What is Latent.Space?
A technical newsletter, podcast (top 10 in US Tech! in audio podcast and YouTube video formats), and community (VERY active Discord online and in person) by and for the rising class of AI Engineers, focusing on the business and technology of this new AI summer.
Quotes
“Great episode, good technical discussion on LLM pretraining 👍” - Andrej Karpathy
“In numbers, there's probably going to be significantly more AI Engineers than there are ML engineers / LLM engineers. One can be quite successful in this role without ever training anything.” - Andrej Karpathy
“This was a wonderful chat w Alessio and @swyx - diving much deeper into the AI infra + compiler + language + implementation details of the Modular AI Engine than ever before.” - Chris Lattner
“A good intro to the tiny corp for people who want to come work here.” - George Hotz (+ Soumith Chintala)
“Probably the highest-leverage 45 mins I spend everyday on catching up with what's going on in AI. So much alpha, organized hierarchically; an exceptionally well curated (and summarized) daily newslettter.” - Soumith Chintala
“This is encouraging for builders like me who focus on solving user problems and the production side of things.” - Eugene Yan
“Probably the most comprehensive overview of where we’ve been, where we’re going, and what’s on our minds at LangChain” - Harrison Chase
“Your pods are a legitimate highlight of life for me. They’re amazing.” - Mckay Wrigley
“This is spot on. Multiverse is the real metaverse.“ - Emad Mostaque
“A very useful summary of the state of play in open source generative AI, including a bunch of projects I hadn't heard about before” - Simon Willison
“I was driving, laughed so hard I almost had to pull over.” - Gene Kim
“I listen to a ton of AI podcasts and by far the best are where the host is a technical expert rather than a journalist or other generalist. Deeper, smarter, more in tune with their audience. The gold standards are @latentspacepod, @CogRev_Podcast and @NoPriorsPod IMO” — Chris Anderson
“Best introduction to LLM benchmarks…It really shows that you always do your homework deeply before addressing any topic, which provides a lot of trust to your listeners. Thank you for that!” — Samuel Path
“I have like 20 tabs open with links now” - Kevin Hu
“Extremely high signal to noise ratio, pretty much the only resource I follow devoutly. Every episode is gold.” - Francisco Pingham
“The All-In podcast equivalent for AI” - Atlantis Labs
“Highest-signal pod” - AI Safety Memes
“Some terrifyingly and delightfully sharp insights here by @swyx into generative AI, capitalism, and how the multiverse will come to our rescue as the metaverse encroaches.” - gabor ugray
"But you cannot own a Multiverse. Whatever virtual land you “own”, my next favorite one is one seed away in a neighboring universe." - Jim OShaughnessy
“Terrific interview” - Eric Ries
“I listen to many podcasts. This has been one of the most interesting conversations I've heard on language models.” - Ben Dickson
What We Cover
Our pet topics:
Fundamentals
Intros: Benchmarks 101, Datasets 101, RLHF 201
Model/Training: FlashAttention, Transformers Math
Executive Summaries: ChatGPT, GPT4, Auto-GPT, GPT4.5, LangChain
AI Engineers - The Rise of the AI Engineer, You are not too old
stay tuned for Latent Space University
AI Engineer Themes
AI UX - It’s Time To Build AI UX, How to Make AI UX your Moat, Retool
Multimodal AI - IDEFICS (HuggingFace M4), TLDraw MakeItReal, Suno AI
AI Agents - The Anatomy of Autonomy, Codium/AutoGPT, Imbue (Kanjun Qiu)
AI Coding Tools - Cursor, Phind, Codeium, Sourcegraph Cody
AI LLM Tooling - Langchain, LlamaIndex, Guardrails, Copilot for X
AI Infra/Compilers/Inference Hardware - MLC.ai, Modular (Chris Lattner), Pytorch (Soumith Chintala), Replicate, Modal, Together
Actually Open AI and Finetuning - How Open Source is Eating AI, Replit, Mosaic, Dolly v1/v2, Llama 2, Fast.ai (Jeremy Howard), Together
AI Startups and Builders - Adept, Lexica, Glean, Roboflow, Seek.ai, Cube.dev, Notion, Amplitude, tiny (George Hotz), Playground (Suhail Doshi)
Recaps
Breaking News: ChatGPT, GPT4, Stable Diffusion 2, Functions, Code Interpreter
Event Recaps: OpenAI DevDay, NeurIPS 2023 Papers and Startups, Latent Space Anniversary
Monthly Recaps: Aug, Sept, Oct, Nov, Dec 2023, Jan, Feb 2024
Misc: Reverse Prompt Engineering, Academia, AI Product/Moats/Eigenquestions
Other AI Podcasts: AI Breakdown, AI Breakdown 2 (part 1 and part 2), Practical AI, Cognitive Revolution, ThursdAI, Thursday Nights in AI, ChinaTalk
Areas we de-emphasize and rarely discuss (bc undifferentiated, covered elsewhere):
Law, Regulation, Politics, Safety and X-risk (check out our friends at PracticalAI, CogRev, and AI Breakdown for that)
Defense/Biotech/Medical/Crypto/High Frequency Trading AI
BigTech power plays (e.g. Google vs Microsoft - tho we covered the No Moat memo)
Aliens
Many of our guests and topics are now also featured at the AI Engineer Summit.
Notes for PR & Media Agencies: We’re interested to work with you ahead of major funding and product announcements (give ~1 month lead time), but we do not accept cold emails. We’ll take a look if you have a warm intro from someone we already know.
Components
The newsletter was first started by swyx in 2022, whose regular blog is more career and engineering focused, and DXTips is more DevRel focused.
The podcast (RSS feed here) was then started in 2023, cohosted by Alessio and swyx, recording mostly in-person with technical deep dives and a real look at the people behind the major projects and news in AI. (see our guest prompt here! and our new YouTube channel). Our podcast intro music is “Ice & Fire - King Canyon”.
Both projects grew out of an informal Discord community that has run since Covid. The community now has an active LLM Paper Club that has unrecorded Zoom meetings every week (join here) covering one important paper or reading in AI.
Origin and other naming notes - from swyx
In 2022 I was undecided on how seriously I wanted to pursue AI, so decided to carve out a new space, and originally hosted it on lspace.swyx.io. With the success of the newsletter and podcast, I realized I was not too old and found I could ramp it up by moving to the latent.space domain thanks to a friendly offer from Brian McMahan, coauthor of Natural Language Processing with PyTorch.
This blog was originally named “L-Space Diaries”, which nominally stands for Latent Space, but also it is a reference to the L-space of Discworld, by Sir Terry Pratchett (GNU).
