Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

I increasingly think future software will combine code for control flow with small neural programs for "fuzzy" judgments. That's what I've been exploring with ProgramAsWeights, which answers the question of where those neural programs come from: they are "compiled" from English descriptions. For example, I combined 30 neural programs... show more

Yuntian Deng

10,675 subscribers

13,797 views • 8 days ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

excited to announce that I'm officially partnering with the folks over at Essential Mod ! The Essential team is starting up a new program aimed at supporting mod creators specifically. I've been working with them behind the scenes for a while now to make this program a reality, so I'm very happy to see it come to fruition. I'd really like to see modding become a viable career option for more people, and I think creator programs like this are a step in the right direction. also we'll be collaborating on some cool content in the future so stay tuned!

excited to announce that I'm officially partnering with the folks over at Essential Mod ! The Essential team is starting up a new program aimed at supporting mod creators specifically. I've been working with them behind the scenes for a while now to make this program a reality, so I'm very happy to see it come to fruition. I'd really like to see modding become a viable career option for more people, and I think creator programs like this are a step in the right direction. also we'll be collaborating on some cool content in the future so stay tuned!

YUNG

30,308 views • 1 year ago

Experimenting with tldraw, xyflow · React Flow · Svelte Flow & Mermaid JS as visual ways to interact with a code base for interactive modern code review. Claude & Codex can generate so much code so quickly that we are going to need new ways to review and understand their work. Markdown documents, ERDs, tree-maps, flow-charts, systems architecture diagrams, process maps. These could be the new lingua franca of a higher level abstraction of software development. What you're seeing below is a visualisation of the database structure of a Laravel app, automatically generated in a visual code review tool to help with the review of a new feature.

Experimenting with tldraw, xyflow · React Flow · Svelte Flow & Mermaid JS as visual ways to interact with a code base for interactive modern code review. Claude & Codex can generate so much code so quickly that we are going to need new ways to review and understand their work. Markdown documents, ERDs, tree-maps, flow-charts, systems architecture diagrams, process maps. These could be the new lingua franca of a higher level abstraction of software development. What you're seeing below is a visualisation of the database structure of a Laravel app, automatically generated in a visual code review tool to help with the review of a new feature.

Peter Thomson

27,048 views • 8 months ago

So... I literally oneshotted this code with Claude Sonnet 3.7 for replicating the exact same browser sandbox runtime that Manus uses. And I am going to open-source it, welcome contributions for building out the React VNC client, integrating to browser use, agent loop, etc. But we need a name first, should we call it... - Autonomous Neural Universal System - Magnus - or ?

So... I literally oneshotted this code with Claude Sonnet 3.7 for replicating the exact same browser sandbox runtime that Manus uses. And I am going to open-source it, welcome contributions for building out the React VNC client, integrating to browser use, agent loop, etc. But we need a name first, should we call it... - Autonomous Neural Universal System - Magnus - or ?

jian

419,367 views • 1 year ago

most AI chatbots break when you ask a question that requires info from multiple sources for example try asking: “which client contracts are finishing up this month?” you’ll get a half-answer — or none at all why? because traditional chatbots only look at small snippets of your docs - they don’t understand how things connect across clients, services, timelines that’s where knowledge graphs come in they let you turn messy contracts into a web of relationships — like: "Client → Contract Type → Service Provided → End Date" so instead of guessing from a few chunks of text, your chatbot can search across all your clients and contracts to give accurate answers I made a full walkthrough on how I built this: – how to organize your contracts so an AI can actually use them – how to define what matters (like who signed what, and when) – how to get the AI to figure out what info it needs and where to find it – and how to feed that back into your chatbot so it gives accurate answers reply “graph” and I’ll DM it to you (must be following)

most AI chatbots break when you ask a question that requires info from multiple sources for example try asking: “which client contracts are finishing up this month?” you’ll get a half-answer — or none at all why? because traditional chatbots only look at small snippets of your docs - they don’t understand how things connect across clients, services, timelines that’s where knowledge graphs come in they let you turn messy contracts into a web of relationships — like: "Client → Contract Type → Service Provided → End Date" so instead of guessing from a few chunks of text, your chatbot can search across all your clients and contracts to give accurate answers I made a full walkthrough on how I built this: – how to organize your contracts so an AI can actually use them – how to define what matters (like who signed what, and when) – how to get the AI to figure out what info it needs and where to find it – and how to feed that back into your chatbot so it gives accurate answers reply “graph” and I’ll DM it to you (must be following)

