Loading video...

Video Failed to Load

Go Home

I've been using Ring-2.6-1T on OpenRouter for my documentation pipeline this week. My usual flow: I dump a raw codebase or a long spec doc into the model and ask it to generate structured docs, flag inconsistencies, and draft a changelog. With other models, I was splitting this into...

Parul_Gautam7's profile picture

Parul Gautam

86,648 subscribers

48,658 views • 3 months ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

ANTHROPIC JUST QUIETLY SHIPPED A FEATURE THAT LETS CLAUDE SPAWN A WHOLE TEAM OF AGENTS THAT MESSAGE EACH OTHER AND REVIEW EACH OTHER'S WORK. It's a Claude Code feature called agent teams. The team lead spawns multiple agents that share a task list and message each other directly, not subagents reporting back, actual peers. In the demo a QA agent caught three bugs, sent the work back to the front-end and back-end devs, they fixed it, app shipped in one pass. How to run it: 1. Enable it. Needs Claude Code v2.1.32+. Add to settings.json: "env": { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1" }. Or paste that to Claude and say "add this to my settings." Restart. 2. Prompt in plain English. Start with a goal (agents wake with zero context), then "create a team of 3 using Sonnet," describe each role, its deliverable, and who it messages when done. 3. The rules: each agent owns its own files, define exact outputs, name who talks to who, keep it to 3-5 agents. Use it for complex work with separate parts running in parallel. Skip it for simple or sequential tasks, teams cost 3-4x the tokens. Bookmark this.
16:29

ANTHROPIC JUST QUIETLY SHIPPED A FEATURE THAT LETS CLAUDE SPAWN A WHOLE TEAM OF AGENTS THAT MESSAGE EACH OTHER AND REVIEW EACH OTHER'S WORK. It's a Claude Code feature called agent teams. The team lead spawns multiple agents that share a task list and message each other directly, not subagents reporting back, actual peers. In the demo a QA agent caught three bugs, sent the work back to the front-end and back-end devs, they fixed it, app shipped in one pass. How to run it: 1. Enable it. Needs Claude Code v2.1.32+. Add to settings.json: "env": { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1" }. Or paste that to Claude and say "add this to my settings." Restart. 2. Prompt in plain English. Start with a goal (agents wake with zero context), then "create a team of 3 using Sonnet," describe each role, its deliverable, and who it messages when done. 3. The rules: each agent owns its own files, define exact outputs, name who talks to who, keep it to 3-5 agents. Use it for complex work with separate parts running in parallel. Skip it for simple or sequential tasks, teams cost 3-4x the tokens. Bookmark this.

Yarchi

461,708 views • 2 months ago

This is exactly what i was hoping for... - Full permissions - no "cowork" like feature which limits agent abilities - Just codex... If you ask for coding task it writes code and gives you a preview. If you ask for a presentation or doc it gives you a presentation or doc. Organized by project on the left side bar. Easy to create skills, easy to @ mention skills / plugins Btw my Readwise automation is insane. Every morning it codex sends me a deck of all my x bookmarks organized by topic. And i highly recommend using the excaladraw skill to render diagrams into docs.
0:33

This is exactly what i was hoping for... - Full permissions - no "cowork" like feature which limits agent abilities - Just codex... If you ask for coding task it writes code and gives you a preview. If you ask for a presentation or doc it gives you a presentation or doc. Organized by project on the left side bar. Easy to create skills, easy to @ mention skills / plugins Btw my Readwise automation is insane. Every morning it codex sends me a deck of all my x bookmarks organized by topic. And i highly recommend using the excaladraw skill to render diagrams into docs.

Riley Brown

65,403 views • 4 months ago

I set up Gemini Spark to write my own weekly AI news digest for me, and I'm still stuck on how little effort I had to put in. This is the exact prompt I gave it, just once: "Every Tuesday at 12:15 PM, create a new Doc in my Drive, find the 5 biggest AI stories from that week, and format each one properly." And here's what it actually produces, every week: 1. A new Doc, created and named on its own 2. The 5 biggest AI stories from that week, found and picked automatically 3. A bold headline and a summary for each one 4. The key impact spelled out, not just the news 5. A source link on every single story This runs on a schedule so I never have to think about it again. I didn't ask anything on Tuesday, I just opened the drive and the news digest doc was already there. Just like this, you can use Spark to automate any recurring task.
1:06

I set up Gemini Spark to write my own weekly AI news digest for me, and I'm still stuck on how little effort I had to put in. This is the exact prompt I gave it, just once: "Every Tuesday at 12:15 PM, create a new Doc in my Drive, find the 5 biggest AI stories from that week, and format each one properly." And here's what it actually produces, every week: 1. A new Doc, created and named on its own 2. The 5 biggest AI stories from that week, found and picked automatically 3. A bold headline and a summary for each one 4. The key impact spelled out, not just the news 5. A source link on every single story This runs on a schedule so I never have to think about it again. I didn't ask anything on Tuesday, I just opened the drive and the news digest doc was already there. Just like this, you can use Spark to automate any recurring task.

