Coding Agents Are Secretly General Agents
Jay Hack, head of AI at ClickUp, discusses the evolution of AI from early computer vision to generalist coding agents, highlighting how 'positive transfer' makes coding an 'AGI-complete' domain. He delves into the brutal economics of AI startups facing foundation model giants, ClickUp's strategy for convergence and first-party data as a moat, and the challenges of verifiability and catastrophic forgetting. The conversation also explores LLMs at the scientific frontier, the 'car wash test' revealing limits of world models, and speculative future applications like LLM resorts and game integration.
Jay Hack, with a background in AI dating back to 2008, traces the evolution of artificial intelligence from early computer vision techniques (Sift features, SVMs) to the transformative impact of large language models. He notes that early code generation companies like TabNine and Kite, though visionary, were too early, demonstrating that momentum is key for startup success. GitHub Copilot’s integration into VS Code and Cursor’s innovative ‘composer mode’ with sidebar chat using Claude Sonnet 3.5 marked significant inflections by seamlessly injecting AI into workflows.
Hack’s personal journey into AI was driven by a fascination with recursive self-improvement and systems of self-reference, inspired by books like “Gödel, Escher, Bach.” The release of GPT-3 in 2020, with its few-shot learning capabilities, convinced him of the technology’s potential. His