Skip to content

Instantly share code, notes, and snippets.

@itproto

itproto/copilot.md

Last active September 11, 2025 15:40
Show Gist options
  • Save itproto/7dcad10de9a2ad1992a8e1be57b190e6 to your computer and use it in GitHub Desktop.
Save itproto/7dcad10de9a2ad1992a8e1be57b190e6 to your computer and use it in GitHub Desktop.
Download ZIP

Revisions

  1. itproto revised this gist Jun 10, 2025. 1 changed file with 80 additions and 0 deletions.
    80 changes: 80 additions & 0 deletions copilot.md
    View file
    Original file line number Diff line number Diff line change
    @@ -79,3 +79,83 @@ Competitors:
    4. __Commits__ summary - auto-generated
    - __PR__ summary → GitHub interface disabled
    - GitHub VSCode_workspaces_workflow → cloud-hosted dev.env

    ## 4.1 Keyboard/Settings
    - `Cmd+I` → Invoke Copilot
    - `Cmd+Shift+I` → Chat window
    - `Cmd+Alt+Shift+L` → Open Copilot settings
    - `Tab/Ctrl+Enter` → Accept suggestion
    - `Esc` → Dismiss suggestion
    - `Ctrl+Enter` → Show all suggestions
    - `Alt+][` → Next suggestion
    - `Cmd+/` → Add to context

    ---

    ## 5. Prompts

    ## 3S:
    1. Short (CONCISE) – LLM chatterbox, order of sentences `sort list|ascending` vs. "can you be so kind…" (can have typos and incomplete)
    2. Simple (no multi-tasking) – break down complex tasks
    3. Specific (CONTEXT) – use `#tags/@agents`

    ---

    ## 5.1 Context (Working set)
    > /help
    ### @AGENTS
    - @workspace
    - @vscode
    - @terminal
    - @directory/@file

    ### /SLASH Shortcuts
    - /new - scaffold
    - /generate
    - /fix
    - /tests /doc
    - /explain

    ## #Tags Contexts
    #file, #editor, #codebase, #sym

    ---

    ## 6. Limits
    - __Limited Scope__ → Check GitHub for the types of projects/languages supported. No latest documentation for openfin/esp.js
    - __Bias__ → Can reproduce some bad practices and common errors.
    - __Security__ → Potential risk of leaking client data (Copilot takes care of it).
    - __Public Code__ → Stochastic machine, can produce public code (license issues, vulnerabilities).
    - __Inaccurate__ → Copilot is not a linting tool or compiler (doesn’t fully understand code). Needs human oversight.
    - __Codex-Only__ → not a ChatGPT to answer any questions (code/context related)

    ---

    ## 7. Demo Time:
    1. Create 3 slides revealJs about Copilot in md format
    2. `As FXO team PM, create a project plan to create FXO tile screen`
    3. // function that generates random prices for 5 major currency pairs
    4. `function generateStock()`
    5. extract common code
    6. generate tests `right_click`
    7. `Q/A comments`
    ```ts

    // Q: what does rxjs stand for?
    // A:

    ```
    8. `#terminalLastCommand`, `#codebase`
    9. __Generator__ → Create simple Next.js app
    10 `@workspace` - I want to document how to run this project so that other developers can get started quickly by reading the README.md file

    ---

    ## 8. Q/A

    ---

    ## Resources
    🔗 [https://learn.microsoft.com/en-us/training/browse/?terms=github%20copilot](https://learn.microsoft.com/en-us/training/browse/?terms=github%20copilot)

  2. itproto revised this gist Jun 10, 2025. 1 changed file with 22 additions and 27 deletions.
    49 changes: 22 additions & 27 deletions copilot.md
    View file
    Original file line number Diff line number Diff line change
    @@ -1,10 +1,9 @@

    ### 💬 My Take on Codex-tools (LLM-based dev tools)
    → Similar to JavaScript → cool/strange/simple at the beginning
    → Evolution of tools/practices for defining context/prompts/LLMs
    - Similar to JavaScript → cool/strange/simple at the beginning
    - Evolution of tools/practices for defining context/prompts/LLMs

    ---
    > 3 months usage of passive use at SCB
    ### 1.1 Day-to-Day:
    - Generating tests
    - Explaining code
    @@ -15,18 +14,16 @@
    ---

    ### 🔍 Observations
    - Proxy settings (WFH)
    - Slow generation at times (stops)
    - Many settings disabled by admin (model, scopes)
    - Generation interrupted by **public code** warning
    - Slow generation
    - Settings disabled by admin (model, scopes)

    ---

    ## 2. Copilot
    ## 2. Copilot J+History
    Jun 2021: GitHub (Microsoft) released Copilot based on Codex (OpenAI GPT-3)
    - Uses FIM (fill-in-the-middle) → new LLM approach for better context
    - GPT-3.5 Turbo = success
    - System to block risky patterns/exploits
    - 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

