Riley Brown

Hyper bullish on OpenAI and Anthropic. No other models worth using right now.

Okay after further testing Muse Wonder might actually be better than Claude Sonnet 3.7

The only benchmark i care about now is my glide benchmark... I'm done with these color coated model cards.

I don't wanna hear "yo bro you gotta use it in the mumbo-jumbo harness with max effort only blah blah"... I been using claude code with opus 5.5 for 3 weeks straight never changed it from high effort. Opus just glides to where i want to go. I can't remember a SINGLE time where i got frustrated unless it was me making a clear blunder on instructions. Opus 5.5 doesn't make mistakes, when it does it realizes right away and fixes them without asking. GPT 6.1 Sol is right there too. It Just lacks the visual awareness and creativity of of Opus 5.5... Muse Wonder is not even in the same league. So I guess Zuck, Wang, Theo and I disagree.

Used Muse Wonder for 5 hours. Was gonna make a video. No need, back to codex and claude code.

Spent the last year going hard on long form... Which caused me to slow on short form. Now i'm locking into short form again. Off to a good start. (also hit 300k on YouTube)

This is wild.

RT Emily Lambert I used Claude Opus 5.5 to design a little blueberry home for one of our Mac minis. The phone is its face, so the enclosure needed a built-in stand. Getting there meant moving the phone so it didn’t block the Mac’s ports, changing the backrest to clear the camera bump, and figuring out how to print it with the filament I actually had. I only had one 250g spool of blue left so I split the design into blue parts and separate green leaves and trim (leaf skirt), then I printed and assembled them on our Bambu H2S. The tiny leaves went in with tweezers. They’re such a small detail, but they make the whole thing and it's so cute.

absolutely, undeniably sick.

Zuck Released a new model. Here's the Opus 5.5 Summary.

RT Agent Native Zuck just released a new model... Called "Wonder" Their focus is on personal and professional agents, not necessarily coding. This model is available - Via API - Via Muse Code - (Soon) inside OpenCode, Devin, Cursor - (Soon) inside the Muse app (paid subscribers) What Meta focused on with this model: - Meta wants this model to work well in the real world. - They want this model to have better active memory, and remember what's important to the user For Muse (their agent app): - It's better at remembering habits - It doesn't give up, it keeps trying things until it achieves its goals

RT Agent Native Went into the fridge, and… All the Watermelons Argon.

We are at 1% of what AI gaming will lead to. Making a ton of money making games used to be about Capital and technical abilities, plus a little cultural understanding. Soon it will basically be all about the culture. Understanding what people want, and then making viral content playing that game.

I'm building an internal super app for my company. What I love about building my own software: I can add any damn toggle I want. Toggles that would never ship in software sold to millions because they're too niche. Because... the niche is my company. The software is mine. I don't need permission. So I can make every action an employee takes PAINFULLY OBVIOUS... AND REQUIRED. Yes, there's a learning curve to using Claude Code to build software for your company. But it beats teaching every employee how to use someone else's software. Notion is the example (sorry, I pick on them a lot). If Notion is your company's database, notes and wiki, you have to teach everyone Notion. Databases with pages with databases with linked pages... the list goes on. When I build software for my company, the goal is to make the software irrelevant. Only the task matters. Claude can change anything, but employees only see what they need. Notion was the first notes app to kind of expose the main primitives. Everything is a database with pages with blocks etc. and that comes with a learning curve. Convex / Supabase is one layer below that. Give Claude a Convex key and it can build any app you want. When i want something from Notion... I Add it. From any other app? Add it. 12 months ago my workflow included Notion, MyMind, Excaladraw, Typefully / Buffer, VidIQ, and Google Drive. Well... I took all the features from those apps that I liked and added it to my own company super app "Native Note". (Still have yet to touch a line of code in my life btw)

Anyone who says software is dead and that AI will kill apps is wrong. I’ve been downloading apps like crazy.

Codex computer use is SO good. I actually can’t live without it now.