Eyisha Zyer

41,454 views • 23 days ago

It was a big day for me. For the FIRST TIME in a long time, I bought more $XRP with the intention of holding it and adding to my long term bag. It’s been tough to want to buy above $2, because I got in so early at $0.17 - $0.50 and packed my bags. The only time i’ve bought more in the past couple years was for the purpose of using it to trade memecoins on the XRPL. Why haven’t I bought more? Because I was stacking my bags with other tokens and expanding my portfolio!
0:47

It was a big day for me. For the FIRST TIME in a long time, I bought more $XRP with the intention of holding it and adding to my long term bag. It’s been tough to want to buy above $2, because I got in so early at $0.17 - $0.50 and packed my bags. The only time i’ve bought more in the past couple years was for the purpose of using it to trade memecoins on the XRPL. Why haven’t I bought more? Because I was stacking my bags with other tokens and expanding my portfolio!

MASON VERSLUIS

65,617 views • 7 months ago

over the weekend i built and shipped Sentra for our docs. we use Mintlify for documentation. the built-in assistant kept running into context issues. wrong answers, missing context, no citations. so we extended Sentra's capability to the docs site directly. now when you're reading Dodo Payments docs, you can ask Sentra questions right there. it pulls context from our entire documentation via RAG (the same contextmcp (completely opensource) + Pinecone vector search with reranking), generates grounded answers, and cites the exact doc pages. built as a single Cloudflare Worker serving both the widget and the chat endpoint. React widget bundled into one self-contained IIFE, mounted in Shadow DOM so it doesn't conflict with Mintlify's styles. OpenAI models routed through Cloudflare AI Gateway. Vercel's AI SDK for streaming and shadcn for beautiful chat components. live now at
0:25

over the weekend i built and shipped Sentra for our docs. we use Mintlify for documentation. the built-in assistant kept running into context issues. wrong answers, missing context, no citations. so we extended Sentra's capability to the docs site directly. now when you're reading Dodo Payments docs, you can ask Sentra questions right there. it pulls context from our entire documentation via RAG (the same contextmcp (completely opensource) + Pinecone vector search with reranking), generates grounded answers, and cites the exact doc pages. built as a single Cloudflare Worker serving both the widget and the chat endpoint. React widget bundled into one self-contained IIFE, mounted in Shadow DOM so it doesn't conflict with Mintlify's styles. OpenAI models routed through Cloudflare AI Gateway. Vercel's AI SDK for streaming and shadcn for beautiful chat components. live now at

Ayush Agarwal

18,001 views • 1 month ago

With the release of Kling AI 2.6, I was eager to put it up against one of its arch-rivals….Google Flow Veo 3.1. I ran four separate tests. First, I tested their ability to generate consistent and stable action shots. Second, I tested camera movement in a crowded scene. Third, I tested their ability to generate consistent facial details and emotion. Fourth, I wanted to see how well they can generated an interaction between two characters. Prompt 1: camera tracking shot, wide-shot. The camera tracks above as two nights ride on horses through a forest. The horses a galloping fast. The camera is tracking the knights from above. Prompt 2: Slow camera push in on a man eating Ramen at a Ramen house in Japan. The alley street is busy with people. Prompt 3: Close-up shot of a man on a sailboat. The man has a scared look on his face. After a short pause, the man says, "I'm gonna need a bigger boat” Prompt 4: Static camera. A man picks up a cookie from a plate and takes a bite. His dog looks up at him; eagerly wanting a bite of the cookie. The man turns, looks at the dog, and says, "no cookies for you mister” Verdict🏆: To be honest, I was surprised by Kling 2.6. While Veo 3.1 outperformed Kling 2.6 in the close-up shot test, it slightly underperformed in every other test. In the Veo 3.1 outputs, objects would randomly appear and camera movements were abrupt or didn’t adhere to the prompt. Don’t get me wrong, Kling 2.6 definitely had its flaws (its audio was quieter and often mismatched the prompt). That said, I was impressed with Kling 2.6 more than I imagined I would be. Although a somewhat small lead, I would give Kling 2.6 the slight advantage for this round of tests.
1:03

With the release of Kling AI 2.6, I was eager to put it up against one of its arch-rivals….Google Flow Veo 3.1. I ran four separate tests. First, I tested their ability to generate consistent and stable action shots. Second, I tested camera movement in a crowded scene. Third, I tested their ability to generate consistent facial details and emotion. Fourth, I wanted to see how well they can generated an interaction between two characters. Prompt 1: camera tracking shot, wide-shot. The camera tracks above as two nights ride on horses through a forest. The horses a galloping fast. The camera is tracking the knights from above. Prompt 2: Slow camera push in on a man eating Ramen at a Ramen house in Japan. The alley street is busy with people. Prompt 3: Close-up shot of a man on a sailboat. The man has a scared look on his face. After a short pause, the man says, "I'm gonna need a bigger boat” Prompt 4: Static camera. A man picks up a cookie from a plate and takes a bite. His dog looks up at him; eagerly wanting a bite of the cookie. The man turns, looks at the dog, and says, "no cookies for you mister” Verdict🏆: To be honest, I was surprised by Kling 2.6. While Veo 3.1 outperformed Kling 2.6 in the close-up shot test, it slightly underperformed in every other test. In the Veo 3.1 outputs, objects would randomly appear and camera movements were abrupt or didn’t adhere to the prompt. Don’t get me wrong, Kling 2.6 definitely had its flaws (its audio was quieter and often mismatched the prompt). That said, I was impressed with Kling 2.6 more than I imagined I would be. Although a somewhat small lead, I would give Kling 2.6 the slight advantage for this round of tests.

