Thinking Phrases

Instead of staring into the abyss while Copilot thinks — contemplating how agents are going to take your job — you could be doing something useful with that dead time.

This hooks into VS Code's chat.agent.thinking.phrases and turns it into a live dashboard. A launchd cron job fetches real-time data and swaps your phrases in the background. Interactive CLI to set it up. That's it.

• 📰 Catch up on the news — Google News, Hacker News, RSS/Atom, any feed
• 🐙 GitHub activity — commits, org feeds, security advisories, your timeline
• 🌦️ Get the weather — current conditions + severe alerts (lol like I go outside)
• 📈 Watch the stock market go brrr while your agents go brrr
• 🧠 Learn something new — VS Code tips, TypeScript snippets, programming languages
• 🎮 Loading screen tips — WoW, League of Legends... because why not
• 🤖 AI-powered summaries — GitHub Models rewrites articles into concise phrases

Quick start

npx thinking-phrases --interactive

No install, no clone, no config files. The interactive CLI walks you through picking sources, previewing phrases, writing to VS Code settings, and optionally installing a macOS scheduler.

Or just grab the static thinking phrases directly.

Live data sources

| Source | What it shows | Example phrase | |--------|--------------|----------------| | RSS / Atom | Any feed URL | Copilot code review now accounts for over 20% of all code reviews on GitHub, reflecting a tenfold increase in usage since its launch. — The GitHub Blog (2d ago) | | Stocks | Live ticker quotes via Yahoo Finance | BTC - USD $66,978.23 ▼ 2.02% 🟢 | | Hacker News | Top/new/best/ask/show/jobs | Multiple cursors enable simultaneous operations on syntax nodes, significantly enhancing bulk editing and refactoring capabilities. — Hacker News @ravenical 378 pts (17h ago) | | Earthquakes | USGS earthquake catalog near a ZIP | M4.2 — 12 km NE of Ridgecrest, CA — USGS (38m ago) | | Weather | Temp, humidity, wind + NWS alerts | 77°F, Mostly Cloudy, Wind E 50 mph — Fort Lauderdale, FL — Weather.gov | | Custom JSON | Any JSON API with field mapping | Article title — My API (1h ago) | | GitHub Commits | Repo or org commits with diffs | fix devtools entrypoint — vscode +1/-1 @deepak1556 (4h ago) | | GitHub Feeds | Org activity, timeline, advisories | opened a pull request in copilot-sdk — @dependabot (3m ago) |

Static phrase packs

Pre-built packs for when you just want vibes:

| Pack | Phrases | |------|---------| | VS Code tips (mac/win/linux) | 80 | | JavaScript tips | 100 | | TypeScript tips | 124 | | Ruby tips | 108 | | League of Legends loading tips | 100 | | WoW loading screen tips | 109 | | Inspirational quotes | 1,614 |

Examples

# GitHub org activity feed
npx thinking-phrases --use-github --github-mode feed --github-feed-kind organization --github-org github --github-max-items 10

# Repo commits with AI summaries
npx thinking-phrases --use-github --github-mode repo-commits --github-repo microsoft/vscode --github-max-items 5 --use-models

# Stocks + RSS
npx thinking-phrases --stocks MSFT,NVDA,TSLA --feed https://github.blog/feed/

# Earthquakes near a ZIP
npx thinking-phrases --use-earthquakes --quake-zip 94103 --quake-min-magnitude 2

# Weather conditions + alerts
npx thinking-phrases --use-weather-alerts --weather-zip 33312

# Hacker News top stories
npx thinking-phrases --use-hacker-news --hn-feed top --hn-max-items 15 --hn-min-score 100

# Custom JSON API
npx thinking-phrases --use-custom-json --json-url "https://hn.algolia.com/api/v1/search?tags=front_page" --json-items-path hits --json-title-field title

# Dry run (preview only, don't write)
npx thinking-phrases --dry-run --stocks MSFT,NVDA --use-hacker-news

Install

# Just run it (no install needed)
npx thinking-phrases --interactive

# Or install globally
npm install -g thinking-phrases
thinking-phrases --interactive

# Or clone for development
git clone https://github.com/austenstone/thinking-phrases.git
cd thinking-phrases
npm install
npm run start:interactive

Data source config

Each source can be configured via CLI flags, config files, or the interactive CLI. Here's what each one looks like in a config file:

Any feed URL. Google News, GitHub Blog, Ars Technica, your company blog — whatever. When AI is enabled, full article HTML is fetched for richer summaries.

