LLM Training Gets Its Definitive Guide

The highlight of the day comes from Hugging Face researcher elie (@eliebakouch), who shared what might be the most comprehensive practical guide to LLM training yet released:

"Training LLMs end to end is hard. Very excited to share our new blog (book?) that cover the full pipeline: pre-training, post-training and infra. 200+ pages of what worked, what didn't, and how to make it run reliably"

The "Smol Training Playbook" covers the complete pipeline—pre-training, post-training, and infrastructure—with an emphasis on practical lessons learned. At 200+ pages, this represents a significant knowledge dump from practitioners who have actually built these systems. The focus on "what didn't work" is particularly valuable; most resources only showcase successes while hiding the failures that often teach more.

For anyone looking to train their own models, this is now required reading.

AI Film Pushes Creative Boundaries

The ongoing debate about AI-generated content as legitimate art continues. Charles Curran (@charliebcurran) threw down a challenge to skeptics:

"If you think AI film can't be art than explain this."

While the specific film wasn't detailed in the post, the provocation highlights the shifting conversation around AI creativity. We've moved past "can AI make art?" to "what makes AI-generated content compelling?"

Crypto Whales Make Big Bets

In the crypto world, the famous "100% win-rate whale" continues to attract attention with massive leveraged positions:

"100% win-rate whale again opens $BTC 13x and $ETH 10x long positions and increases $SOL 10x. Overall, all positions now valued at $275 million with floating loss of $1.6 million."

A $275 million position with only $1.6 million in floating losses demonstrates the confidence (or perhaps recklessness) of large players in the current market.

Key Takeaways

• For ML practitioners: The Smol Training Playbook from Hugging Face is a must-read for anyone serious about training language models
• For the AI curious: The AI art debate continues to evolve as tools improve and creators push boundaries
• For market watchers: Big money continues to flow into crypto with significant leverage