Curious Refuge

24,695 views • 8 months ago

I’ve been using Grok Bot in early access for a little over a week now, and it has been a massive boost to my game dev workflow and my productivity in general. One of the first things I asked it to do was to visit my custom art generation tool webpage and generate art assets to replace all the image placeholders in my game. It looked through my codebase to understand what each asset was, wrote a custom art generation prompt for every single one, generated the images, cropped and cleaned them up into transparent PNGs, and then wired them back into the game. In about 2 hours, it generated 74 (cleaned and cropped) art assets for the cards and integrated all of them into my game. Now, ALL my art assets workflows go through Grok bot. This blew my mind because, before this, I was doing everything manually, one asset at a time. It basically solved my entire art asset workflow and allowed me to put it on autopilot, which is amazing. I’ve also been using Grok Bot for: Playtesting my game: The bot was a little slow initially, so I asked it to build a skill specifically designed to playtest my game, with intimate knowledge of its rules and mechanics. It’s much faster now, though I’m still looking for ways to optimise its speed. Setting up my itchio page for my game: It also automated a workflow that uploads a new build to itch whenever I push one to GitHub. Going to the itch website and uploading builds manually every single time was always a pain. Flowcharts and UX design: I’m a UX designer by day, so I decided to give it a much bigger task. I fed the agent a PRD and asked it to design the UX flows and wireframes. With the help of the Figma MCP, it actually delivered. Auditing all my paid subscriptions: I had it go through my emails and hunt down subscriptions that I might have forgotten about. Unsubscribing me from marketing newsletters: It did this beautifully, though it did miss a few newsletters. What I’ve learned is that it takes time to discover all the possible use cases for Grok bot. I’ve had to reevaluate my day-to-day workflows and figure out which parts I can delegate to it. So far, it has been a huge productivity unlock for me and I am discovering more uses for my Grok bots every other day.
0:22

I’ve been using Grok Bot in early access for a little over a week now, and it has been a massive boost to my game dev workflow and my productivity in general. One of the first things I asked it to do was to visit my custom art generation tool webpage and generate art assets to replace all the image placeholders in my game. It looked through my codebase to understand what each asset was, wrote a custom art generation prompt for every single one, generated the images, cropped and cleaned them up into transparent PNGs, and then wired them back into the game. In about 2 hours, it generated 74 (cleaned and cropped) art assets for the cards and integrated all of them into my game. Now, ALL my art assets workflows go through Grok bot. This blew my mind because, before this, I was doing everything manually, one asset at a time. It basically solved my entire art asset workflow and allowed me to put it on autopilot, which is amazing. I’ve also been using Grok Bot for: Playtesting my game: The bot was a little slow initially, so I asked it to build a skill specifically designed to playtest my game, with intimate knowledge of its rules and mechanics. It’s much faster now, though I’m still looking for ways to optimise its speed. Setting up my itchio page for my game: It also automated a workflow that uploads a new build to itch whenever I push one to GitHub. Going to the itch website and uploading builds manually every single time was always a pain. Flowcharts and UX design: I’m a UX designer by day, so I decided to give it a much bigger task. I fed the agent a PRD and asked it to design the UX flows and wireframes. With the help of the Figma MCP, it actually delivered. Auditing all my paid subscriptions: I had it go through my emails and hunt down subscriptions that I might have forgotten about. Unsubscribing me from marketing newsletters: It did this beautifully, though it did miss a few newsletters. What I’ve learned is that it takes time to discover all the possible use cases for Grok bot. I’ve had to reevaluate my day-to-day workflows and figure out which parts I can delegate to it. So far, it has been a huge productivity unlock for me and I am discovering more uses for my Grok bots every other day.

Danny Limanseta

404,057 views • 16 days ago

Quick demo of how I use session to session messaging in Claude Code + Ghostty (Save this and try it in your own workflows) 1- I always start with a regular session and a panel running claude agents 2- The first session becomes the orchestrator. It creates the initial context, then distributes the work across other sessions using forks 3- Each fork starts with that context + a prompt telling it what to work on 4- I rename each session with ctrl + r to keep things organized and make it easy to message them from the main session 5- Using @ I send messages to the sessions that are running and ask them to report their task status back, so the main session always knows what everyone is working on This way, I only need 2 or 3 panels open and can switch between tasks as needed The days of opening endless terminal panels and running Claude Code 100 times just to look productive are over! What we need now is ORCHESTATION one main session coordinating everything through session messaging + Agent View And the key is context, build it in the main session first, then fork from there That’s what makes the entire workflow click
0:28

Quick demo of how I use session to session messaging in Claude Code + Ghostty (Save this and try it in your own workflows) 1- I always start with a regular session and a panel running claude agents 2- The first session becomes the orchestrator. It creates the initial context, then distributes the work across other sessions using forks 3- Each fork starts with that context + a prompt telling it what to work on 4- I rename each session with ctrl + r to keep things organized and make it easy to message them from the main session 5- Using @ I send messages to the sessions that are running and ask them to report their task status back, so the main session always knows what everyone is working on This way, I only need 2 or 3 panels open and can switch between tasks as needed The days of opening endless terminal panels and running Claude Code 100 times just to look productive are over! What we need now is ORCHESTATION one main session coordinating everything through session messaging + Agent View And the key is context, build it in the main session first, then fork from there That’s what makes the entire workflow click

