Is coding assistant as good as we thought
summary
Claims that coding assistants boost productivity and test coverage, but evidence is specific: a homebuilt agent (a failed precursor) runs a deepseek LLaMA file locally on GPU, returns a verbose chain‑of‑thought that gets stripped, and outputs a JSON action plan directing Python tools to list and delete files. Benchmarks and a GitHub survey across US India Brazil report efficiency gains, yet a surprising bug appears: the agent suggests a deprecated Django logout API because of its knowledge cutoff, so generated code needs careful review.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:50AI agents and where they appear
- 5:06How AI coding agents work
- 11:52Building a simplified agent
- 17:17Why developers use agents
- 18:43Pitfalls of AI coding agents
- 21:33Using AI agents better
- 23:39Agent limitations and risks
- 25:45Five Bs for responsible AI
- 27:56Are AI coding agents as good?
related talks
30mHow AI Agents Will Change the Future of Programming
41mGoing beyond copilot with AI Agents
36mGoing Beyond Copilot with AI Agents
26mFrom autocomplete to agents: AI coding assistance state of play
30mBeyond the Hype: Using AI Coding Agents Today
49mFrom autocomplete to agents: AI coding assistance state of play
45mThe AI Revolution in Software Development: From Code Writing to Quality Assurance
20mAI Coding Assistants: Hype, Help, or Hindrance?