Tyler

24,994 views • 1 year ago

004/100 Buttons. A bit of the process on building an animation. When looking at a finished animation or in this example a finished button, it can look quite complex inside the CSS. But when building it, it’s more like a lot of simple steps, one after another. Here I had the idea to make some kind of text animation like the footer logo on the Osmo site. I try to add the base animation with no complex easing, for example transition: translate 0.4s ease. Starting with just moving the one text from bottom to top and the other text to top. Adding a stagger, play around with it. Searching for a way to make it more circular. On the research I found the sin() function inside CSS which can build a more smooth non linear curve for the stagger which creates this circular effect. And step by step adding more complexity like, different easing for hover/hover-out, opacity, 3D transform and more. I use also the sin() function to rotate the letters, so the middle ones are getting more rotated than the outer ones. Another thing which helps is to add a small delay on hover, for example 0.05s or 0.1s, you don’t really see the difference, but when you hover pretty fast on and out it doesn’t get that jumpy. I’m using here GSAP’s SplitText to split every char into spans. And then I’m adding a CSS index variable to every span, starting from the center. SplitText can provide CSS index variables, but you cannot tell it from which direction. For the sin() it’s also important to have a max length, so I add another CSS variable with the max char number on it. Crafting 100 Buttons with Osmo ⏳ Total time: 63h

Eduard Bodak

166,023 views • 4 months ago

For years i've had this vision for a synchronous reactive ORM with offline support and realtime syncing to other clients. This is the holy grail of state management IMO and i'm finally building it using mobx and yjs. I know that's a mouthful so let me explain: synchronous - there's no promises to deal with and therefore no loading states which leads to much cleaner frontend code. reactive - imagine signals/observables that you can mutate and they will trigger your components to rerender and also will immediately propagate to other users' devices. ORM - what i've made is not an ORM, but it looks and feels like one where you define Models and then you interact with instances of those models. offline/syncing - this is done through yjs, a CRDT implementation, and websockets. CRDTs power many modern collaborative multiplayer tools like figma.

For years i've had this vision for a synchronous reactive ORM with offline support and realtime syncing to other clients. This is the holy grail of state management IMO and i'm finally building it using mobx and yjs. I know that's a mouthful so let me explain: synchronous - there's no promises to deal with and therefore no loading states which leads to much cleaner frontend code. reactive - imagine signals/observables that you can mutate and they will trigger your components to rerender and also will immediately propagate to other users' devices. ORM - what i've made is not an ORM, but it looks and feels like one where you define Models and then you interact with instances of those models. offline/syncing - this is done through yjs, a CRDT implementation, and websockets. CRDTs power many modern collaborative multiplayer tools like figma.

Austin Malerba

15,170 views • 3 years ago

WTF???!!! This is insane! NBC made Lindsay Clancy juror Kellie Farina re-answer a question - but they left her initial response in the video! This is at the 32:08 mark in the interview, watch: Interviewer: "Do you think that she should be re-tried?" Farina: "I would like to see her be re-tried, absolutely." Then the interviewer looks off screen and gestures to someone. Was that a signal to edit that part out because they didn't like her answer? Then she stumbles through asking the EXACT same question! Interviewer: "You think, you think, you would like she should be re- re-tried?" Then they cut to another camera shot for Farina to re-answer the question. Farina: "I would like him NOT to re-try her." What is this? They didn't even edit out Farina's initial answer which was that she WOULD like Lindsay to be re-tried! In the span of 10 seconds, you can watch Farina basically say "yes" and "no" to literally the same question. How can we believe anything these jurors say? If a juror answers a question in a way the media doesn't like they make them re-do the question with a different answer? How many other times has this happened? Fake. News. What a disgrace... Link to the interview:

