AI Engineer Melbourne 2026 Keynote Livestream | Day 2
TL;DR
Jeremy Howard argues that AI coding tools risk trapping developers in addictive 'dark flow' states that diminish psychological well-being, drawing on Self-Determination Theory to advocate for intentional AI use that augments human mastery and autonomy rather than outsourcing complexity.
🧠 The Psychology of Flourishing 3 insights
Self-Determination Theory fundamentals
Research spanning 30 years and hundreds of experiments identifies autonomy, mastery, relatedness, and purpose as essential nutrients for human vitality and authentic motivation.
Eudaimonia versus hedonia
Lasting well-being comes from effortful growth and actualizing capacities rather than frictionless pleasure, with behavioral activation proving more effective than medication for treating depression.
Authentic flow versus dark flow
Optimal experience requires balancing high challenge with high skill in goal-directed systems, distinct from addictive 'junk flow' that creates an illusion of control without genuine growth.
⚠️ AI and the Risk of 'Dark Flow' 3 insights
The vibe coding trap
AI-assisted coding can trigger casino-like addiction cycles where developers experience dopamine hits from rapid initial progress that becomes decoupled from external validation or shippable results.
Productivity paradoxes
Engineers report achieving 95% completion in hours then spending days debugging, while companies like Uber implement strict token budgets after failing to find ROI in unconstrained AI use.
Accumulating technical debt
Teams with 200,000 lines of AI-generated code face verification nightmares and slowed shipping velocity, discovering during quarterly reviews that perceived progress lacked actual business outcomes.
💻 Augmenting Human Intellect 3 insights
Historical precedents for amplification
Pioneers including Ivan Sutherland (Sketchpad), Douglas Engelbart (Mother of All Demos), and Kenneth Iverson (APL) designed computational tools to amplify human reasoning rather than replace it.
Notation as a tool for thought
Kenneth Iverson's APL demonstrated how symbolic abstraction enables new mathematical insights, such as proving properties of entire function classes through generalized inner product operators.
Protecting autonomy and mastery
AI vendors optimize for output metrics and token consumption rather than user growth, requiring developers to intentionally select tools that teach foundational principles instead of outsourcing complexity.
Bottom Line
Treat AI as a pedagogical tool for expanding your capabilities and mastering underlying principles rather than a replacement for understanding, or risk degrading the autonomy and mastery essential to both professional effectiveness and psychological well-being.
More from AI Engineer
View all
Think You Can Build a Game with AI? Think Again! - Danielle An & David Hoe, Meta
Meta engineers Danielle An and David Hoe argue that while AI has democratized basic game creation, true differentiation requires human taste, cohesive aesthetics powered by key art anchoring, and innovative runtime LLMs that enable unscripted, dynamically personalized gameplay experiences previously impossible in traditional development.
Beyond the Harness: A Journey Towards Adaptative Engineering - Rajiv Chandegra, Annicha Labs
Rajiv Chandegra introduces 'adaptive engineering,' a paradigm shift from fixed AI harnesses (like Cursor or Claude Code) to dynamic, self-organizing systems that emerge during runtime, enabling AI to handle complex, real-world messes beyond deterministic software environments.
What if the harness mattered more than the model? - Aditya Bhargava, Etsy
Aditya Bhargava argues that sophisticated agent harnesses can compensate for weaker open-source models, enabling local AI to match proprietary performance while reducing vendor dependency.
Frontier results, on device - RL Nabors, Arize
Rachel Lee Neighbors introduces a framework for replacing expensive cloud-based frontier models with Small Language Models (SLMs) running on-device, demonstrating how a systematic 'prototype big, deploy small' approach using evaluation tools like Phoenix can cut inference costs to zero while maintaining 90% accuracy and enabling offline functionality.