Jun's Vision: Augmenting Human Capabilities Through Next-Gen AI Interfaces

Join me, Jun, founding engineer at Tusk, as I share my thoughts on building the next generation of AI interfaces that put humans in the center, augmenting our capabilities, and helping us be more thoughtful and creative.

  • 1. Jun is a founding engineer at Tusk and will discuss building the next generation of AI interfaces.
  • 2. The focus is on AI systems that put humans in the center, augmenting capabilities and helping with thoughtfulness and creativity.
  • 3. It is a collection of ideas Jun has been thinking about, some more speculative, to encourage builders in the space to consider these patterns and principles.
  • 4. In 2025, agents are expected to be everywhere, performing research, browsing for users, and automating tasks.
  • 5. Agent-based tooling and protocols are becoming more sophisticated, with large chunks of knowledge work anticipated to be automated in the future.
  • 6. Many AI agents focus on automating discrete tasks, which is easier to quantify, sell, benchmark, and compare one system against another.
  • 7. Over-reliance on automation can lead to general laziness and atrophy of skills; many high-judgment domains like coding and design still require tight human supervision.
  • 8. The main thesis of the talk is to help humans produce high-quality work instead of attempting to automate complex tasks suboptimally.
  • 9. Introducing ideas for augmentation-based UX, looking at interaction patterns for AI helping users review blind spots, spark creativity, and amplify thoughtful decision-making.
  • 10. Principles for designing AI products to emphasize and grow human capabilities and trustworthy human-AI partnerships will be discussed.
  • 11. Comparing automation and augmentation approaches, with automation writing the entire email or code and sending it on behalf of the user, while augmentation helps users brainstorm key points, sugge
  • 12. In augmentation, the human is still in control, with an AI thinking partner reviewing work and suggesting improvements.
  • 13. The shift in mindset moves responsibility for the task to the AI system, transforming it from an offshore contractor to a team member that grows together with the user.
  • 14. The first core interaction pattern is blind spot detection, which is immediately compelling because everyone has blind spots in their thinking.
  • 15. AI should be designed to reveal something at the right time when a unit of work can be assumed to be ready.
  • 16. Task, an AI testing platform, finds edge cases and bugs based on pull requests; it validates potential issues and surfaces them for review.
  • 17. Users can provide feedback by clicking thumbs up or thumbs down or explaining their reasoning, helping the AI learn from user reviews.
  • 18. The second pattern is cognitive partnership, moving from stateless answering machines to systems that adapt to a user's mental models.
  • 19. Building personalization without being creepy is essential; users need to feel understood but not surveyed.
  • 20. Proactive guidance, the third pattern, is the hardest to get right; it should feel like serendipity and not interruption.
  • 21. To build trust in augmentation systems, trust must be progressive, contextual, and bidirectional.
  • 22. Trust should be built with low-stake suggestions before moving to high-impact decisions, allowing a system to prove itself on small things first.
  • 23. AI should facilitate skill growth, visualize skill development, evolve with the user, and create an emotional connection with users.
  • 24. The future of AI interfaces should focus on becoming more fully human, amplifying intuition, taste, and creativity rather than replacing human capabilities.

Source: AI Engineer via YouTube

❓ What do you think? What are the most critical aspects of building trustworthy AI-human partnerships, and how can we prioritize these elements in designing augmentation systems? Feel free to share your thoughts in the comments!