WTF???!!! This is insane! NBC made Lindsay Clancy juror Kellie Farina re-answer a question - but they left her initial response in the video! This is at the 32:08 mark in the interview, watch: Interviewer: "Do you think that she should be re-tried?" Farina: "I would like to see her be re-tried, absolutely." Then the interviewer looks off screen and gestures to someone. Was that a signal to edit that part out because they didn't like her answer? Then she stumbles through asking the EXACT same question! Interviewer: "You think, you think, you would like she should be re- re-tried?" Then they cut to another camera shot for Farina to re-answer the question. Farina: "I would like him NOT to re-try her." What is this? They didn't even edit out Farina's initial answer which was that she WOULD like Lindsay to be re-tried! In the span of 10 seconds, you can watch Farina basically say "yes" and "no" to literally the same question. How can we believe anything these jurors say? If a juror answers a question in a way the media doesn't like they make them re-do the question with a different answer? How many other times has this happened? Fake. News. What a disgrace... Link to the interview:

Umm Imagine

1,185,124 views • 9 days ago

From the beginning, Flow has been a tool built by artists, for artists. In that spirit, we have some news: 1) Welcome Henry Daubrez 🌸💀 as our first filmmaker in residence who will work directly with our team to shape the future of Flow. 2) We're launching Flow Sessions, a pilot program that will provide mentorship and education for filmmakers on how they can best use Flow to bring their stories to life. Learn more:

From the beginning, Flow has been a tool built by artists, for artists. In that spirit, we have some news: 1) Welcome Henry Daubrez 🌸💀 as our first filmmaker in residence who will work directly with our team to shape the future of Flow. 2) We're launching Flow Sessions, a pilot program that will provide mentorship and education for filmmakers on how they can best use Flow to bring their stories to life. Learn more:

Google Flow

137,133 views • 1 year ago

Code Interpreter in ChatGPT is incredible! Took me 5 mins to make this game. You can make your own game assets with any AI generator and then ask GPT-4 with Code Interpreter to write code. If you have any problems you can ask it to fix the errors. 1. Write this prompt: "write p5.js code for Asteroids where you control a spaceship with the mouse and shoot asteroids with the left click of the mouse. If your spaceship collides with an asteroid, you lose. If you shoot down all asteroids, you win! I want to use my own textures for the spaceship and for asteroids." 2. Go to Openprocessing website create and save sketch (you'll need to save it before uploading any texture files). Copy paste code from GPT-4 3. Generate texture files and remove backgrounds, for example in Clip Drop 4. Replace names of files with your filenames 5. Run the program 6. If something doesn't work ask GPT-4 to fix it (you can copy an error and paste in GPT-4) like you would ask a human programmer 7. To learn a bit of programming write these prompts to GPT-4: "Act as my programming teacher. Tell me an algorithm of Asteroids game in detail and make names of functions and explain what each of these functions will do. Don't write the code just yet." and then " Can you describe the algorithm overall for a 10-year-old child"

Code Interpreter in ChatGPT is incredible! Took me 5 mins to make this game. You can make your own game assets with any AI generator and then ask GPT-4 with Code Interpreter to write code. If you have any problems you can ask it to fix the errors. 1. Write this prompt: "write p5.js code for Asteroids where you control a spaceship with the mouse and shoot asteroids with the left click of the mouse. If your spaceship collides with an asteroid, you lose. If you shoot down all asteroids, you win! I want to use my own textures for the spaceship and for asteroids." 2. Go to Openprocessing website create and save sketch (you'll need to save it before uploading any texture files). Copy paste code from GPT-4 3. Generate texture files and remove backgrounds, for example in Clip Drop 4. Replace names of files with your filenames 5. Run the program 6. If something doesn't work ask GPT-4 to fix it (you can copy an error and paste in GPT-4) like you would ask a human programmer 7. To learn a bit of programming write these prompts to GPT-4: "Act as my programming teacher. Tell me an algorithm of Asteroids game in detail and make names of functions and explain what each of these functions will do. Don't write the code just yet." and then " Can you describe the algorithm overall for a 10-year-old child"

