Open Source AI Coding Assistants: The Future of Software Development
summary
Sourcegraph builds an open-core AI coding assistant called Cody that anchors LLM outputs to a parsed code Knowledge Graph for much lower hallucination rates. Demos generate a React app using the Alpha Vantage API and recharts, render a table and chart (one run hits API rate limits but the generated app runs), and editor integrations (VS Code, JetBrains) add commands and context-aware inline completions. Early telemetry shows ~30% completion acceptance within two weeks versus competitors hitting that after months.
AI-written summary · about
chapters (11)
- 0:00Introduction to open code AI
- 1:05Common frustrations with AI coding
- 3:12AI coding interfaces today
- 4:58Context from code examples
- 7:24The squirrel test
- 11:38Building new applications
- 15:52Working in legacy codebases
- 19:26Under the hood of Cody
- 22:32Why context matters most
- 25:42Automating software development
- 26:43Q&A
related talks
31mCody, an open source AI coding assistant
41mGoing beyond copilot with AI Agents
33mAI assisted software delivery
27mAI for software development: A reality check
57mAccelerate Your Developer Productivity with AI Embrace the Future Now!
39mAI and the Democratization of Software Development
14mDeveloper productivity in a post-AI world
35mGoing Beyond Copilot with AI Agents
