Build without limits: lessons learned from using generative AI dev tool
summary
18 months of using AI coding assistants shows measurable gains and clear limits: teams report faster prototyping and higher productivity but the single biggest real use is stack trace debugging, not bulk code generation. Examples: a one-hour GitHub data-collector prototype, Stanford studies showing variable productivity uplift, and AWS work on formal verification to curb hallucinations. Practical failures include hallucinated classes, malicious npm packages, and dependency files that pin insecure versions; success requires prompt engineering, spec-driven development, and breaking work into small, reviewable increments.
AI-written summary · about
chapters (11)
- 0:00Introduction to AI coding assistants
- 3:00Evolution of developer tools and AI
- 6:00How AI enables faster prototyping and learning
- 9:00Improving developer flow and reducing context switching
- 12:00Automation, debugging, and real-world use cases
- 15:30Risks: hallucinations, security, and bad code
- 19:00Limitations with large codebases and dependencies
- 22:00Prompt engineering and context best practices
- 25:00Spec-driven development vs vibe coding
- 28:00New developer skills in the AI era
- 31:00Future trends and augmented developers
related talks
47mLessons learned from using generative AI dev tools
42mBuild Without Limits - Lessons Learned Working with Data and AI Coding Assistants
33mAI assisted software delivery
39mAI-assisted development: Common pitfalls and smarter practices
01h 02mCan you become a 10x developer using generative AI?
43mBecome a 10x developer by harnessing the power of Generative AI
41mAI-Augmented Testing: How Generative AI, Prompt Eng Turn Testers Into Superheroes
01h 02mPANEL - Unleashing the Future: The Impact of Gen AI on Development Productivity