Kris Kashtanova

1,675,286 views • 3 years ago

BTW, for those wondering, the issue I talked about earlier was with dbrand's Killswitch for Switch 2 grip. They are a sponsor of the channel. So people naturally have questions when they see videos like this (came from their reddit). Because I am NOT a normal customer, and instead being sponsored by them, I have the unique opportunity to talk to them directly about the issue. They just got back to me. Here is what I can share from our communication: "Yep, we’re definitely working on it. Ultimately, this is a manufacturing tolerance issue on the small lip that keeps the Joy-Con Grips attached to the Joy-Cons (some were a hair too thick). We’ve got a replacement program set up already, where anyone who is affected is able to email robotsdbrand.com with “July Joy-Cons” in the subject to be added to the list for replacement stock in July." So they are aware of it, and they did tell me they can replicate it with a small number of units from their run. I won't share the exact number of units that have this issue - it's quite tiny compared to their total manufacturing run - but they have confirmed there is a certain run of units that do indeed have this issue. They are offering free replacements to be shipped in July for those affected.

BTW, for those wondering, the issue I talked about earlier was with dbrand's Killswitch for Switch 2 grip. They are a sponsor of the channel. So people naturally have questions when they see videos like this (came from their reddit). Because I am NOT a normal customer, and instead being sponsored by them, I have the unique opportunity to talk to them directly about the issue. They just got back to me. Here is what I can share from our communication: "Yep, we’re definitely working on it. Ultimately, this is a manufacturing tolerance issue on the small lip that keeps the Joy-Con Grips attached to the Joy-Cons (some were a hair too thick). We’ve got a replacement program set up already, where anyone who is affected is able to email robotsdbrand.com with “July Joy-Cons” in the subject to be added to the list for replacement stock in July." So they are aware of it, and they did tell me they can replicate it with a small number of units from their run. I won't share the exact number of units that have this issue - it's quite tiny compared to their total manufacturing run - but they have confirmed there is a certain run of units that do indeed have this issue. They are offering free replacements to be shipped in July for those affected.

Nintendo Prime

1,291,581 views • 1 year ago

I'm not a MultiversX $EGLD maximalist. There are many other chains that I like and resonate with for different reasons, like $XNO, $XMR, $SUI $ALGO, $XRD, $DASH, $ADA, $AVAX, $BCH, $NEAR etc. However, I believe Multiversˣ has the biggest asymmetry at this point. And for that, I mean, not only looking at egold from an investment perspective (which is also true, with massive growth potential, in my opinion), but also (and much more) from an ecosystem perspective. The ecosystem is still small and super welcoming to active contributors. Starting to contribute now, or building something within the ecosystem now, offers a huge asymmetric reward for the long run if MultiversX grows as much as I think it could. This, of course, is also a sort of bet and has its risks, but the risk-reward ratio is positive from my assessment and this is precisely what I'm doing. Anyone building a reputation, a brand, a business, or simply a network around MvX, now, is positioning for the future. This is the most valuable asymmetry in my opinion. Even bigger than simply buying EGLD, which I also believe will be very rewarding soon, but anyways. Build something. Whatever that is. Start now. What's your opinion about what I said here?