RT Jen Why I keep pleading with Dems and progressives to stop pretending AI is a fad like NFTs. If we don’t understand this tech we can’t govern it and tech bros will. Video explainer of Andrew Karpathy’s tweet today, the cofounder of OpenAI (ChatGPT), now at Anthropic (Claude).

I've sent this tweet to like 10+ software engineers and they seem deeply bothered by it lmao. https://x.com/growing_daniel/status/2108679983787016256?s=20

yes.

Here's Opus 5.5 with a visual reading of Karpathy's latest tweet. Watch on 2x speed. (one prompt btw)

And yes this works with any agent. Codex, Muse, Instinct etc.

RT Agent Native Connect Claude Code to GrokBot with a Webhook https://agentnative.inc/resources/grokbot-claude-code-webhook

RT Peter Orban 18 agent updates in 20 minutes, from Claude and Codex to Google's latest. Massive week for builders, this recap is pure fuel!

If all it takes is 3 hours to prompt it into existence people will not be paying for it. It will be turned into an “artifact type” by the labs. Anthropic just launched “Motion” as an artifact type which is a collaborative motion graphic video editor. GPT sites has built in storage, auth, db, hosting and more. Claude Design will just send you a figma style board you can edit when you need a design. These are all so unbelievably useful. OpenAI just released GPT space. Right now it’s pages and sites, soon it will be any file type that you might need to create in a company, built into the platform.

Open-Descript Coming Soon.

All software is de facto open source now. Might as well be open source and get the benefits.

Massive week in the world of agents. New Updates to Claude, GrokBot & Codex. Here are the 18 biggest updates in 20 minutes. (Details and links down below) 00:00 Intro Claude 00:19 Claude Dashboards & Claude Motion 02:30 AI Video Editing with Opus 5.5 04:01 Claude Projects Rolls Out 06:08 Haiku 5.5 Released OpenAI 06:45 28 Days of Codex 08:21 ChatGPT Intelligent UI 09:20 OpenAI Math Breakthroughs GrokbBot 09:58 GrokBot Now Uses All Models 11:28 GrokBot Team & Primary Bots Google 12:49 Nano Banana 2.1 13:41 Gemini 4 Argon Rumors 14:07 Antigravity More 15:34 AI Is Recreating Adobe Software 16:50 Video Games Reverse-Engineered & Merged 17:23 Play Console Games in Your Browser 18:09 New Personal Agent Products 19:33 Hermes Agent Raises $90M 20:35 Outro

Every AI agent hack from 3 months ago is already obsolete.

The future of software is open source, ship at lightning speed, free offer to build huge community, backed by founder who understands content. Theo w/t3 code is great example.

If you want to feel what it’s like to have a personal agent platform created by @AnthropicAI go to the Claude app right now and create a new Claude Code Project, name it Tony or something. Use this prompt. “this project is responsible for moving my life forward, you are to help me get things done in order of how important they are. Analyze all my connections including email, slack, and any platform and suggest threads and automations that we can set up now. One of your threads should be called the Organizer, which does the prioritizing and planning, which keeps an up to date editable document of our todo lists and important documents.” Projects is very good at delegating. That’s what it was made for. There’s a coordinator chat, and then threads (sub agents). What’s cool, is these sub agents can be cloud or local. And since I have my MacBook and 2 Mac minis running 24/7 some of the threads run there. But the coordinator thread is in the cloud so it’s always available. Your coordinator thread can be Opus 5.5, but threads can run different models. Just ask for haiku or sonnet if they are responsible for very routine tasks.

Besides getting faster what do you want the next models to get better at? I’m genuinely curious where do you think Opus and GPT 6 fall short rn?

let's see how this goes.

Ohhhhh my god i can't believe how visually creative Opus is... i swear Anthropic took gave this model vitural psychedelics or something.

RT Agent Native I don’t get it. Where is it?

I highly suspect the @X team will try and make this big feature on the platform. X really is becoming the everything app.