Daniel San

47,525 views • 8 days ago

Jordan Spieth shot a bogey free 62 yesterday, with 6 birdies in a row, and a closing 9 of 29, to climb into the top 10 of the CJ Cup Byron Nelson. After the round he said he feels as good as he has in 9 years and then went into more detail about why: “It's physically and my mechanics. I got pretty off for a long time. I've been trying to build it back, and then I'd compensate and do what worked. “This last off-season I said no more compensating because, to be consistent, I've got to get it back to a certain place, and it's been work from then to try to get there. It's all in mechanics and health.” He then detailed what part of his game worked well in the 62: “It was putting today. I've been driving the ball the best of my life, and I drove it horribly today, like really badly. But putting was -- to go to your actual question, putting was the best. “I felt fluid with it. I felt like I -- in what I've been trying to do, I've been trying to put it all together. I know what needs to happen, but putting it all together into a fluid stroke and then being able to be outwardly focused has been the goal. “I felt like last night I did a really good job out here of that being the goal, find that, and then carry it into this morning and then trust it on the course. “Frankly, I mean, I told Michael -- I've been telling Michael for a while. I just feel like one lips in instead of lips out, and I feel like the lid comes off. I lipped out a few yesterday. I had one on No. 10 go across the lip today, but then the one on No. 1 could have been short and fell in. It was like, oh, wow, there is a hole there. “Sometimes I've certainly shown that I can get streaky with it and was able to kind of get on the right side of that today.” Jordan Spieth PGA TOUR
1:49

Jordan Spieth shot a bogey free 62 yesterday, with 6 birdies in a row, and a closing 9 of 29, to climb into the top 10 of the CJ Cup Byron Nelson. After the round he said he feels as good as he has in 9 years and then went into more detail about why: “It's physically and my mechanics. I got pretty off for a long time. I've been trying to build it back, and then I'd compensate and do what worked. “This last off-season I said no more compensating because, to be consistent, I've got to get it back to a certain place, and it's been work from then to try to get there. It's all in mechanics and health.” He then detailed what part of his game worked well in the 62: “It was putting today. I've been driving the ball the best of my life, and I drove it horribly today, like really badly. But putting was -- to go to your actual question, putting was the best. “I felt fluid with it. I felt like I -- in what I've been trying to do, I've been trying to put it all together. I know what needs to happen, but putting it all together into a fluid stroke and then being able to be outwardly focused has been the goal. “I felt like last night I did a really good job out here of that being the goal, find that, and then carry it into this morning and then trust it on the course. “Frankly, I mean, I told Michael -- I've been telling Michael for a while. I just feel like one lips in instead of lips out, and I feel like the lid comes off. I lipped out a few yesterday. I had one on No. 10 go across the lip today, but then the one on No. 1 could have been short and fell in. It was like, oh, wow, there is a hole there. “Sometimes I've certainly shown that I can get streaky with it and was able to kind of get on the right side of that today.” Jordan Spieth PGA TOUR

Flushing It

15,564 views • 3 months ago

"I burned roughly $5,000 of API tokens in a few hours for $200. It's 100x subsidized. This is insanity" Kain on running 500+ agents to decompose an entire codebase on Claude Opus 4.8 "I set up my own personal account to test this on Opus 4.8. I know enough that I can come up with a task that's gonna burn a fuckload. In the course of four hours I used something like 200 million tokens and didn't hit any limits" "It was hundreds of agents, like 500 or 600 agents decomposing a codebase. It would have been somewhere on the order of $5,000 with API metered tokens, and I know how much tokens cost because that's where I live" "I ran this overnight multiple times. It's very possible I could have burned through $20,000 worth of API credits for $200. I knew it was subsidized, but it's subsidized to a level that's crazy. It's 100x subsidized"
2:06

"I burned roughly $5,000 of API tokens in a few hours for $200. It's 100x subsidized. This is insanity" Kain on running 500+ agents to decompose an entire codebase on Claude Opus 4.8 "I set up my own personal account to test this on Opus 4.8. I know enough that I can come up with a task that's gonna burn a fuckload. In the course of four hours I used something like 200 million tokens and didn't hit any limits" "It was hundreds of agents, like 500 or 600 agents decomposing a codebase. It would have been somewhere on the order of $5,000 with API metered tokens, and I know how much tokens cost because that's where I live" "I ran this overnight multiple times. It's very possible I could have burned through $20,000 worth of API credits for $200. I knew it was subsidized, but it's subsidized to a level that's crazy. It's 100x subsidized"

