Workflow · saas · workflow
How GitHub Copilot is getting better at understanding your code
GitHub Copilot's underlying transformers could process only about 6,000 characters at a time, meaning not all of a developer's code could be used as context; the first version was also limited to the single file a developer was actively working in.
How it works
Common implementation structure
How this type of workflow is generally built, generalized across documented cases — not tied to any one vendor's stack. Click any stage to read what happens there. Specific products that implement these stages appear in “Tools commonly seen” below.
Stage 1 · Developer codes in IDE
GitHub Copilot generates coding suggestions whether the developer is currently writing, just finished a comment, or is in the middle of code.
Tools used
GitHub CopilotCodexGPT-3vector database
Outcome
Neighboring tabs increased user acceptance of GitHub Copilot's suggestions by 5% and Fill-in-the-Middle gave a 10% relative boost in performance; GitHub's quantitative research found developers code up to 55% faster while using the pair programmer.
What failed first
Prior to Fill-in-the-Middle, only code before the cursor was included in prompts; code after the cursor was entirely ignored. The original version could also only consider the single active file.
Results
Volume5%
Running sinceJune 2022 (generally available)
Grounding & classification
Source type: technical build writeup
22 fields verified against source quotes, 1 dropped as unverifiable.
code generationragcode diff prknowledge basefailure mode describedmetric backedproduction runtime claimedsource backedtools describedsoftwareaccuracy improvementemployee productivitytechnical build writeuprag answering