MIT AI Club Newsletter

AI for Sloanies is a newsletter built and run for members of the MIT Sloan AI & ML Club. Every Wednesday morning, subscribers receive a curated take on the week's most relevant AI news (big company drama, regulatory battles, early-stage startup funding), while tying the news to concepts and learnings in our Sloan MBA curriculum.

MVP+ features to be added in the future include: industry-specific selection, AI startup recruiting section, and an AI influencer spotlight.

Staying current on AI as an MBA student is imperative but given how quickly things change in AI, the barrage of classes, recruiting, and everything else, MBA students do not have the bandwidth to sift through Twitter and LinkedIn to find what actually matters. I wanted to find a solution to provide value to our club members.

The project is a two-stage pipeline deployed on Modal, a serverless cloud platform:

• Stage 1 (automated, every Wednesday 8am ET): Scrapes up to 30 articles from NewsAPI and five RSS feeds, runs three-layer deduplication (exact URL → title fuzzy match → content fingerprint), then makes five sequential Claude API calls to categorize, write, and assemble a fully styled HTML email. The newsletter is cached in Modal's persistent key-value store and a preview is sent to a reviewer.
• Stage 2 (human-in-the-loop approval): A styled HTML email is sent to the reviewer with Approve and Request Changes buttons. Approval triggers a POST-only endpoint that sends via Gmail SMTP to all subscribers via BCC. If changes are requested, the submitter submits freeform feedback, which is injected into the next generation run, up to 10 revision cycles per week.
• Tech Stack: Python, Anthropic Claude API (claude-sonnet), Modal (serverless + cron + FastAPI endpoints), NewsAPI, Gmail SMTP, feedparser.

The newsletter sends reliably each week with no manual intervention beyond a single reviewer click. The human-in-the-loop approval step keeps quality high while keeping operational overhead near zero. It's been a genuinely fun project to build and am looking forward to building on top of this first version.