Laura Shin

85,135 views • 2 months ago

I asked Garry Tan how to use meta prompting to get better at AI: "My partners at YC Jared Friedman and Pete Koomen showed me how to do this. You can take almost anything that you do all the time and just drop it into a context window. And then say, “Here’s a bunch of inputs and outputs." And maybe you also add a bunch of notes. And then you tell it, “Write me a prompt that can act as an agent that takes this input and makes this output over here.” You can do this for almost any type of knowledge work. And you can even introspect. "What are things you notice that I did to convert this from the input to the output?”. And then you can just start using the prompt. Initially, it’s going to suck. Because it’s just not that smart yet. But what’s funny is now, I also use it to Iterate my writing. You can be very direct, "I would never say that", "Don’t say it like this", or "Oh, you used the long word there, use the short word". Just speak to it conversationally. And then when you're happy with the output, you can use that new output to make a new prompt. "Based on this conversation, give me a better initial prompt that incorporates all the things we talked about." And you can do this with literally everything. And in theory, there’s so much it applies to that people do day-to-day. You could use it for tweets. You could use it for editing podcasts. You can use it for pretty much everything. I have a folder of prompts that I use all the time. My YouTube prompt is on v27 or something. I'll go through this process with all the different max models. I'll use GPT 5.2 Pro. I’ll use Grok. I'll use Claude. Then, I’ll take all the outputs from all the models and put them into Claude and say "Here’s my prompt, here’s the output from four LLMs, including yourself. Rate each response and tell me what the pros and cons of each approach are." And I usually say "give it to me in numbered form". And then you can agree with one, disagree with two, tell it three is this or that. And then after that, you say given all of this, synthesize it."
4:16

I asked Garry Tan how to use meta prompting to get better at AI: "My partners at YC Jared Friedman and Pete Koomen showed me how to do this. You can take almost anything that you do all the time and just drop it into a context window. And then say, “Here’s a bunch of inputs and outputs." And maybe you also add a bunch of notes. And then you tell it, “Write me a prompt that can act as an agent that takes this input and makes this output over here.” You can do this for almost any type of knowledge work. And you can even introspect. "What are things you notice that I did to convert this from the input to the output?”. And then you can just start using the prompt. Initially, it’s going to suck. Because it’s just not that smart yet. But what’s funny is now, I also use it to Iterate my writing. You can be very direct, "I would never say that", "Don’t say it like this", or "Oh, you used the long word there, use the short word". Just speak to it conversationally. And then when you're happy with the output, you can use that new output to make a new prompt. "Based on this conversation, give me a better initial prompt that incorporates all the things we talked about." And you can do this with literally everything. And in theory, there’s so much it applies to that people do day-to-day. You could use it for tweets. You could use it for editing podcasts. You can use it for pretty much everything. I have a folder of prompts that I use all the time. My YouTube prompt is on v27 or something. I'll go through this process with all the different max models. I'll use GPT 5.2 Pro. I’ll use Grok. I'll use Claude. Then, I’ll take all the outputs from all the models and put them into Claude and say "Here’s my prompt, here’s the output from four LLMs, including yourself. Rate each response and tell me what the pros and cons of each approach are." And I usually say "give it to me in numbered form". And then you can agree with one, disagree with two, tell it three is this or that. And then after that, you say given all of this, synthesize it."

The Peel

51,632 views • 6 months ago

🔥 💻 🎥 How to provide o1 Pro with your FULL CODEBASE through the ChatGPT Cursor connection yesterday I recorded a video connecting o1 Pro to Cursor through the ChatGPT Desktop app more than one person commented on how limiting it is for o1 Pro to only have access to a single open file in Cursor here's a walkthrough of how you can provide o1 Pro with your FULL CODEBASE as context instead of just a single file: 1. Write a Python function that concatenates full codebase into a single snapshot in a .txt file (link in comment) 2. Open that "codebase_snapshot.txt" in a separate Cursor pane 3. Go to the ChatGPT app and ask it what context it has access to through Cursor - should say it has two files with one of them being the codebase snapshot and boom there ya go. o1 Pro has access to everything in your codebase full walkthrough here 👇
6:14

🔥 💻 🎥 How to provide o1 Pro with your FULL CODEBASE through the ChatGPT Cursor connection yesterday I recorded a video connecting o1 Pro to Cursor through the ChatGPT Desktop app more than one person commented on how limiting it is for o1 Pro to only have access to a single open file in Cursor here's a walkthrough of how you can provide o1 Pro with your FULL CODEBASE as context instead of just a single file: 1. Write a Python function that concatenates full codebase into a single snapshot in a .txt file (link in comment) 2. Open that "codebase_snapshot.txt" in a separate Cursor pane 3. Go to the ChatGPT app and ask it what context it has access to through Cursor - should say it has two files with one of them being the codebase snapshot and boom there ya go. o1 Pro has access to everything in your codebase full walkthrough here 👇

Dan McAteer

99,719 views • 1 year ago

🚨Olivia Rodrigo in a new interview on her split from Louis Partridge: 🎙️Interviewer: Fans were devastated when they heard you and Louis had split. How are you doing, and what led to the breakup? 🎙️Olivia Rodrigo: “Honestly, I didn’t want to speak about it because the split took a huge emotional toll on me. My touring and work schedule kept me away from him, but my heart was still with him. Then I found out he had been hóoking up other women behind my back while I was gone.💔 I guess some people just can't handle the distance, but loyalty shouldn't be conditional on geography. I ended up pouring all of that grief straight into the studio, because that's the only way to survive that kind of pain. I was absolutely devastated and it shattered my trust. It was a really dark time, but it's all in the past now. I've processed it, and I'm just focusing on myself and moving forward.".”
0:28

🚨Olivia Rodrigo in a new interview on her split from Louis Partridge: 🎙️Interviewer: Fans were devastated when they heard you and Louis had split. How are you doing, and what led to the breakup? 🎙️Olivia Rodrigo: “Honestly, I didn’t want to speak about it because the split took a huge emotional toll on me. My touring and work schedule kept me away from him, but my heart was still with him. Then I found out he had been hóoking up other women behind my back while I was gone.💔 I guess some people just can't handle the distance, but loyalty shouldn't be conditional on geography. I ended up pouring all of that grief straight into the studio, because that's the only way to survive that kind of pain. I was absolutely devastated and it shattered my trust. It was a really dark time, but it's all in the past now. I've processed it, and I'm just focusing on myself and moving forward.".”