"feeds": [
  { "url": "https://github.blog/feed/" },
  { "url": "https://news.google.com/rss/headlines/section/topic/TECHNOLOGY?hl=en-US&gl=US&ceid=US:en" }
]

Default refresh: every 6 hours.

Live prices via Yahoo Finance. Pre-market, after-hours, closed labels.

"stockQuotes": {
  "enabled": true,
  "symbols": ["MSFT", "NVDA", "TSLA", "AMZN", "GOOGL", "AMD"],
  "includeMarketState": true,
  "showClosed": false,
  "fetchIntervalSeconds": 60
}

Top, new, best, ask, show, or jobs feed. Configurable minimum score.

"hackerNews": {
  "enabled": true,
  "feed": "top",
  "maxItems": 10,
  "minScore": 50,
  "fetchIntervalSeconds": 300
}

Enter a ZIP code or place name → resolves to lat/lon → queries USGS within a radius.

"earthquakes": {
  "enabled": true,
  "zipCode": "94103",
  "minMagnitude": 2,
  "radiusKm": 500,
  "windowHours": 24,
  "limit": 10,
  "orderBy": "time",
  "fetchIntervalSeconds": 1800
}

Current conditions (temp, humidity, wind, description) + active severe weather alerts. Auto-detects location in interactive mode.

"weatherAlerts": {
  "enabled": true,
  "zipCode": "33312",
  "area": "FL",
  "minimumSeverity": "moderate",
  "limit": 10,
  "fetchIntervalSeconds": 1800
}

Severity levels: minor, moderate, severe, extreme.

Point at any JSON endpoint and map the fields. Multiple sources supported via customJsonSources[].

"customJson": {
  "enabled": true,
  "url": "https://hn.algolia.com/api/v1/search?tags=front_page",
  "itemsPath": "hits",
  "titleField": "title",
  "contentField": "summary",
  "linkField": "url",
  "sourceLabel": "HN API",
  "dateField": "created_at",
  "idField": "objectID",
  "maxItems": 10,
  "fetchIntervalSeconds": 3600
}

Three modes: repo-commits (recent commits with diffs), org-commits (push events across an org), and feed (Atom feeds — org activity, timeline, security advisories, etc.).

"githubActivity": {
  "enabled": true,
  "mode": "repo-commits",
  "repo": "microsoft/vscode",
  "branch": "main",
  "maxItems": 10,
  "sinceHours": 24,
  "tokenEnvVar": "GITHUB_TOKEN",
  "fetchIntervalSeconds": 300
}

Feed kinds: timeline, current-user-public, current-user, current-user-actor, security-advisories, organization, custom-url.

AI-powered formatting

Enable GitHub Models and each article gets individually rewritten by AI into a concise phrase. Works with gpt-4o-mini, gpt-5, o3, or any OpenAI-compatible model. Source attribution (— Source (time)) is appended automatically — the model just focuses on content.

Results are cached per-article (7-day TTL) so re-runs don't burn tokens. Falls back to template formatting if auth or inference fails.

"githubModels": {
  "enabled": true,
  "model": "openai/gpt-4o-mini",
  "fetchArticleContent": true
}

Auth resolves in order: GITHUB_MODELS_TOKEN → GITHUB_TOKEN → gh auth token.

Phrase format

Customizable templates with %variable% substitution. The engine auto-strips empty brackets and collapses whitespace.

