In a recent LinkedIn post, Teresa Torres highlighted the innovative work of Zero Gravity, a UK-based platform dedicated to helping disadvantaged students access elite career opportunities. Torres shared insights from an episode of “Just Now Possible” featuring Product Manager Elliot Little and Software Engineer Dan St. Paul, who discussed their approach to building an AI career co-pilot.
The core challenge Zero Gravity addresses, as detailed in Torres’s coverage, is the “knowing-doing gap” – the difficulty students face in acting on opportunities even when aware of them. Their AI co-pilot is designed not as an automation tool, but as an orchestrator, bridging this critical divide.
“You’ll hear how they: Started with grand visions of AI mentors and synthetic avatars, then scaled back to something simpler and more effective… Discovered that hiding the ‘LLM magic’ backfired—students needed to feel the personalization.”
Rethinking AI Implementation for Student Engagement
Teresa Torres points out that Zero Gravity’s journey involved significant iteration. Initially, the team envisioned sophisticated AI mentors and avatars. However, they found a more grounded, simpler approach to be more effective. A key learning, as shared by Torres, was the realization that concealing the AI’s underlying technology, the “LLM magic,” proved counterproductive. Students responded better when they could perceive and feel the personalization being offered.
According to Torres, this led to a strategic shift: embracing transparency to foster student trust and engagement. The platform’s design prioritizes making the AI’s supportive role evident, rather than abstract.
Navigating Complexities in EdTech AI Development
The discussion, as relayed by Torres, also delved into the technical and ethical considerations of building AI for young learners. Zero Gravity’s team focused on practical strategies for managing long-term student interactions.
“Built context management strategies to handle multi-month student journeys without blowing up token counts… Approached safeguarding as a first-class concern when building AI for 16-year-olds.”
Teresa Torres highlights their meticulous approach to context management, essential for supporting students over extended periods without overwhelming the AI’s processing capabilities. Furthermore, safeguarding was treated as a paramount concern from the outset, especially given the target demographic of 16-year-olds.
Technical Strategies and Safeguarding
Torres notes that the Zero Gravity team employed application logic over complex Retrieval-Augmented Generation (RAG) architectures for managing tool availability and ensuring context freshness. This practical approach allowed them to maintain control over the AI’s interactions and information accuracy.
The emphasis on safeguarding, as covered by Torres, involved implementing robust moderation endpoints and collaborating with external verification services like Unitary. This layered security approach is crucial when deploying AI in educational settings.
“It’s a practical look at building AI products that augment human relationships rather than replace them—from a team navigating the unique challenges of educational technology.”
In summary, Teresa Torres’s coverage of Zero Gravity’s work on LinkedIn offers a valuable case study. It demonstrates how AI can be thoughtfully applied in education to augment human relationships and support student development, particularly for those facing significant barriers to opportunity. The insights underscore the importance of user-centric design, ethical considerations, and practical technical solutions in the EdTech space.
📝 About This Content
This article is based on insights shared by Teresa Torres on LinkedIn.
📅 Originally posted on January 8, 2026 | View original post on LinkedIn →