So many technical people ask me how to grow on YouTube and my concise advice is this. Make a screenshare video explaining something that you find (1) interesting and (2) helpful and close to the edge of AI capability. - Start the video showing you and the screenshare. Show some visual on the screen on an excaladraw board. if the video is showing how you can build internal tools with claude code show a claude logo with an arrow pointing at a high quality internal tool screenshot. - You don't need a highly edited intro. Just a good hook and visual that communciates what the video is about. - Your hook should follow what i call the "BRENS" Framework. Big, Relatable, Easy, New, Safe. 1. Big - It's a big deal 2. Relatable - find a topic to latch onto that many people relate 3. Easy - "It's not as hard as you think if you understand these 5 things" 4. New - This new thing CHANGES how we do ____ 5. Safe - I'm credible, you're not going to waste your time. You can say "i'm rick and i have 10m dollars" or you can hint at it in various ways. As concisely as possible try and get at least 3 of these in the intro. Example: OpenAI [relatable topic] just released astra [new] and everyone is talking about how powerful it is [big]. And i've spent $3000 on tokens trying to figure out how to implement it into my business, and today i'm going to share exactly what i've learned. [safe] It basically comes down to 5 things [easy] that i've learned which I'll show you today. THEN IMMEDIATELY cut to the first one without wasting any time... Then introduce some structure - visually. (see the image) This was the number one video released on Astra (in views) besides Then... think to yourself... what questions are most people going to be asking while listening to your intro. Write those 10 down, then narrow it to the top 3-5. Answer those questions as early into the video as possible. And then go to the first idea. And use rehooks when you can. "The first thing that ....." ...

1000 subs in first 4 videos on 2nd channel. Basically I just yap, then let ai edit and post the video automatically, and I make adjustments to the workflow slightly every day. I actually don’t know if the editing (besides cutting) is necessary, I think people just want more raw content.

The reason this role is so hard to hire for is because these people are making 7 figures as independent creators working on their own time following their own curiosities.

I think print is going to have a huge comeback. I’m tired of reading on a screen.

RT Agent Native Want to keep track of the most powerful agent tools? Introducing Agent Native Landscape and Agent Updates. Many more resources coming. It's free, forever. https://agentnative.inc/landscape

Video editing is going through what coding did in Jan of 2026. I'm actually blown away right now.

Okay I’ll bite… what does “decompile” mean?

RT Agent Native Anthropic just launched Haiku 5.5, and by looking at the benchmarks, it’s better than Luna. Testing right now. The only 5.5 model we don’t yet have… Is Fable.

True…

A computer is a completely different thing now. I’ve reached a point where what I want, I get. I don’t need to ask any developer, I just ask ai to do it. As we speak codex is building a new Mac mini case via 3D printer with a phone mount… for bluey. Why am I doing this? because I can! It’s fun. and Opus 5.5 is making improvements to my internal tool app (which has 6 apps built within it). I’ve cancelled 6 tools because of this.. and it’s better for our workflow. When I need keyboard shortcuts changed in any app I just ask codex trusting that it’ll figure it out with computer use. I wanted a cool background animation app for my dual monitors which are in the background of my videos, it never once occurred to me to go look for an existing app. I took the 40 seconds to yap into my computer and it made a desktop app pinned to my dock that I can use for cool background animations. When I want to go outside and continue working I CAN. I can put in air pods and talk to an AI in real time voice that can spin up codex sessions. While walking I can have that AI summarize all the emails I need to respond to. And then the AI can actually respond based on what I tell it. I can have it go through all my files, notes or videos on my computer for context. This are just examples of the past 3 hours. Every day I’m surprised by what is actually possible by talking to a computer.

This changes things...

I'm building a resource library for agentic video editing, fully in the open that i will be using to fully automate my 2nd youtube channel's video editing. Here are overlays / motion graphics that you or your agents can use when editing videos. Will be making improvements + adding hundreds more overlay graphics that you and agents will be able to use as examples when editing talking head content. https://agentnative.inc/resources/video-graphics This will always be free. Will be adding a ton more resources for video editing here including scrapers, logo libraries, curated music that's good for videos and more... completely free. just started this an hour ago, so be patient.

