Miguel Ángel

just a happy boy from a small town in Spain, here in SF thanks @HeyGen_Official @joshua_xu_ @liu8in for making it possible ❤️

@thsottiaux another bank reset?

Tomorrow might be 8M active user celebration day. Just saying

@antoniosarosi 🐐

Years ago, before AI, I wrote a memory allocator in Rust from scratch. It was the first time I actually did serious systems engineering. The first time I built something that composes itself from the ground up to build something much bigger. This here is a "Bucket" 🧵 1/9

really great video made with @HyperFrames_, congrats on the launch @jsonphile!

@zaidmukaddam did you use hyperframes?👁️👁️

Apple had a chance to build this in two years, but they didn’t. So I did. Introducing swift-ai-sdk. Ship real AI in your iOS app tonight! → swift-ai-sdk.dev

@trevin @HyperFrames_ @threejs yes

@Miguel07Code @HyperFrames_ @threejs Need you to try and figure out threejs + illo + hyperframes 😀

comedy club made with @HyperFrames_ + @threejs using GPT 5.6 Sol 🤣

@gavinpurcell @HyperFrames_ @threejs fable 5 cooked gpt tbh

@Miguel07Code @HyperFrames_ @threejs haha nice work

@gootecks @HyperFrames_ this is the second time 🤣🤣

@Miguel07Code @HyperFrames_ Wow is this the first external hf full time job posting???

if you are proficient in @HyperFrames_ you can get a job!

cc @Rames_Jusso

@monocleaaron can you share on DM a google drive link with a zip containing the full session of the coding agent to debug it pls?

@Miguel07Code Yes help pls - my motion is so bad.

If you are a HyperFrames power user and you want to have a call discussing feedback to improve HyperFrames, feel free to DM me with your email! However if you have specific feedback and you don't want a call, just DM'ing me that is fine too!

@dhasandev insane!

I recently interviewed an intern; here are the questions i asked 1. How do you evaluate AI agents? I asked how he evaluates agents: what libraries and frameworks he uses, and how he thinks about reinforcement learning, environments, and evaluation systems. The real question was about scale. How would you evaluate three to five agents? And how would you evaluate three to five thousand? System design matters most when you work with agentic systems. At that scale, humans cannot manually evaluate every agent and every trace. You need to build a system, potentially agents themselves, to do that evaluation work. 2. How would you replicate a product like Notion using today’s technology? This is a product-engineering question. I am not looking for someone who simply knows what tools like Codex or Claude Code are. I want to understand how they would use them to solve a real problem. That starts with discovery: understanding the product claims, the features, how they connect, and how the system would serve millions of users. Then: how would you use modern coding agents to build that system step by step? How would you evaluate whether it works? What would the development workflow look like? Every engineer is developing their own workflow for these general-purpose tools. I want to see how a candidate thinks through that workflow. 3. How would you manage an agent system as it grows from five agents to thousands? Imagine a company starts with three to five agents. Then it has ten, twenty, thirty, and eventually thousands. Some agents handle security review. Others review legal documents or triage bugs. How do you scale the system that manages them? The older approach was role-heavy: tell the model, “You are an auditor,” “You are an implementer,” or “You are a reviewer.” A more scalable approach is to keep personas generic, such as explorer, reviewer, or worker. Then package the actual work into skills, and compose those skills to create the behavior you need. That makes individual skills easier to evaluate, and it makes the agents themselves easier to manage as the system grows. These questions help me understand where a candidate stands in their understanding of the current market, modern engineering workflows, and the systems required to make AI useful at scale.

@SolutionsJoeG @trevin @HyperFrames_ cc @Rames_Jusso

Just made the switch from Remotion to Hyperframes, the one thing most people don't understand about it ... small edits can be made directly in the studio! I had a graphic position that I wanted to move, and in a few clicks and seconds, it was done and ready to preview! This is huge for saving tokens! Loving it so far! What I am still trying to figure out though, is the Story Board .. if you've a got a video or article on it, point me the direction!

Attention all @HyperFrames_ power users, take @Miguel07Code up on this offer to give feedback!

@gavinpurcell crazy, this can become a benchmark, the pixar benchmark

well, i connected GPT-5.6 Sol to Blender and... let's just say the world of 3D graphics (at least via Blender's MCP) is safe for now... but still... this whole scene was generated by GPT-5.6 with very little human intervention feels like *EARLY* pixar, like too early lol

@Rames_Jusso @HyperFrames_ 🤣🤣🤣🤣

New @HyperFrames_ use case unlocked @Miguel07Code should we see how gpt 5.6 fairs? Maybe a two part comedy special with gpt 5.6 and fable 5 roasting each other? 🤣

@trevin @HyperFrames_ ❤️

@trevin ☠️☠️☠️

uh what the hell. This is saying the DEFAULT OPT-IN is sharing our entire code base?!

@xavier_mitjana has probado @HyperFrames_? te podría ahorrar mucha pasta

Seedance es el mejor modelo de video. Pero mi experiencia me dice que el combo Kling + Omni es el más útil para crear B-roll para youtube. Y más ahora que con Omni puedes trabajar sobre tus propios clips. Aquí podéis ver la intro de mi último video ANTES de pasar por After Effects para añadir rotulación. 👇

@Imad_Phelps @HyperFrames_ man cooking so hard it created a fire in the building

Me Linting Ep. 9 of @HyperFrames_

I will keep saying it if you’re sleeping on @HyperFrames_ you’re missing out. It’s free and has so many use cases. You will thank me if you try it. It’s been super helpful for me.