BUSTED🧌

364,438 views • 2 days ago

THE $4,000 DGX SPARK BENCHMARKS AT 342 T/S ON PROMPT PROCESSING - THEN DROPS TO 73 WHEN IT ACTUALLY GENERATES. that clip is a dgx spark running llama-bench on qwen3-coder 30b, side by side with a mac. two numbers decide everything here: pp512, prompt processing = 342 tokens/sec tg128, token generation = 73 tokens/sec that gap is the whole lesson. local ai isn't one speed, it's two. prompt processing is reading your input: your files, your long context. the spark is compute-heavy, so it eats that fast. token generation is writing the answer, and it's bound by memory bandwidth, where the spark is modest and unified-memory macs stay close. so "which box is faster" has no single winner. feeding huge context or running agents that read a lot? prompt processing wins, the spark shines. long chat replies, one token at a time? generation matters more, and a mac holds its own. no single benchmark number, no universal "fastest box", no upgrade that wins every workload. the uncomfortable part isn't that the spark is fast. it's that "tokens per second" was always two numbers, and the hype only ever quotes the flattering one. follow RetroChainer for the benchmark truth nobody screenshots, and bookmark this before you buy on one number.
1:14

THE $4,000 DGX SPARK BENCHMARKS AT 342 T/S ON PROMPT PROCESSING - THEN DROPS TO 73 WHEN IT ACTUALLY GENERATES. that clip is a dgx spark running llama-bench on qwen3-coder 30b, side by side with a mac. two numbers decide everything here: pp512, prompt processing = 342 tokens/sec tg128, token generation = 73 tokens/sec that gap is the whole lesson. local ai isn't one speed, it's two. prompt processing is reading your input: your files, your long context. the spark is compute-heavy, so it eats that fast. token generation is writing the answer, and it's bound by memory bandwidth, where the spark is modest and unified-memory macs stay close. so "which box is faster" has no single winner. feeding huge context or running agents that read a lot? prompt processing wins, the spark shines. long chat replies, one token at a time? generation matters more, and a mac holds its own. no single benchmark number, no universal "fastest box", no upgrade that wins every workload. the uncomfortable part isn't that the spark is fast. it's that "tokens per second" was always two numbers, and the hype only ever quotes the flattering one. follow RetroChainer for the benchmark truth nobody screenshots, and bookmark this before you buy on one number.

RetroChainer

42,451 views • 19 days ago

First fight ever and I really put my heart on the line. 3v1 is no joke — at moments I genuinely felt like I was gonna die in that ring, but I kept pushing because quitting was never an option. I trained through stress, doubt, noise and pressure and still showed up. Big respect to the other fighters for stepping in and giving it everything — no easy work in there. Massive thank you to my coach for believing in me and guiding me through it, and to Unhinged Promotions for putting on a proper show. This wasn’t just a fight, this was a test of character. Proud of myself. On to the next. 🥊
0:44

First fight ever and I really put my heart on the line. 3v1 is no joke — at moments I genuinely felt like I was gonna die in that ring, but I kept pushing because quitting was never an option. I trained through stress, doubt, noise and pressure and still showed up. Big respect to the other fighters for stepping in and giving it everything — no easy work in there. Massive thank you to my coach for believing in me and guiding me through it, and to Unhinged Promotions for putting on a proper show. This wasn’t just a fight, this was a test of character. Proud of myself. On to the next. 🥊

Mizzy

52,981 views • 7 months ago

Being able to move fluidly between Figma Make and the canvas is a pretty big deal Recent example 👇 1️⃣ Started exploring a minified recording toolbar directly on the canvas 2️⃣ Once I had some rough shapes I moved it into Make, imported my style guide, and started bringing it to life with expandable controls, hover states, and tooltips for keyboard shortcuts 3️⃣ A few prompts later, I had something interesting… but the layout wasn’t quite right. 4️⃣ So I jumped back into the canvas and made ~8 quick variations with a bigger video container. 5️⃣ Once I got somewhere interesting I brought that back into Make. All I had to do was copy and paste the frames for my default and hover state of the component and it pretty much nailed the core layout 6️⃣ Then I layered on the timer and made it interactive. But once I got it back into Make I thought "hmm... what if the toolbar layout was based on which quadrant of the screen the component was in 🤔" I was able to one shot this interaction and after playing with it a bunch I kinda love it? 7️⃣ Sent it to the team in Inflight so everyone could play with the prototype. 8️⃣ Brought the up to date UI back into the canvas to polish some of the tooltip details Before this, it always felt like I had to choose between two very different workflows. Now I can bounce between them seamlessly and it’s genuinely changing how I explore ideas.
0:28