I want all my agents to have the same skills with access to every useful api - every ai model every scraper everything, and I want to be billed on one account. Too hard to ask?

2 years ago we’ve come so far.

Can you please add the ability to add images and videos to comments

ok let's do it.

Locked in on my video editing workflows. Opus 5.5 high is cooking. Back to using Premiere Pro. I don't like Davinci Resolve at all.

RT Agent Native Every major Agent product is moving from many chats & agents to a single agent orchestrator. - GrokBot now has a team leader - Claude Code Projects has an agent that delegates to threads - Dot can Delegate to codex

GrokBot recently dropped 4 new features that I found incredibly useful in my business... and not enough people are talking about them. 00:00 Intro 00:20 Primary / Main Bot 07:11 Team Bot 14:30 Inline Mini Apps for Slack and Gmail 17:40 Bypass Bot Blockers 19:37 Bonus Tip

This is awesome…

Using Frontier models like Astra and Opus is like Coaching Kobe. Good motivational speeches about the vision, good spacing, have everyone get out of his way.

RT Agent Native This week will be a gem. Buckle up.

Can anyone from @Anthropic plz tell me when the new Claude Code Projects will be rolled out publicly. I’ve made 2 videos talking about it and I have hundreds of DMs asking how to get access…

Spent the entire weekend building. Will be open sourcing all @agentnative_ projects. 5 videos next week as well.

Like this post if you want to see a touchdown.

I agree - when a task is of high complexity and importance I’ll use Astra. But for most other stuff esp frontent / documents and most coding tasks opus does the job well. And feels unlimited on the 200/month plan.

who's building this?

RT John KM (YouTube Strategy) This @rileybrown video was a 9 of 10 early. Now it's a 2 of 10 and tracking for 100k+ Here's how I did it 🧵

Yes

Vibe Coding has completely evolved. It’s gotten so good at vibe coding “SAAS apps” that it’s no longer interesting. Now, people are vibe-coding full multiplayer video game mods, adding agents to old hardware devices, and 3D printing any object they can imagine using Codex and Claude code.

RT Sean Collins-Cruz Riley shipped Bluey in an hour last night. I thought it was awesome and decided to make a version for my fully mobile workflow! Riley, you were legitimately my FIRST follow in the ai space! You always had the vision for cool projects and efficient workflows! Keep on pushing! 🫡

RT Alexandr Wang modretro 🤝 muse @PalmerLuckey load up the chopper

i turned my ModRetro into a smart device. by plugging it in and asking codex "is this device compatible with https://gadgets.muse.ai/"

Bought an Apple display monitor with my Twitter (x) debit card and it didn’t feel real. Like it was Monopoly money.

Got Muse on my @modretro chromatic.

Building something cool.

YESSSSSS

woah woah hold on let's talk about this. what if... hear me out.... what if we added an ADDITIONAL left side bar, and a new an improved bottom bar?

Is there an API for a physical magazine? Like I want to be able to create a PDF and then send it via API then have it be delivered to my house in like 3 days as a magazine. Who’s building?

i'm so glad i came to NYC

So many ideas, wow.

Yes.

“Yet” assumes that all households will pay. What if peaks at 3%?

I don’t think people realize what’s about to happen. Between this and ai being able to mod any video game, and remix them, nearly perfectly in a few prompts.

Re Here’s the video https://youtube.com/post/UgkxUrB7aYtb_yppDvBB0ZkzNASVvnFg3kzU?si=dj8lkDZsJ3NNydXq

Is this speculation about “routing to fable 5.5” or is it legit.

i have noticed this as well.

Specializing in a skill that’s done on a computer is over. You need to rapidly expand to learn the whole stack of whatever you’re doing, all the way to storytelling and sales.

With a proper setup. You could get Opus 5.5 to create the top 50 apps in the utility category in a single prompt.