    ---

    @@ -35,8 +32,8 @@ Jun 2021: GitHub (Microsoft) released Copilot based on Codex (OpenAI GPT-3)

    1. Subscription:
    - Individual
    - Business
    - Enterprise (learns codebase, COBOL, scans PRs for best practices)
    - Business
    - Enterprise (learns codebase, COBOL, scans PRs for the best practices)

    2. Models:
    - GPT-4.0, D3
    @@ -46,15 +43,13 @@ Jun 2021: GitHub (Microsoft) released Copilot based on Codex (OpenAI GPT-3)
    ---

    ## 2.1 Internals
    ### NLP, ML
    > NLP, ML
    1. Input Preprocess – prepare for model context and prompt (code related) → `processed_prompt`
    2. LM Analysis – neural network analyzes `processed_prompt` (large body of text)
    3. Generation – LM generates code/suggestion
    4. Output Format – indent/highlight

    ```html
    <img src="img/cop-cycle.jpg" style="width: 3000px" />
    ```
    <img src="https://images.ctfassets.net/wfutmusr1t3h/16gHTMjSRTr8UtY7rb9rjG/ed6323cd793ddf005ff37fe969e3ae61/2000x1200-Inline-Graphic.jpg?w=1280&q=75" style="width: 3000px" />

    ---

    @@ -67,20 +62,20 @@ Jun 2021: GitHub (Microsoft) released Copilot based on Codex (OpenAI GPT-3)
    Competitors:
    - Amazon CodeWhisperer
    - Cursor VSCode Fork
    - Claude Code
    - Claude Code (recent impressive results)

    ---

    ## 4. Features:
    → Completion → VSCode (ghost text)
    → Smart Actions:
    1. __Completion__ → VSCode (ghost text)
    2. __Smart Actions__:
    - __Explain__
    - __Fix__
    - __GenTest__
    - __GenDocs__
    → Cmd+I → change/refactor
    → 3.2 Cmd+I in __terminal__
    → 3.3 QuickChat
    → Commits_summary → auto-generated
    → PR_summary → GitHub interface disabled
    → GitHub VSCode_workspaces_workflow → cloud-hosted dev.env
    3. Cmd+I → change/refactor
    3.2 Cmd+I in __terminal__
    3.3 QuickChat
    4. __Commits__ summary - auto-generated
    - __PR__ summary → GitHub interface disabled
    - GitHub VSCode_workspaces_workflow → cloud-hosted dev.env
  3. itproto created this gist Jun 10, 2025.
    86 changes: 86 additions & 0 deletions copilot.md
    View file
    Original file line number Diff line number Diff line change
    @@ -0,0 +1,86 @@

    ### 💬 My Take on Codex-tools (LLM-based dev tools)
    → Similar to JavaScript → cool/strange/simple at the beginning
    → Evolution of tools/practices for defining context/prompts/LLMs

    > 3 months usage of passive use at SCB
    ### 1.1 Day-to-Day:
    - Generating tests
    - Explaining code
    - Simple refactoring
    - Code generation
    - Reading errors/extracting values

    ---

    ### 🔍 Observations
    - Proxy settings (WFH)
    - Slow generation at times (stops)
    - Many settings disabled by admin (model, scopes)
    - Generation interrupted by **public code** warning

    ---

    ## 2. Copilot
    Jun 2021: GitHub (Microsoft) released Copilot based on Codex (OpenAI GPT-3)
    - Uses FIM (fill-in-the-middle) → new LLM approach for better context
    - GPT-3.5 Turbo = success
    - System to block risky patterns/exploits

    ---

    ### 🔧 Settings
    🔗 https://github.com/settings/copilot

    1. Subscription:
    - Individual
    - Business
    - Enterprise (learns codebase, COBOL, scans PRs for best practices)

    2. Models:
    - GPT-4.0, D3
    - Anthropic Claude 3.7
    - Gemini 2.0 Flash

    ---

    ## 2.1 Internals
    ### NLP, ML
    1. Input Preprocess – prepare for model context and prompt (code related) → `processed_prompt`
    2. LM Analysis – neural network analyzes `processed_prompt` (large body of text)
    3. Generation – LM generates code/suggestion
    4. Output Format – indent/highlight

    ```html
    <img src="img/cop-cycle.jpg" style="width: 3000px" />
    ```

    ---

    ## 3. Stats
    - **AMD 2023**: Specific HDL Language – 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

    ---

    ## 4. Features:
    → Completion → VSCode (ghost text)
    → Smart Actions:
    - __Explain__
    - __Fix__
    - __GenTest__
    - __GenDocs__
    → Cmd+I → change/refactor
    → 3.2 Cmd+I in __terminal__
    → 3.3 QuickChat
    → Commits_summary → auto-generated
    → PR_summary → GitHub interface disabled
    → GitHub VSCode_workspaces_workflow → cloud-hosted dev.env