- Similar to JavaScript β cool/strange/simple at the beginning
- Evolution of tools/practices for defining context/prompts/LLMs
3 months usage of passive use at SCB
- Generating tests
- Explaining code
- Simple refactoring
- Code generation
- Reading errors/extracting values
- Slow generation
- Settings disabled by admin (model, scopes)
Jun 2021: GitHub (Microsoft) released Copilot based on Codex (OpenAI GPT-3)
- Uses FIM (fill-in-the-middle) - new LLM approach for capuring context
- GPT-3.5 Turbo - first success
- Introduced system to block risky patterns/exploits
π https://github.com/settings/copilot
-
Subscription:
- Individual
- Business
- Enterprise (learns codebase, COBOL, scans PRs for the best practices)
-
Models:
- GPT-4.0, D3
- Anthropic Claude 3.7
- Gemini 2.0 Flash
NLP, ML
- Input Preprocess β prepare for model context and prompt (code related) β
processed_prompt - LM Analysis β neural network analyzes
processed_prompt(large body of text) - Generation β LM generates code/suggestion
- Output Format β indent/highlight
- AMD 2023: Specific HDL Language β generated code better aligned with standards.
- Shopify: Out of 2,000 developers, 75% used Copilot, with 26% accepting suggestions.
- Accenture: Out of 500 developers, 35% accepted suggestions.
- General Satisfaction: 75% of users are satisfied with Copilot.
Competitors:
- Amazon CodeWhisperer
- Cursor VSCode Fork
- Claude Code (recent impressive results)
- Completion β VSCode (ghost text)
- Smart Actions:
- Explain
- Fix
- GenTest
- GenDocs
- Cmd+I β change/refactor
3.2 Cmd+I in terminal
3.3 QuickChat - Commits summary - auto-generated
- PR summary β GitHub interface disabled
- GitHub VSCode_workspaces_workflow β cloud-hosted dev.env