Being able to move fluidly between Figma Make and the canvas is a pretty big deal Recent example 👇 1️⃣ Started exploring a minified recording toolbar directly on the canvas 2️⃣ Once I had some rough shapes I moved it into Make, imported my style guide, and started bringing it to life with expandable controls, hover states, and tooltips for keyboard shortcuts 3️⃣ A few prompts later, I had something interesting… but the layout wasn’t quite right. 4️⃣ So I jumped back into the canvas and made ~8 quick variations with a bigger video container. 5️⃣ Once I got somewhere interesting I brought that back into Make. All I had to do was copy and paste the frames for my default and hover state of the component and it pretty much nailed the core layout 6️⃣ Then I layered on the timer and made it interactive. But once I got it back into Make I thought "hmm... what if the toolbar layout was based on which quadrant of the screen the component was in 🤔" I was able to one shot this interaction and after playing with it a bunch I kinda love it? 7️⃣ Sent it to the team in Inflight so everyone could play with the prototype. 8️⃣ Brought the up to date UI back into the canvas to polish some of the tooltip details Before this, it always felt like I had to choose between two very different workflows. Now I can bounce between them seamlessly and it’s genuinely changing how I explore ideas.

Ridd 🤿

27,665 views • 10 months ago

🧵1/3 I partnered with Kling to make a promo for their new 3.0 model. I came up with the concept, created it and delivered it all on my own in 3 days of early access, I wanted to make something that showed how Kling could be used to tell a diverse range of stories in a diverse range of styles. I've honestly been blown away be how incredible the model is and I haven't even scratched the surface of features like the new 3.0 OMNI model. For me Kling 2.6 was already best in class for most things but 3.0 sees massive improvements to video quality and detail (no need to add grain to mask imperfections) and prompt adherence. Also you can time your generations to be as long as 15 seconds and as short as 3, this is fantastic for pacing your scenes. Read on to find out more about the new Performance and Multi shot features!👇
1:26

🧵1/3 I partnered with Kling to make a promo for their new 3.0 model. I came up with the concept, created it and delivered it all on my own in 3 days of early access, I wanted to make something that showed how Kling could be used to tell a diverse range of stories in a diverse range of styles. I've honestly been blown away be how incredible the model is and I haven't even scratched the surface of features like the new 3.0 OMNI model. For me Kling 2.6 was already best in class for most things but 3.0 sees massive improvements to video quality and detail (no need to add grain to mask imperfections) and prompt adherence. Also you can time your generations to be as long as 15 seconds and as short as 3, this is fantastic for pacing your scenes. Read on to find out more about the new Performance and Multi shot features!👇

Uncanny Harry AI

29,194 views • 6 months ago

Colton McKivitz was asked if there are any regrets in signing a contract extension rather than testing free agency: “No, not one. My goal never was to be the highest paid. Obviously you’re in the NFL and money’s the main thing. We all want to get paid and have generational wealth… With the contract I was offered and honestly I don’t want to leave here. Obviously being cut I got some scars there, but to be cut, be a draft pick here, a gold helmet guy— whatever that means, but there was a reason why I was drafted here, why I was cut here, why I still play here and now I have had 3 good seasons starting and going into the 4th. The grass isn’t always greener on the other side. I saw a lot of guys not get paid in free agency this year. I like it here and I want to be on that 10 year wall.”
1:37

Colton McKivitz was asked if there are any regrets in signing a contract extension rather than testing free agency: “No, not one. My goal never was to be the highest paid. Obviously you’re in the NFL and money’s the main thing. We all want to get paid and have generational wealth… With the contract I was offered and honestly I don’t want to leave here. Obviously being cut I got some scars there, but to be cut, be a draft pick here, a gold helmet guy— whatever that means, but there was a reason why I was drafted here, why I was cut here, why I still play here and now I have had 3 good seasons starting and going into the 4th. The grass isn’t always greener on the other side. I saw a lot of guys not get paid in free agency this year. I like it here and I want to be on that 10 year wall.”

Coach Yac 🗣

112,870 views • 3 months ago

I nearly 2x'd the speed while only using +1GB VRAM with the new MTP update in llama.cpp 🤯 You need to add these flags to start using it: --spec-type draft-mtp \ --spec-draft-p-min 0.75 \ --spec-draft-n-max 2 My results with Qwen3.6 27B on a single RTX 5080 ↓ ⚪️ no flag (without mtp) → 54.3 tok/s with 13.26GB VRAM 🔵 --spec-draft-n-max 2 → 90.7 tok/s with 14.29GB VRAM 🔴 --spec-draft-n-max 2 --spec-draft-p-min 0.75 → 93.9 tok/s with 14.30GB VRAM 🟢 --spec-draft-n-max 6 --spec-draft-p-min 0.75 → 93.9 tok/s with 14.87GB VRAM Increasing to 6 draft tokens didn't help my setup for some reason. I made sure to test with a low context length to have enough headroom and eliminate risk of vram stress. From my understanding: 1) The speed gains are very task-dependent. You need to test across a wide range of tasks to get a realistic idea of the benefits 2) We’re already running heavily quantized GGUF models (Q3, Q4, Q6, etc.), so we already benefit from strong speed/performance thanks to the reduced size. That’s why some people are seeing little to no improvement compared to MLX or other quantized versions The progress over the past few days has been insane to say the least. However, MTP now consumes significantly more VRAM. Personally 16GB just isn't enough to use MTP and run it with a good context size. Time to upgrade lads, 24GB+ users are eating GOOD today 🔥 Full setup below ↓
0:33

