Tom Lynch

@_adishj @ycombinator @motion_so Motion

introducing Motion, a video agent for tasteful motion design. this launch video was made entirely with Motion. 👇🏽 comment "MOTION" to get early access + free credits. tag @motion_so in any post on your X feed for a surprise. here’s how it works + examples (thread):

@HilaShmuel This is incredibly useful, for someone about to launch! Thank you 🙏

Hey everyone, I saw a lot of you saved my reply where I explained how I made yesterday’s promo video for runcabinet.com, so I thought it’s worth making a proper post about it. tools: Claude Code + Remotion + Eleven Labs (narrator) + Sumo (music) + sound effects + skills.sh + my brain 🧠 The foundation: Remotion (powerful framework for creating real videos with React, and inside you can import a lot) + Claude Code. 1. on Claude Code, copy that page and paste in your claude code. ask him to install and run the remotion web video editor on your browser: remotion.dev/docs/ai/claude… 2.on the the remotion web video editor, add skills. I've added a lot, almost anything relevant. 3. I searched for copywriting and scripts editing skills on , installed/pasted. 4. I told Claude to review my Cabinet website + the full Cabinet code so it would understand the project vision and messaging. I also asked Grok to summarise my style of posts and replies about Cabinet to get “my tone and style”. But in the end, I wrote the first few sentences myself :) 5. I asked claude: "create a script in a table for a promo 30-40 seconds video about cabinet. Have columns for: # Scene, Text on Scene, Visuals, Narration text. ". 6. After few iterations with claude, I've asked remotion to create the video. I've iterated a lot ("Change that text, move it these, let's create iMessage text bubble"). it was all fun process, but took 2+ hours. 7. When video was ready - let's handle Audio: "Please use Eleven labs to narrate this. suggest me few female young excited voices for product demo. Here is my API Key: ...". I did iterations on a few voices. Important tip: Eleven Labs has an option to generate tags that tell the narrator exactly how to deliver the line, so I told Claude to use those. more "please regenerate the voice on scene 7" but overall fast. using the eleven labs skills, It generated all the audio files by itself. Pure magic ✨ 8. "Please write a sumo prompt for a music for this video". I copied the Sumo song link(I requested “Apple keynote music” 😅) and pasted, and walla - we have music! 9. "using the remotion sound effects skill, add sound effects for this video, consider what we see on screen on each scene". 10. Render! I worked on the video for a few hours yesterday but I think it looks awesome 🤩 Next time it’ll run much faster for me 🚀 I think I will add "AI video generation studio" to Cabinet, might be cool to have system to generate this videos in the future easily. Anyone planning to try this combo? Let me know!

@dunkhippo33 Big parts of distribution can be autonomous too though - things like blog posts, tweets, etc

I tried Paperclip - an open source project that lets you set up an "autonomous company" with multiple AI agents. The concept is wild: you act as a board member setting vision while agents coordinate to build and run the company. More >>

What's causing this? Is it similar for other journals?

@hazhubble why is @wylansford eating dog food

OFFER: if you want to come to SF but budget is stopping you, i want to offer you FREE housing in our hacker house! we have the best views in SF, fancy chairs and monitors, 1gbps internet, and queen size beds waiting for you if you know someone who should be in SF tag them below DMs open

@PI010101 Huge achievement. I wonder if anyone will get Creating convexity...

Every subgaussian is a sum of Gaussians -- Antoine Song resolves Talagrand's conjectures arxiv.org/pdf/2602.22342

@arlanr how did you set this up? Want something similar for my saas

Introducing Agent Auth. Human onboarding is becoming obsolete. Now any agent can sign up for Nia and get access to 15+ tools to index and search across any data, including docs, PDFs, research papers, wikis, and more.

@LocasaleLab It's easier to pump your citations with high-volume self-referential papers than to do impactful scientific work. How do you solve this though? AI peer review?

There is something to consider here. Metrics for scientific merit like impact factor and citation indices have already shown how easily evaluation systems can be gamed once they become targets. At the same time, the volume of science now makes some form of structured assessment unavoidable. The traditional model of relying on a small number of deeply knowledgeable scientists has eroded as those individuals are marginalized within larger institutional systems. So some kind of scoring system may be necessary. The challenge is that any metric, once adopted, will be optimized and abused. The question is how to design them in a way that resists that.

@Hesamation More time spent doing Science than writing about it. Could be a good thing

the state of academia right now: → a bunch of code written by AI → become papers written by AI → to be evaluated by AI → so AI can summarize them and give us the important bits

@emollick Humans need HTML and agents need .md death to PDF

The fact that every scientific paper in 2026 is still uploaded only as fully formatted PDFs to academic archive sites that often limit downloads tells you everything you need to know about how quickly the scientific system is adjusting to the potential of AI to accelerate science

@emollick dropping a tool to fix this on Thursday

@yacineMTB more time doing science and less time writing about it, its a good thing

peer review is doomed AI reviewing AI written papers its over

@Faheem_uh 2026 will be the year of AI in Science the way 2025 was the year of AI in software engineering

I am ranked as the #1 voice in AI & Research in Australia. Here are my 5 predictions about AI & Research for 2026 1. The ones rejecting AI will also start using AI in research 2. AI detection tools will become even more inaccurate 3. The boundary between ethical vs unethical use of AI will become more clearer 4. Universities & journal will define more concrete policies for the use of AI 5. Methods and tools for catching unethical use of AI will be introduced I thank you all for supporting and engaging with my content. What is your one prediction for AI & research for 2026?

I've always liked this breakdown from @Nature on a high-quality abstract

@Faheem_uh I've always liked this guide from @Nature on how to write an abstract

How to write a journal paper for publication? Your paper should have the following 9 sections. 1. Abstract 2. Introduction 3. Literature Review 4. Methodology 5. Results 6. Discussion 7. Conclusion 8. Acknowledgements 9. References

@Faheem_uh cool concept. Is this quite similar to cursor? I'll give it a try anyway

How to create a scientific diagram in seconds? Follow these 3 simple steps. 1. Go to sciclaw.ai and log in 2. From the skills, select scientific schematics 3. Enter the requirements of your diagram. SciClaw will create a perfect diagram for you. You can download it as PNG. Try it today for FREE: sciclaw.ai

AI Slop research is a becoming a crisis. Partly because it’s an easy way to game the H-index. Peer review could fix this, but it’s not fast enough to keep up with the frontier. What’s the solution? AI peer review?

@yacineMTB obviously based, you’re in one

i still haven't decided if world models are a stinky meme or if they are based and cool as heck

@DavidHumme5859 @yacineMTB subtle genius answer

@yacineMTB Let’s do a sim with world models