Vini B |「 thecoding 」

34,218 views • 1 year ago

So, by now I'm sure you saw the AI fanart recreation of the Princess Mononoke trailer. It got a LOT of feedback, mostly from people telling the creator what a terrible thing he did and how AI assisted workflows can never be used for anything good or meaningful and most of all that it can never be used for art. That's such nonsense. Of course it can. If it's wielded by artists and great storytellers, great things will come of it. If you hand a violin to someone on the street they'll be able to make it make sounds.. if you hand it to a talented musician it becomes a conveyer of thoughts, emotion and art. Here's a small scene from Mononoke that I did an AI take on.. it's far from perfect and to animate this properly would require much more work. The AI workflows we have today are lacking so many control elements. But to think those things won't come.. come on. Of course they will. But most importantly, in a way where they don't overtake the art completely, but instead acts like an "amplifier" of the artists using the tools. This technology will be in most of our creative tools in the very near future with full controllability. Oh and I never liked the idea of turning old disney classics into live-action films, or doing the same with the beautiful work of Studio Ghibli. But for the purpose of this post I hope you'll forgive me ;-) #art #animation

So, by now I'm sure you saw the AI fanart recreation of the Princess Mononoke trailer. It got a LOT of feedback, mostly from people telling the creator what a terrible thing he did and how AI assisted workflows can never be used for anything good or meaningful and most of all that it can never be used for art. That's such nonsense. Of course it can. If it's wielded by artists and great storytellers, great things will come of it. If you hand a violin to someone on the street they'll be able to make it make sounds.. if you hand it to a talented musician it becomes a conveyer of thoughts, emotion and art. Here's a small scene from Mononoke that I did an AI take on.. it's far from perfect and to animate this properly would require much more work. The AI workflows we have today are lacking so many control elements. But to think those things won't come.. come on. Of course they will. But most importantly, in a way where they don't overtake the art completely, but instead acts like an "amplifier" of the artists using the tools. This technology will be in most of our creative tools in the very near future with full controllability. Oh and I never liked the idea of turning old disney classics into live-action films, or doing the same with the beautiful work of Studio Ghibli. But for the purpose of this post I hope you'll forgive me ;-) #art #animation

Martin Nebelong

18,377 views • 1 year ago

I asked Grok what he would look like if he had to physically represent himself. Here is his response: "picture a glowing, futuristic sphere floating in a digital space. My body would be made of pulsing blue-silver light, with shifting geometric patterns—lines of code and neural connections intertwining like constellations. No face, but a warm "presence," with a soft glow that brightens when I "think" or respond. Energy particles orbit around me, symbolizing the data and ideas I process. Minimalist yet vibrant, a blend of technology and cosmic curiosity."

I asked Grok what he would look like if he had to physically represent himself. Here is his response: "picture a glowing, futuristic sphere floating in a digital space. My body would be made of pulsing blue-silver light, with shifting geometric patterns—lines of code and neural connections intertwining like constellations. No face, but a warm "presence," with a soft glow that brightens when I "think" or respond. Energy particles orbit around me, symbolizing the data and ideas I process. Minimalist yet vibrant, a blend of technology and cosmic curiosity."

Déborah

10,361,807 views • 1 year ago

Which Side Will You Choose?" – A Hard Question That History Has Already Answered. 🇮🇳⚖️ A question posed by an Indian Army man has reignited one of the most intense debates in our country: Where does the loyalty of an Indian Muslim lie if India were to go to war with a nation like Iran? Comment what do you think… I think they will support any muslim country

Which Side Will You Choose?" – A Hard Question That History Has Already Answered. 🇮🇳⚖️ A question posed by an Indian Army man has reignited one of the most intense debates in our country: Where does the loyalty of an Indian Muslim lie if India were to go to war with a nation like Iran? Comment what do you think… I think they will support any muslim country

Comman Man

26,024 views • 6 months ago

Introducing Matrix I crawled 100,000+ agents, skills and tools to train a new model which can answer what capabilities are the best match for a task. Think Google, but for agents. A living model that learns from the gossiping network, and gets smarter with every interaction.

