Going beyond copilot with AI Agents
summary
Claim: current code assistants produce outputs that see 50% of generated code modified or removed within two weeks. Demo builds an agented pipeline in Python using OpenAI models plus a react (reason+act) tool loop (ping, bash, web search) to give the model real environment grounding. Product prototype runs, compiles and tests generated code via an agentic repair pipeline, reducing hallucinations; examples show it actually executes commands, reads package versions and fetches live web data. Also references open-source projects SweAgent and OpenHands.
AI-written summary · about
chapters (5)
- 0:00Introduction and motivation
- 12:34Why we need AI agents
- 25:58Demo of AI agents in action
- 31:24Future of AI and developers
- 34:17Q&A session
related talks
36mGoing Beyond Copilot with AI Agents
26mAI Agents as Your Co-Developer
29mPairing with the machine: How AI Agents are rewriting the rules of software development
35mGoing Beyond Copilot with AI Agents
30mHow AI Agents Will Change the Future of Programming
16mCoding Reinvented How AI Agents Are Shaping the Future of Software Development
29mDemystifying AI Agents with Python Code
26mFrom autocomplete to agents: AI coding assistance state of play