Former Google and Apple Researchers Launch a Startup to Build AI’s Missing Feedback Loop

Editor's take

Trajectory, a new startup founded by former Google and Apple AI researchers, aims to address the critical gap in AI development by creating a continuous feedback mechanism for learning models. This initiative seeks to replicate the fast-paced iteration seen in areas like vibe-coding, applying it to broader AI product development.

The venture matters because the current AI development paradigm often relies on static datasets and periodic retraining, leading to slower adaptation and potential performance degradation. By building a robust feedback loop, Trajectory could empower companies to create AI systems that evolve and improve in real-time, enhancing user experience and product efficacy across diverse applications, from customer service chatbots to complex predictive analytics.

Future developments to monitor include Trajectory's ability to secure significant early customer adoption and demonstrate tangible improvements in model performance for a range of use cases. Success here would hinge on their capacity to build a scalable, secure, and privacy-preserving feedback infrastructure that integrates seamlessly with existing AI workflows, potentially influencing how companies like OpenAI or Anthropic approach model updates beyond their current release cycles.