Long-form, technical AI analysis
I’m a researcher training state-of-the-art language models at the Allen Institute for AI. This blog shares the insights I have. It is mostly post-training and open-source AI, but I cover all the major events too.
What you get:
Weekly essays on the specific aspects of current AI training, deployment, systems, or impacts. These are normally Wednesday mornings at 5:00AM PST (and are available in audio form).
Regular interviews with leading AI researchers. Normally these are Thursdays, but there’s some variance in it.
For the full audio feed (with voiceovers) subscribe where-ever you get your podcasts.
For a feed with interviews only, check out YouTube or this secondary Spotify feed.
[Paid only] Monthly roundups of the latest open models, datasets, and interesting resources for keeping up with AI.
What this often contains:
I write out the feelings of understanding a new research idea in AI,
Follow latest models and what they mean (not the marketing), and
More direct access to popular and hidden gems of technical researchers.
Praise for Interconnects
Here are some highlights of folks describing Interconnects in their own words:
From Jordan Schneider, ChinaTalk on Twitter.
nathan i've given up on reading papers now you're my only hope
From Christopher Manning, renowned computer science professor at Stanford and Sasha Rush at Cornell:
From Matthew Lynley of Supervised:
Nathan Lambert, an expert currently working in the field, synthesizes some of the most complicated new developments in AI in an incredibly relatable and entertaining way.
From many others:
The best AI Substack right now.
And to be clear, we’re just getting started.
There are not enough people trying to explain complicated topics in AI research and how it integrates with where technology and society is going. Interconnects is trying to pioneer a new method of AI communication.
Paid subscribers get:
The ability to always comment on posts, which is the place I respond to the most over email, Twitter DMs, etc.
Access to the subscriber Discord, so you can know what I’m thinking about and access ML experts. You can even steer what I write about.
Monthly roundups of the top models, Artifacts logs.
The ~10% of my essays that are paywalled explicitly.
The joy of supporting my work and encouraging me to take risks, such as the new audio experiment.
(I have paid upsell on in the Substack settings, so sometimes you may get an extra email — my most engaged people are the people I want to hang out with in the Interconnects Discord)
Upgrading to paid supports my memberships and API bills for understanding ML and creating this content. The growing list of subscriptions I use (or have used) to create Interconnects includes ChatGPT+, OpenAI API, MidJourney, Runway.ml, ElevenLabs, Transistor.fm, and more.
Contact
mail at interconnects.ai if you have any questions!
I love hearing from people and uplifting those in my community, so please comment, join my chat (you can ask questions too), and Tweet my articles with a question or more.
If you want to propose a guest post, please email me a completed draft.
If you’re an author of a recent ML paper, I’d love to hear from you and help you tell the story of your paper.
Student discount
I’m offering 80% off paid plans for students, to redeem this use an educational email and go to interconnects.ai/student
No LLMs used to write this
I only use LLMs for editing in the form of Grammarly for typo detection and a secondary full pass of Claude and/or ChatGPT (which does normally find interesting typos). LLMs are not used for prose or diagrams; just occasionally for title brainstorming.
Disclosures
I was a paid advisor to Tola Capital in 2024.
As part of paying for Eleven Labs API to generate my voiceovers, I enrolled in their referral program, which currently nets me $0 (try it here).
Otherwise, I have accepted no advertising dollars and pay for the services I use.
Interconnects AI is a newsletter owned and operated by Interconnects AI, LLC.
Thank you for your support!