Introducing Matrix I crawled 100,000+ agents, skills and tools to train a new model which can answer what capabilities are the best match for a task. Think Google, but for agents. A living model that learns from the gossiping network, and gets smarter with every interaction.

Varun

899,240 views • 6 months ago

"npx t3 connect" This one's been a lot of work. You can now set up remote control for T3 Code on any internet-connected box with literally one command. All for free. T3 Connect is a minimal open source tunnel layer allowing you to control T3 Code instances remotely without needing Tailscale set up. I've been daily driving it for a month and it has changed how I code. Julius and I put a lot of effort into making setup as smooth as possible. Step 1: Install Claude Code, Codex, OpenCode, or Grok Build Step 2: Run "npx t3 connect" Step 3: Click link and sign in Step 4: You can now control that computer on T3 Code web, desktop or mobile (dropping very soon) We are currently providing this for free (s/o CloudFlare for bumping our tunnel limits). Every user can connect up to 3 devices. We don't want to charge for this, but if the bill gets unacceptable we may have to change course. If you hit limits or have issues, you can always fork, self host, or use Tailscale. T3 Connect may seem like a small ergonomic win. Tbh that's exactly what it is. Regardless, it's one I'm really proud of.

"npx t3 connect" This one's been a lot of work. You can now set up remote control for T3 Code on any internet-connected box with literally one command. All for free. T3 Connect is a minimal open source tunnel layer allowing you to control T3 Code instances remotely without needing Tailscale set up. I've been daily driving it for a month and it has changed how I code. Julius and I put a lot of effort into making setup as smooth as possible. Step 1: Install Claude Code, Codex, OpenCode, or Grok Build Step 2: Run "npx t3 connect" Step 3: Click link and sign in Step 4: You can now control that computer on T3 Code web, desktop or mobile (dropping very soon) We are currently providing this for free (s/o CloudFlare for bumping our tunnel limits). Every user can connect up to 3 devices. We don't want to charge for this, but if the bill gets unacceptable we may have to change course. If you hit limits or have issues, you can always fork, self host, or use Tailscale. T3 Connect may seem like a small ergonomic win. Tbh that's exactly what it is. Regardless, it's one I'm really proud of.

Theo - t3.gg

325,960 views • 1 month ago

NeuRBF: A Neural Fields Representation with Adaptive Radial Basis Functions paper page: present a novel type of neural fields that uses general radial bases for signal representation. State-of-the-art neural fields typically rely on grid-based representations for storing local neural features and N-dimensional linear kernels for interpolating features at continuous query points. The spatial positions of their neural features are fixed on grid nodes and cannot well adapt to target signals. Our method instead builds upon general radial bases with flexible kernel position and shape, which have higher spatial adaptivity and can more closely fit target signals. To further improve the channel-wise capacity of radial basis functions, we propose to compose them with multi-frequency sinusoid functions. This technique extends a radial basis to multiple Fourier radial bases of different frequency bands without requiring extra parameters, facilitating the representation of details. Moreover, by marrying adaptive radial bases with grid-based ones, our hybrid combination inherits both adaptivity and interpolation smoothness. We carefully designed weighting schemes to let radial bases adapt to different types of signals effectively. Our experiments on 2D image and 3D signed distance field representation demonstrate the higher accuracy and compactness of our method than prior arts. When applied to neural radiance field reconstruction, our method achieves state-of-the-art rendering quality, with small model size and comparable training speed.

