⚡ Behind Latent Space: A Rare Look Inside the Pod | AIEi Shanghai

The AI engineer title that built a conference

Shawn Wang turned a blog about Stable Diffusion into Latent Space and the AI Engineer conference by naming a discipline and building a community around it. The job title stuck even as hot topics moved from RAG to coding agents, and AIE grew into a three-day World's Fair with more than 7,000 attendees. He now applies that lens to software, arguing mid-tier SaaS is vulnerable because its network effects are thin, while SMBs are in a golden age. His worries are an AI winter and safety.

Takeaways

• AIE grew into a 7,000-person World's Fair. The first AIE summit was three years earlier. The current three-day event had 30 tracks, beat its 6,000 target with more than 7,000 attendees and 1.5m online viewers, and transcribes every talk. Wang wants to build the Wikipedia of AI, with datasets at ai.engineer/data.
• Naming the AI engineer built a durable discipline. Wang named the AI engineer discipline, and the title held up as hot topics moved from RAG to coding agents. He says if you give people a job title that pays money, they will be loyal to you.
• Mid-tier SaaS is vulnerable while Salesforce is safe. His conference software budget is about $300,000 per event, roughly 10% of a $3m to $10m budget. He offered $10,000 to clone a $40,000 product and got 69 submissions. He now runs his own Calendly, Drive, GitHub, Dropbox, Slack and DocuSign.
• Learning in public lets you pick your boss. Wang argues finance is zero-sum while tech is positive-sum, so telling someone something increases its value. He says learning in public lets you choose your own boss, and that he advises technical founders to involve marketing earlier so go-to-market is built into the product.
• Chinese founders excel at tech, need marketing help. A Chinese founder breaking into the US is good at tech and bad at marketing, so should talk to someone who is good at it. Wang points to 996, ComfyUI, Canva and fal AI, and says Chinese founders would kill Amazon and Facebook.
• Continual learning and world models excite Wang. Continual learning, where the model learns as you work with it rather than through pre-training and post-training, is important. He names interaction models and world models as what excites him. His worries are an AI winter and safety, with AI doom as his default assumption.

Wang's path to AIE

1. Spent eight years in trading and investment banking, including hedge funds, before moving to tech at Amazon and startups.
2. Went independent four years ago to write Latent Space and start AIE.
3. Started the blog when Stable Diffusion came out, before ChatGPT. Emad Mostaque read his analysis and said it was good.