and formed a business partnership. Through our engineering experience and their decades of enterprise expertise, we’ve grown and expanded, working quietly and developing financial products such as on-chain trading along with security solutions involving novel cryptography.

Now, after careful thought, we’re happy to share our plans for the next phase of Dusk Labs.

AI Research & Development

AI has become an important part of how we build software and support our clients. Outside of writing software, We’ve used it for tasks such as security testing, infrastructure monitoring, and cost reduction.

Moving forward, we are actively investing into producing our own AI research work to hone our expertise around the technology. In doing so, we hope to offer novel specialized expertise to our clients while contributing to the expansion of the frontier.

Cybersecurity

The introduction of capable open weight models like DeepSeek v4.1 Flash have continued to change the landscape of the cybersecurity space. With agent-powered actors, threats can emerge and develop faster than many organizations can respond. Through Polaris, we have worked with Symmatrics on symmetric key delivery and post-quantum encryption, and built expertise around the potential risks AI-accelerated research poses to newer post-quantum methods such as ML-KEM.

With careful testing and monitoring only serving as a baseline defense, we are working with Tracebit to bolster our security offerings and bring a proactive approach to our clients. In alignment with our general initiative around AI R&D, we will develop persistent workflows around blue team and red team agents, enhancing the ability of businesses to react to AI-assisted attackers at machine speed, along with expanding general awareness around new AI-resistent techniques such as context bombs and canary credentials.

Open Source

Open source has played an important role in our growth as engineers and as a company. Learning from other builders has allowed us to move quickly and stay close to new ideas and better ways of working. Now that we are in a position to give back, we plan on contributing both funding and development to projects in the space.

In the coming months, we will be releasing a harness for persistent AI agent workflows, utilizing a blend of novel context and memory management techniques. Learnings and thoughts synthesized throughout the engineering process will be published through our blog. With this contribution, we hope to expand on the exploration of how humans and AI agents can work together over long periods without losing information, judgement, or control that good work requires.

Conclusion

The past five years have taught us how to solve difficult problems behind the scenes. We look forward to the future where we’ll be sharing more of our work with others and fostering further growth and opportunities in the engineering space.