• AI Exec
  • Posts
  • AI models that fine-tune themselves

AI models that fine-tune themselves

MIT and Anthropic researchers explain how

👋 Good morning/evening (wherever you are). It’s Friday.

Two papers I’ll be reading this weekend:

MIT’s SEAL and Anthropic’s ICM share a common goal: reducing reliance on human supervision.

But they differ in purpose, mechanism, and philosophy.

Here’s a simple breakdown:

  • Anthropic’s ICM teaches an AI to judge itself and learn from what makes sense. Think of it as grading its own homework without a teacher.

  • MIT’s SEAL teaches an AI to rewrite parts of itself to improve at new tasks. Imagine a robot editing its own brain code after learning something new.

Switching gears:

Missed my LinkedIn post on Monday? It’s worth checking out.

Jacob Adler’s Grand Prix–winning video essay Total Pixel Space (winner of AIFF 2025) is becoming a cult favorite.

OK let’s keep going ↓

Here’s what you should know:

The numbers:

  • Meta finalized a $14.3B investment in Scale AI, valuing it at $29B.

  • Clay raised Series C at a $3B valuation, led by CapitalG.

  • Fireflies hit $1B via tender offer and launched Talk to Fireflies with Perplexity.

  • Canary raised $80M to scale hospitality AI, now valued at $600M.

  • Tastewise raised $50M Series B for GenAI in food and beverage, led by TELUS.

  • Landbase raised $30M Series A to build AI sales tools, backed by Sound and Picus.

  • Bolo AI raised an $8.1M seed round for industrial AI, led by True Ventures.

  • Weavy raised a $4M seed round for AI image and video editing, led by Entree.

Thought starters:

Meme of the day:

Thanks for reading,

Eddie

Are we friends on LinkedIn and X? Let’s connect — DM me or hit the follow button.

P.S. If this was valuable, forward it to a friend. If you’re that smart friend, subscribe here.

P.P.S. Interested in reaching other ambitious readers like you? To become a sponsor, reply to this email.