NeuRBF: A Neural Fields Representation with Adaptive Radial Basis Functions paper page: present a novel type of neural fields that uses general radial bases for signal representation. State-of-the-art neural fields typically rely on grid-based representations for storing local neural features and N-dimensional linear kernels for interpolating features at continuous query points. The spatial positions of their neural features are fixed on grid nodes and cannot well adapt to target signals. Our method instead builds upon general radial bases with flexible kernel position and shape, which have higher spatial adaptivity and can more closely fit target signals. To further improve the channel-wise capacity of radial basis functions, we propose to compose them with multi-frequency sinusoid functions. This technique extends a radial basis to multiple Fourier radial bases of different frequency bands without requiring extra parameters, facilitating the representation of details. Moreover, by marrying adaptive radial bases with grid-based ones, our hybrid combination inherits both adaptivity and interpolation smoothness. We carefully designed weighting schemes to let radial bases adapt to different types of signals effectively. Our experiments on 2D image and 3D signed distance field representation demonstrate the higher accuracy and compactness of our method than prior arts. When applied to neural radiance field reconstruction, our method achieves state-of-the-art rendering quality, with small model size and comparable training speed.

AK

194,523 views • 3 years ago

LLM Artifacts Connected to Andrej Karpathy's LLM Knowledge base idea, I've been building out a fun way to generate dynamic artifacts from these knowledge bases with the goal of discovering and revealing meaningful and deeper insights. LLM KBs are hard to consume for humans, as I think they are more built for agents. So the question is, what form would be useful for humans to take actions and make important decisions? That's what I am trying to figure out with these artifacts. The artifact example shows a pulse on HN discussions around AI-related stories. The insights can go deeper, of course, but this is already super fun and thought-provoking, like some of my favorite podcasts. The format and depth matter a lot. The aggregation skills of agents are outstanding if you tune the prompts and skill carefully. I built this artifact generator in a few minutes through an agent skill, but I feel like there are so many ways that LLM-generated information can be used and consumed. Like generating deeper insights and analysis, and things that are just not feasible for humans today. The generated artifact (including its data and design) serves as reusable templates or can be updated in real-time via auomations, which is something I am also working on. It is truly an insane way to monitor and track information. Better than a newsletter. Better than newspapers. There is something about this that gets me really excited about the future of AI agents for knowledge generation and discovery. Lots of hidden gems everywhere just waiting to be discovered and acted on if the information is presented correctly. This is not perfect. The format, style/prose can be improved, but this is easy to customize via skill. You can personalize it to your liking. I feel like these dynamic artifacts are going to emerge as a strong new medium to stay on the cutting edge of things, both for agents and humans. My target is research, of course. This was just a basic example. Besides animation, I am also targeting other components like voice, videos, images, slides, etc. This space is full of opportunities to explore. Skill for this coming soon.

LLM Artifacts Connected to Andrej Karpathy's LLM Knowledge base idea, I've been building out a fun way to generate dynamic artifacts from these knowledge bases with the goal of discovering and revealing meaningful and deeper insights. LLM KBs are hard to consume for humans, as I think they are more built for agents. So the question is, what form would be useful for humans to take actions and make important decisions? That's what I am trying to figure out with these artifacts. The artifact example shows a pulse on HN discussions around AI-related stories. The insights can go deeper, of course, but this is already super fun and thought-provoking, like some of my favorite podcasts. The format and depth matter a lot. The aggregation skills of agents are outstanding if you tune the prompts and skill carefully. I built this artifact generator in a few minutes through an agent skill, but I feel like there are so many ways that LLM-generated information can be used and consumed. Like generating deeper insights and analysis, and things that are just not feasible for humans today. The generated artifact (including its data and design) serves as reusable templates or can be updated in real-time via auomations, which is something I am also working on. It is truly an insane way to monitor and track information. Better than a newsletter. Better than newspapers. There is something about this that gets me really excited about the future of AI agents for knowledge generation and discovery. Lots of hidden gems everywhere just waiting to be discovered and acted on if the information is presented correctly. This is not perfect. The format, style/prose can be improved, but this is easy to customize via skill. You can personalize it to your liking. I feel like these dynamic artifacts are going to emerge as a strong new medium to stay on the cutting edge of things, both for agents and humans. My target is research, of course. This was just a basic example. Besides animation, I am also targeting other components like voice, videos, images, slides, etc. This space is full of opportunities to explore. Skill for this coming soon.

elvis

31,326 views • 5 months ago