Usefulness: 10/10 Ease of Explaining the Usefulness: 1/10 As i plan my videos talking about Dots, i'm confronted with so many challenges. I'm trying to think about how to explain this agent but there are SOOOO MANY LAYERS. It's very hard to explain to someone, unless they are technical. Almost noone knows what a "vm" is. And when you say the agent has a computer they don't understand why that's useful or interesting... Also the agent can spin up Codex Sessions! that's awesome!!...... if you 1. know what Codex is 2. know what a session is 3. know why codex is useful One of the hardest parts about agents right now is EXPLAINING them. Every little feature, every layer, makes it much harder. But that's what makes educational content fun i guess.

RT Agent Native 4 safety researchers just left OpenAI. Their next model should be really good.

Spent 24 hours using my Dot (Bluey) - The voice experience is the highlight. Truly remarkable. (As you approach 30min+ I’ve run into problems) Here’s an unfiltered list of feature requests and thoughts - make a better view for seeing codex threads it spins up. I think paying attention to the way Claude does Claude Code Projects (new version) would be useful. The way you can spin up new threads is quite nice. But the “projects” analogy doesn’t work. An agent (like dot) makes more sense. - being able to voice call multiple dots and switch between the dot your talking to would be really cool. Especially if they have different voices. This allows you to do something while it’s working - better animations when working. Sometimes hard to tell when the dot is working. - simplify the vocab around browser and computer - for instance, I asked my dot to use the browser… and it asked me “my vm browser, the codex browser, or chrome browser?” it was at this moment where I was like, omg there’s ALOT going on here. - however I was skeptical of dot being in the same app as chat, and codex. But now I’m starting to see the vision. It’s going to require ruthless prioritization and cutting surfaces. This is more codex feedback - even my most technical friends are telling me that chatgpt desktop is too intimating. - tbf I remember just 9 months ago I was like OpenAi is overcooking with Sora, Atlas, etc then they did 2 months of prioritization and they made codex… so I expect that’ll happen again now.

RT Riley Brown Re NYC 10/10 AUSTIN 8.5/10 SF 4/10 Remember, my view of SF is warped by startup culture. Some thoughts that come to mind. I could see myself moving back to Austin when I have family. For me I value working remotely, walkable city, meeting creative people. NYC is great for that. Austin is good for that if you live center of downtown which I did. Austin the people are so nice and so detached from work and want to have fun. Which is better than SF, where people have a frantic existential dread and constant worry of perminent underclass. At least in tech scene. There are lots of rich people in NYC, SF and to a lesser degree, ATX. The difference is that SF rich people are unbearable. NYC has a lot of cool rich creative people. Food in NYC is unmatched. Austin has good food. SF food is mid at best. Nothing is open late in SF and ATX. NYC People do have a frantic pace, almost culturally embedded… but it’s a buzz. It feels like a property of the city. Whereas in SF there’s no buzz. Just GRIND. GRIND GRIND. GRIND baby. NYC people aren’t “nice” like in ATX, that’s not the right word. They are not overtly friendly. This is really cool about Austin… people are just really nice. but in NYC people look after each other. Theres a city pride to NYC. Like they have a duty to look after each other. People are “nice”… in a NY way. It feels the most real. In SF i felt none of this. People aren’t nice or proud to be there. Again my opinion could be warped. Age anxiety. The people I know in SF… in their mid 20s feel like they are very old. One of my best friends there is 22, whole life ahead of him. “Feels like he’s very far behind”. The startup scene does this to people.

Okay just called my dot for 45 minutes and handled all my email / slack tasks for the day… phone in pocket, AirPods in. I have so many thoughts, making a YouTube video today on everything I’ve learned / hopes for new features etc

I’m so glad I left SF.

Meet Bluey. He’s my pal.

💯

My 3D printer is here. Time for Opus and GPT 6 to cook.

Yo can i get access Logan

RT Agent Native Big release from Gemini. Still one more big announcement coming this week.

hey @OpenAI my team is now using Spaces and Pages... how do i make it so that i can share it with anyone with the link.

RT Dan Shipper turns out media is a good business :)

just out of curiosity... would you rather separate or all in one?

It’s working.