Eric Rachlin
@eerac
Yo, I’m a scientist! NYC➡️CS PhD➡️Body Labs cofounder (acq. by AMZN)➡️BK dad➡️principal scientist➡️corporate dropout➡️http://withcount.com
Human language evolved ~100,000 years ago Written language is much more recent, only about 5,500 years old Written language was the catalyst for global civilization. Its inception changed the planet The impact of software being able to understand language is hard to overstate
I don't know where AI art is headed, but one thing I'm sure of is that there's going to be a lot more graphic novels out there. My 9 year old already has access to (and reads) way more comic-style books than I did as a kid. Demand is high, and cost is dropping fast.
# of meals I ate today in Brooklyn that ended with a soufflé pancake: 2
This deserves more attention. Even forgetting about the sharing feature, I’ve often wanted Claude/GPT to loop over data (which it can easily write code to do) but then process each entry with AI. Previously this meant writing an app, but now you can do it directly in Claude.
Introducing two new ways to create with Claude: A dedicated space for building, hosting, and sharing artifacts, and the ability to embed AI capabilities directly into your creations.
Huge UI improvement! Runway is getting much closer to being able to go from a script + assets (e.g. ref images) to sequence of clips that get edited together to make a movie. I’m expecting something like “Cursor for video” next year, but UI will need to be pretty different.
Introducing Chat Mode. A new way to create with Gen-4 Images, Videos and References. Now you can generate anything you want, all from within a single conversational interface. Available for all users.
3D. My preferred seat when traveling to #CVPR2025 DM if you’re there and want to meetup.
Now is a great time to take your favorite simple app and make an "AI first" version that both people and AI users can control. Start simple, a weather or stock app perhaps, then move to something fancier like a podcast player. Why will we want podcast players AI can’t control?
Products with extensive/rich UIs lots of sliders, switches, menus, with no scripting support, and built on opaque, custom, binary formats are ngmi in the era of heavy human+AI collaboration. If an LLM can't read the underlying representations and manipulate them and all of the…
Had a great time yesterday hosting 10-year old’s party at McCarren park. Fun times, super chill.
Been using ChatGPT desktop app with o3 connected to Cursor. It kind of sucks, but if I were Cursor I'd still be worried. Right now it can only see/edit one file at a time, and the edits aren't super reliable, but what about 6 months from now? I'm expecting rapid improvement.
Even if you aren't all in on AI, it's hard to not see the utility of this. Even within a completely traditional workflow, if you need footage for flashback, or perhaps an additional shot you wish you'd taken, this seems like a great option.
Today we are releasing Gen-4 References to all paid plans. Now anyone can generate consistent characters, locations and more. With References, you can use photos, generated images, 3D models or selfies to place yourself or others into any scene you can imagine. More examples…
Went to Just Salad today for lunch instead of my usual Sweetgreens. Wild times. Never felt more alive.
Has anyone tried this, seems incredibly appealing given that normally this type of user research takes a huge amount of time and effort. Also intriguing that they could use AI to create profiles of their users and then simulate them. People obviously wouldn't pay as much for…
AI writes your code. Now it talks to your users. We raised $27M from @Sequoia to build @ListenLabs. Listen runs thousands of interviews to uncover what users want, why they churn, and what makes them convert. See how @Microsoft and @canva use it:
“Add excitement to any game that requires dice” As both a founder and consumer, I appreciate this cleared-headed assessment of the products value prop.
Non-professionals get excited when AI demos show them how they can "turn an idea into X", but professionals don't generally want this. They want to more easily create the asset libraries and technical specifications needed to create high quality final outputs. AI's new found…
Mood boards aren’t world building & storyboards aren’t blocking. They are the first “sketching” step that gives you the right shape of things so your world building or blocking are more efficient. The spacial management of assets is where the actual world building & blocking is.
This is a good point and becomes ever more apparent in the world of LLM-powered coding. More and more command for a way in which a tool like Cursor can "handle everything in one place" (even a single file in many cases!)
The reality of building web apps in 2025 is that it's a bit like assembling IKEA furniture. There's no "full-stack" product with batteries included, you have to piece together and configure many individual services: - frontend / backend (e.g. React, Next.js, APIs) - hosting…