name: ai-news
description: Check AI news sources for recent developments. Use when user says "ai news", "check ai news", "what's new in AI", or wants to catch up on AI developments.

AI News Check

Fetch recent AI news from curated sources and summarize key developments. Quick scan for staying current without deep research.

Trigger

Use when:
- User says "ai news", "check ai news", "what's new in AI"
- User wants to catch up on AI developments
- Interview prep needs talking points

Default Lens: Capco Principal Consultant

Terry is a Principal Consultant, AI Solution Lead at Capco. Frame AI news through this lens:
- Governance & regulatory: What are banks struggling to comply with?
- Client advisory angle: What would you tell a bank CRO/CTO?
- Competitive landscape: What are other consultancies saying?
- Gaps: Where is adoption outpacing oversight?

Always include governance-focused web searches alongside regular sources:
- "AI governance financial services banking [year]"
- "HKMA GenAI regulation" (for HK context)
- "site:capco.com AI" (know your employer's positions)

Inputs

• mode (optional): "quick" (Tier 1 only) | "deep" (all sources) | default is quick
• smol_deep (optional): Fetch last 3-5 full Smol AI issues for deep analysis

Sources

Sources are defined in sources.yaml in this skill directory. Categories:
- Web sources: Blogs, newsletters (WebFetch)
- Bank tech blogs: Financial services AI (WebFetch)
- Chinese sources: 机器之心, 量子位, 新智元 (RSS)
- X accounts: @karpathy, @emollick, etc. (via bird CLI)
- YouTube: Video search + transcript extraction

Tier 1: Always fetch (~15 sources, fast)
Tier 2: Deep scan only (~50+ sources, comprehensive)

Workflow

Quick Mode (default)

1. Load sources from sources.yaml, filter to Tier 1

2. Fetch web sources in parallel (WebFetch):

3. Smol AI News, Simon Willison, Eugene Yan
4. Anthropic Blog, OpenAI Developer Blog
5. Layer 6, Monzo, Plaid, Sardine, Chainalysis

6. 机器之心, 量子位 (RSS)

7. Extract & summarize:

8. Last 3-5 articles from each source
9. Title, date, one-line summary

10. Highlight relevance to AI/ML in banking

11. Output to chat + append to [[AI News Log]] in vault

Deep Mode (user says "deep", "full", "all sources")

1. Fetch all Tier 1 + Tier 2 sources

2. Fetch X accounts (via bird CLI — portable):
   bash bird user-tweets karpathy -n 5 --cookie-source chrome bird user-tweets emollick -n 5 --cookie-source chrome

3. All accounts from both tiers
4. Skip if user says "quick"

5. See /x-twitter skill for full CLI reference

6. YouTube videos (Brave Video Search):

7. Search AI-related queries from sources.yaml
8. Pick 2-3 most relevant videos

9. Extract transcripts using youtube-transcript skill

10. WeChat scan (Chinese practitioner content):

11. WebSearch with site:mp.weixin.qq.com/s/ queries from sources.yaml
12. Filter to short URLs only (format: /s/ABC123)
13. Extract via wechat.imagenie.us/extract API
14. Limit to 5 articles to avoid rate limits

15. See wechat-article skill for full extraction details

16. Apply stale filtering: Skip sources with no articles <30 days old

Smol AI Deep Read

When user says "smol deep" or "full smol":
- Fetch last 3-5 full issues from news.smol.ai/issues/YY-MM-DD-slug
- Full issues contain: swyx's analysis, community insights, technical highlights, smaller stories

Error Handling

• If site unreachable: Report error, continue with other sources
• If paywall/login required: Report and skip
• If too much content: Limit to 5 most recent per source
• If source stale (>30 days): Note for future pruning

Output

## Smol AI News (swyx)
- **[Title]** (Date) — One-line summary

## Simon Willison
- **[Title]** (Date) — One-line summary

## Bank Tech Blogs
### Layer 6 (TD Bank)
- **[Title]** (Date) — One-line summary

## Chinese AI News (中文)
### 机器之心 (Synced)
- **[Title]** (Date) — One-line summary

## WeChat Articles (微信公众号) [Deep mode only]
- **[Title]** — One-line summary
- **[Title]** — One-line summary

## X Highlights
- **@karpathy:** [Recent post summary]

---
*Notable for your context:* [Relevance to AI/ML in banking, interview prep]

Save to: Append to [[AI News Log]] with date header

Files

• Sources config: sources.yaml (in this skill directory)
• Related: /x-twitter skill for X account fetching