"phraseFormatting": {
  "includeSource": true,
  "includeTime": true,
  "maxLength": 140,
  "templates": {
    "article": "%title% — %source% (%time%)",
    "hackerNews": "%title% — HN %score% (%time%)",
    "stock": "%symbol% %price% %change% %market%",
    "githubCommit": "%headline% — %repo% %delta% @%author% (%time%)",
    "githubFeed": "%action% — @%handle% (%time%)"
  }
}

Presets

The interactive CLI includes presets to get started fast:

| Preset | What you get | |--------|-------------| | Dev Pulse | Google Tech news + Hacker News top stories | | Market Watch | MSFT, NVDA, AMZN, GOOGL, AMD, TSLA stock quotes | | World Signals | Earthquakes (M4.5+) + severe weather + Hacker News best |

Scheduler (macOS)

A launchd job that refreshes phrases in the background. The interactive CLI can set this up, or do it manually:

npm run schedule             # every 1 hour (default)
npm run schedule -- 300      # every 5 minutes
npm run schedule:trigger     # run now
npm run schedule:remove      # remove it

Also available via CLI flags: --install-scheduler, --trigger-scheduler-now, --uninstall-scheduler.

Static packs

VS Code tips

| Category | Count | Description | |----------|-------|-------------| | ⌨️ Shortcuts | 25 | Multi-cursor, navigation, selection, formatting | | 🤖 Copilot | 15 | Chat, agent mode, inline chat, prompt files | | 🔀 Git | 8 | Source control, blame, staging, timeline | | 💻 Terminal | 7 | Splits, suggestions, integrated terminal | | 🐛 Debugging | 7 | Breakpoints, logpoints, conditional stops | | ✏️ Editor | 10 | Zen mode, sticky scroll, Emmet | | 😄 Funny | 8 | Flavor text |

OS-aware format — each tip has mac, windows, and linux variants:

{
  "mac": "⌘+D selects the next occurrence of a word. Keep pressing for more!",
  "windows": "Ctrl+D selects the next occurrence of a word. Keep pressing for more!",
  "linux": "Ctrl+D selects the next occurrence of a word. Keep pressing for more!"
}

Other packs

Standalone string arrays — just drop them in:

• JavaScript tips (100)
• TypeScript tips (124)
• Ruby tips (108)
• League of Legends loading tips (100)
• WoW loading screen tips (109)
• Inspirational quotes (1,614 — from dwyl/quotes, GPL-2.0)

Install a static pack directly:

npx thinking-phrases --static-pack out/typescript-tips.json

CLI reference

Commands

| Command | Description | |---------|-------------| | npx thinking-phrases --interactive | Guided interactive setup | | npx thinking-phrases --dry-run | Preview phrases without writing | | npx thinking-phrases | Write phrases to VS Code settings | | npx thinking-phrases --uninstall | Remove thinking phrases from settings | | npx thinking-phrases --static-pack <path> | Install a static phrase pack |

Global flags

| Flag | Description | |------|-------------| | --config <path> | Load a saved config file | | --settings <path> | Custom VS Code settings.json path | | --dry-run | Preview phrases without writing to settings | | --verbose | Verbose output | | --debug | Debug output (includes verbose) | | --limit <num> | Max total phrases (default: 100) | | --mode append\|replace | Append to or replace existing phrases | | --target auto\|insiders\|stable | Which VS Code edition to target | | --max-length <num> | Max phrase length in characters (default: 140) | | --no-source | Omit source attribution from phrases | | --no-time | Omit relative time from phrases |

