AI Agents as Your Co-Developer
summary
Agents now take issue descriptions and produce branches, run tests, and open PRs with tools like GitHub Copilot agent mode, Bolt, Replit, and Cursor, and on SWE-bench top models resolve 72–77% of tasks versus ~30% a year earlier. Live coding can scaffold full apps but often needs followup work; industry surveys report 41% of code AI-generated, 20–40% productivity gains, yet 96% of developers don’t fully trust output and AI causes about one in five security breaches. Safe adoption uses progressive autonomy, agent-validate-agent, and auto-merge where tests justify it.
AI-written summary · about
chapters (12)
- 0:00AI agents as co-developers
- 1:04Autocomplete to autonomous agents
- 6:34Agentic loop five properties
- 8:19AI coding tools and benchmarks
- 10:30AI agents in practice
- 14:06Adoption and productivity gains
- 15:30Trust and security risks
- 17:39Hallucinations and context limits
- 19:48Safe AI agent integration
- 23:44Measuring AI code impact
- 24:05Future of self-healing systems
- 25:29Closing remarks
related talks
30mHow AI Agents Will Change the Future of Programming
01mAI Agents are the Future of Coding?
16mCoding Reinvented How AI Agents Are Shaping the Future of Software Development
27mAgentic Engineering: Working With AI, Not Just Using It
26mFrom autocomplete to agents: AI coding assistance state of play
29mPairing with the machine: How AI Agents are rewriting the rules of software development
49mFrom autocomplete to agents: AI coding assistance state of play
27mSoftware engineering with GenAI