weekly-ai-recap

name	weekly-ai-recap
description	Research, write, and publish the weekly "What happened in AI this week" article for Lablab. Run every Monday — searches for the past week's top AI news, writes a skimmable recap article, and produces a social post brief for Gosia/Soto. Updates the Notion task on completion.
compatibility	Designed for Claude Code. Requires Exa MCP or Brave Search MCP (WebSearch fallback), Cloudinary for images, and Notion token in ~/claude-workspace-automation/.env (NOTION_TOKEN).

Steps

1. Confirm the publication date

Ask the user (if not already clear):

• What Monday is this for? (default: today's date)
• Any specific stories, launches, or topics to definitely include?

Derive:

• WEEK_END = publication date (the Monday)
• WEEK_START = 7 days prior (the previous Monday)
• SLUG = this-week-in-ai-YYYY-MM-DD (using the publication date)
• FILE_PATH = blog/en/this-week-in-ai-YYYY-MM-DD.mdx
• BRANCH = add-this-week-in-ai-YYYY-MM-DD

2. Create a feature branch

Per the git workflow in CLAUDE.md, all new content goes on a feature branch:

git checkout -b add-this-week-in-ai-YYYY-MM-DD

Never write directly to main.

3. Check Notion for the task

Fetch the matching task from DB 2cab4088-66ca-4d1f-aeb9-8fe29dafb470:

• Use the workspace automation token from ~/claude-workspace-automation/.env (NOTION_TOKEN)
• Filter: Workspace = Lablab, Category = Content, title contains "AI News Recap" or "AI recap"
• If found: read the Notes field for any pre-captured story ideas or instructions
• If not found: create a task using the notion-task global skill before continuing

4. Research — find the top AI news of the week

Search for major AI news from WEEK_START to WEEK_END. Run all available search tools in parallel:

Priority queries (always run):

• "AI news this week" [WEEK_END date]
• new AI model release [month year]
• AI breakthrough announcement [month year]
• major AI company update [month year]
• AI funding round [month year]

Supplementary queries (run if time allows):

• OpenAI GPT announcement [month year]
• Anthropic Claude update [month year]
• Google Gemini news [month year]
• Meta AI [month year]
• open source AI model release [month year]

Curate: Select the 5–8 most significant stories. Prioritise:

1. New model releases or major version updates
2. Big-company announcements (OpenAI, Anthropic, Google, Meta, xAI, Mistral)
3. Genuine AI breakthroughs or research with real-world implications
4. Significant funding / acquisitions
5. Notable open-source releases

Discard: minor blog posts, opinion pieces, redundant coverage of the same story.

5. Write the article

Write to blog/en/this-week-in-ai-YYYY-MM-DD.mdx. The file must start with MDX frontmatter:

---
title: "[Story 1], [Story 2], [Story 3]"
description: "[2-sentence SEO summary naming key companies and the month/year. No hype words.]"
image: "https://res.cloudinary.com/dygkv9gam/image/upload/v1/lablab-static/this-week-in-ai-cover.jpg"
authorUsername: "stevekimoi"
---

Then the article body:

# [Story 1], [Story 2], [Story 3]

*This Week in AI — [Month Day–Day, Year]*

[2–3 sentence thematic intro. Capture the mood and the thread connecting this week's stories —
not a list of headlines, but a single observation about what the week meant. No "In this article..." openers.]

## Key Takeaways

- **[Company/topic]:** [1–2 sentences: what happened + the direct implication for builders or
  the industry. Write it as if explaining to a smart friend, not a press release.]
- **[Company/topic]:** [same]
- **[Company/topic]:** [same]
- [3–5 bullet points total — only the stories with a clear "so what"]

---

## [Thematic Section Title — groups 2–3 related stories, e.g. "The Open-Source Race" or "OpenAI vs. Everyone"]

### [Story Headline]
[2–3 sentences: what happened, why it matters, any concrete numbers. Source linked inline on
the most important claim. End with the implication — one sentence, no hedging.]

### [Story Headline]
[same]

---

## [Next Thematic Section]

### [Story Headline]
[same format]

### [Story Headline]
[same format]

---

## Quick Hits

- **[Company/topic]:** [One sentence. Source linked.]
- **[Company/topic]:** [One sentence. Source linked.]
- [4–6 bullet points for smaller stories that don't need a full breakdown]

---

*This Week in AI is published every Monday by the [Lablab](https://lablab.ai) team.*

Writing rules:

• Intro: 2–3 sentences that name the theme of the week, not a summary of every story. The reader should feel oriented before they've read a single headline.
• Key Takeaways: Implications, not just facts. Each bullet ends with what a developer or founder should do or think differently as a result.
• Section titles: Group stories by theme (e.g. "Meta Goes Closed-Source", "AI Meets the Real World") — never just "News" or "Updates"
• Story write-ups: 2–3 sentences, one inline source link on the key claim, end with the implication. No hanging "time will tell" sentences.
• Quick Hits: Stories that matter but don't need context — funding rounds, minor releases, policy moves, notable tools
• Voice: Direct and confident, like a smart colleague briefing you before a meeting. Not a journalist, not a hype merchant. Casual enough to be readable, precise enough to be trusted.
• Banned words: "revolutionary", "game-changing", "powerful", "cutting-edge", "groundbreaking", "exciting", "innovative"
• Every factual claim has a source — linked inline. No hallucinated stats; if a number can't be verified, omit it.
• Total word count: 700–1,100 words. Skimmable but substantive.
• SEO: The intro and Key Takeaways naturally include the major company names and "AI news [month year]" — do not force it, let coverage drive it.

6. Produce the social post brief

After the article is drafted, create a separate brief for Gosia and Soto at the bottom of the article file under a --- separator, or as a note to the user:

SOCIAL POST BRIEF — [publication date]

Format: Carousel (Instagram/LinkedIn)
Post on: Same day as article publication

Slide 1 — Cover: "What happened in AI this week 🧠" + date
Slide 2–[N] — One story per slide: bold headline + 1-sentence summary
Last slide — CTA: "Full breakdown at lablab.ai → link in bio"

Stories to cover (pick top 4–5 for carousel):
1. [Story 1 headline]
2. [Story 2 headline]
3. [Story 3 headline]
4. [Story 4 headline]
5. [Story 5 headline]

Tone: Clean, minimal, no buzzwords. Reference: @evolving.ai on Instagram.
Template: Soto to create reusable carousel template for this series.

7. Run publish-check

Before marking done, run the publish-check skill on the draft:

• Confirm no placeholders, no skeleton sections, no unverified stats
• Confirm MDX frontmatter is complete (title, description, image, authorUsername)
• Confirm word count is 700–1,100 words
• Confirm all source links are present

Fix any issues before continuing.

8. Update Notion

Once publish-check passes:

• Set Notion task Status → In Progress (article written, pending publication)
• Append to Notes: Draft written. File: [FILE_PATH]. Branch: [BRANCH]. Social brief included. Date: [today].

9. Hand off

Confirm to the user:

• Feature branch created: [BRANCH]
• Article draft written at [FILE_PATH]
• MDX frontmatter complete
• Social post brief produced
• Notion task updated to In Progress
• Publish-check passed

Next steps for the user:

1. Review the article and edit as needed
2. Share the social brief with Gosia + Soto
3. Open a PR from [BRANCH] → main targeting the lablab-ai/community-content upstream
4. After PR merges and article is live, run publish-check again and set Notion task → Done
5. Gosia posts the carousel same day

Series tracking

This skill is designed to repeat every Monday for the 5-week experiment (Apr 14 – May 12, 2026). After 5 editions, review performance data before committing to ongoing publication.