Source flags

| Flag | Description | |------|-------------| | --feed <url> | Add an RSS/Atom feed URL (repeatable) | | --use-stocks / --no-stocks | Enable/disable stock quotes | | --stocks <SYMBOLS> | Comma-separated stock tickers | | --use-hacker-news | Enable Hacker News | | --hn-feed <type> | top, new, best, ask, show, or jobs | | --hn-max-items <num> | Max HN items | | --hn-min-score <num> | Minimum HN score filter | | --use-earthquakes | Enable earthquake source | | --quake-zip <zip> | ZIP code for earthquake search center | | --quake-place <name> | Place name for earthquake search | | --quake-min-magnitude <num> | Minimum magnitude filter | | --quake-radius-km <num> | Search radius in km | | --quake-hours <num> | Time window in hours | | --quake-order time\|magnitude | Sort order | | --quake-limit <num> | Max earthquake results | | --use-weather-alerts | Enable weather source | | --weather-zip <zip> | ZIP code for weather | | --weather-area <area> | US state code (e.g. FL) | | --weather-severity <level> | Minimum severity (minor/moderate/severe/extreme) | | --weather-limit <num> | Max weather alert results | | --use-custom-json | Enable custom JSON source | | --json-url <url> | JSON endpoint URL | | --json-items-path <path> | JSONPath to items array | | --json-title-field <field> | Field name for title | | --json-content-field <field> | Field name for content | | --json-link-field <field> | Field name for link | | --json-source-field <field> | Field name for source | | --json-source-label <label> | Fallback source label | | --json-date-field <field> | Field name for date | | --json-id-field <field> | Field name for ID | | --json-max-items <num> | Max JSON items | | --use-github | Enable GitHub activity | | --github-mode <mode> | repo-commits, org-commits, or feed | | --github-repo <owner/repo> | Target repository | | --github-org <org> | Target organization | | --github-branch <branch> | Branch filter | | --github-feed-kind <kind> | Feed type (see GitHub activity section) | | --github-feed-url <url> | Custom feed URL | | --github-max-items <num> | Max GitHub items | | --github-since-hours <num> | Lookback window in hours | | --github-token-env <var> | Env var name for GitHub token |

AI model flags

| Flag | Description | |------|-------------| | --use-models / --no-models | Enable/disable AI rewriting | | --model <name> | Model ID (default: openai/gpt-4o-mini) | | --models-endpoint <url> | Inference endpoint | | --models-token-env <var> | Env var for model auth token | | --models-max-concurrency <num> | Parallel inference requests | | --models-max-input-items <num> | Max articles sent to model | | --models-max-input-tokens <num> | Max input token budget | | --models-max-tokens <num> | Max output tokens per request | | --models-max-phrases-per-article <num> | Phrases generated per article | | --models-temperature <0-1> | Sampling temperature | | --fetch-article-content / --no-fetch-article-content | Fetch full article HTML for AI | | --max-article-content-length <num> | Max chars of article body sent to model |

Scheduler flags

| Flag | Description | |------|-------------| | --install-scheduler | Install/update macOS launchd job | | --trigger-scheduler-now | Trigger the scheduler immediately | | --uninstall-scheduler | Remove the launchd job |

How it works

Sources → Normalize → Format → Cache → Write

1. Sources fetch live data (RSS, stocks, GitHub, USGS, NOAA, JSON APIs) — each respects its own refresh interval
2. Core normalizes everything into ArticleItem or StockItem objects
3. Formatter builds display phrases from customizable templates — content first, source/metadata suffix appended
4. AI (optional) rewrites phrases via GitHub Models, with per-article caching to avoid redundant inference
5. Phrase store persists phrases per-source in ~/.cache/thinking-phrases/ so different refresh intervals don't clobber each other
6. Sink writes the merged phrases into VS Code settings.json using jsonc-parser (preserves comments and formatting)

The source catalog is modular. Each source is a { type, isEnabled, fetch } object registered in the catalog. Adding a new source means writing one file and registering it.

Portability

Settings path auto-detection works on macOS, Linux, and Windows. Supports both VS Code Stable and Insiders. You can also pass --settings to point at any path or --target to force a specific edition.

References

• VS Code Custom Thinking Phrases
• GitHub Models
• GitHub REST API — Feeds
• GitHub REST API — Commits
• USGS Earthquake Catalog
• NOAA/NWS Alerts API
• Yahoo Finance

License

MIT