I nearly 2x'd the speed while only using +1GB VRAM with the new MTP update in llama.cpp 🤯 You need to add these flags to start using it: --spec-type draft-mtp \ --spec-draft-p-min 0.75 \ --spec-draft-n-max 2 My results with Qwen3.6 27B on a single RTX 5080 ↓ ⚪️ no flag (without mtp) → 54.3 tok/s with 13.26GB VRAM 🔵 --spec-draft-n-max 2 → 90.7 tok/s with 14.29GB VRAM 🔴 --spec-draft-n-max 2 --spec-draft-p-min 0.75 → 93.9 tok/s with 14.30GB VRAM 🟢 --spec-draft-n-max 6 --spec-draft-p-min 0.75 → 93.9 tok/s with 14.87GB VRAM Increasing to 6 draft tokens didn't help my setup for some reason. I made sure to test with a low context length to have enough headroom and eliminate risk of vram stress. From my understanding: 1) The speed gains are very task-dependent. You need to test across a wide range of tasks to get a realistic idea of the benefits 2) We’re already running heavily quantized GGUF models (Q3, Q4, Q6, etc.), so we already benefit from strong speed/performance thanks to the reduced size. That’s why some people are seeing little to no improvement compared to MLX or other quantized versions The progress over the past few days has been insane to say the least. However, MTP now consumes significantly more VRAM. Personally 16GB just isn't enough to use MTP and run it with a good context size. Time to upgrade lads, 24GB+ users are eating GOOD today 🔥 Full setup below ↓

left curve dev

30,086 views • 3 months ago

New Short Course: Getting Structured LLM Output! Learn how to get structured outputs from your LLM applications in this course, built in partnership with .txt, and taught by Will Kurt, a Founding Engineer, and , Developer Relations Engineer. It's challenging for software to automatically parse through an LLM's freeform text outputs. Structured outputs—like JSON—solve this by converting natural language into consistent, clear, data that a machine can read and process. This course teaches you how to generate structured outputs while building several use cases, including a social media analysis agent. You’ll learn about structured outputs and efficient ways to generate outputs in your defined schema or format. You’ll begin by using structured output APIs, then use re-prompting libraries like “instructor” to generate structured output. Finally, you’ll learn how constrained decoding works; this is a very clever technique in which constraints are applied on each subsequent token generated, blocking any tokens that don’t fit your defined schema. In detail, you’ll: - Learn why structured outputs are important, how they allow for scalable software development, and the different approaches to generate them, including vendor-provided APIs, re-prompting libraries, and structured generation. - Build a simple social media agent using OpenAI’s structured output API, learn how to define a model's desired structured output using Pydantic, and perform basic programming with your outputs, such as importing structured data into a data frame using pandas. - Learn how to use the open-source library "instructor," which checks the structured output of the model and re-prompts the model until it validates the desired output, and explore the limitations of this approach. - Understand how structured generation by the “outlines” library works by modifying LLM logits, on a per-generated-token basis based on the desired format, to give a particular output structure. - Learn how regular expressions, which outlines works with, are represented as finite-state machines, and how they can be used to develop a range of structured outputs beyond JSON. By the end of this course, you’ll have broadened your knowledge of the approaches you can use to get structured outputs from your LLM applications. Please sign up here:
2:05

New Short Course: Getting Structured LLM Output! Learn how to get structured outputs from your LLM applications in this course, built in partnership with .txt, and taught by Will Kurt, a Founding Engineer, and , Developer Relations Engineer. It's challenging for software to automatically parse through an LLM's freeform text outputs. Structured outputs—like JSON—solve this by converting natural language into consistent, clear, data that a machine can read and process. This course teaches you how to generate structured outputs while building several use cases, including a social media analysis agent. You’ll learn about structured outputs and efficient ways to generate outputs in your defined schema or format. You’ll begin by using structured output APIs, then use re-prompting libraries like “instructor” to generate structured output. Finally, you’ll learn how constrained decoding works; this is a very clever technique in which constraints are applied on each subsequent token generated, blocking any tokens that don’t fit your defined schema. In detail, you’ll: - Learn why structured outputs are important, how they allow for scalable software development, and the different approaches to generate them, including vendor-provided APIs, re-prompting libraries, and structured generation. - Build a simple social media agent using OpenAI’s structured output API, learn how to define a model's desired structured output using Pydantic, and perform basic programming with your outputs, such as importing structured data into a data frame using pandas. - Learn how to use the open-source library "instructor," which checks the structured output of the model and re-prompts the model until it validates the desired output, and explore the limitations of this approach. - Understand how structured generation by the “outlines” library works by modifying LLM logits, on a per-generated-token basis based on the desired format, to give a particular output structure. - Learn how regular expressions, which outlines works with, are represented as finite-state machines, and how they can be used to develop a range of structured outputs beyond JSON. By the end of this course, you’ll have broadened your knowledge of the approaches you can use to get structured outputs from your LLM applications. Please sign up here:

Andrew Ng

89,792 views • 1 year ago