The Next Coding Horizon - Proactive LLMs for Enhanced Productivity
summary
Large Language Models cut coding time dramatically: internal JetBrains research reports 91% of users see time savings, 71% faster task completion and 58% reduced cognitive load, while a 40% acceptance rate for the Melum code completions improves latency and relevance and will be open sourced soon. New agentic tools like Windsurf and Juny generate, build and self-validate projects (sometimes failing and needing iterations), rely on RAG and large context windows, and push toward VIP/vibe coding where regeneration beats manual debugging.
AI-written summary · about
chapters (11)
- 0:00Welcome and conference intro
- 3:43The AI evolution
- 8:34Developer sentiment on AI
- 12:44Efficacy and productivity
- 16:54Live demo: Windsurf
- 26:01Choosing the right LLM
- 31:29Local models and RAG
- 37:42Levels of AI assistance
- 42:21Agentic coding and future
- 47:08Mental models and validation
- 52:42Vibe coding and new roles
related talks
34mHow to train a fast LLM for coding tasks
26mMastering Large Language Models: Prompt Engineering for Precision & Performance
40mLLM for Coding, the State and lnitiative
43mMachine Learning for Developer Productivity
26mAI-powered Productivity using Open-Source LLMs
44mJoint Keynote Address: Accelerating Software Development: The LLM (R)evolution
01mSpeed Up Your Code With LLM's!- 44mAccelerating Software Development: The LLM (R)evolution