正在加载视频...

视频加载失败

People keep saying AI coding agents can only build basic, cookie-cutter apps. I decided to prove them wrong. For my first major public demo, I spent some time pushing Replit ⠕ 's AI agent to its absolute limit. The result? I rebuilt macOS entirely on the web. No templates....

158,392 次观看 • 7 个月前 •via X (Twitter)

71 条评论

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

Here's the iOS themed version which gets shown whenever this app is opened on an iPhone 📱

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

Special thank you to @zhenthebuilder for pushing me to make this video and @mattyp for advising me. and obviously @pirroh & @amasad for believing in me 🧡

adi 的头像
adi7 个月前

@Replit How much did this cost you? I love Replit but find myself leaving the ecosystem because it’s gets very expensive very quickly

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit I’ll find the exact cost tomorrow morning but it was built pre cost optimization & was still <$100

adi 的头像
adi7 个月前

@Replit Awesome! Curious to know the results I have been building quite extensively on Replit and have at few occasions been surprised at the cost (before the recent launch and optimizations!

cyrus 的头像
cyrus7 个月前

@Replit Dog shit web slop buddy sorry

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit You’re too kind!

Lando Toll 的头像
Lando Toll7 个月前

@Replit Yo chat, I think this guy is cheating. No way Replit can do all this

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit 100%

Brandon M. Crenshaw 的头像
Brandon M. Crenshaw7 个月前

@Replit I just shipped the Clawberry

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Love that!

Brandon M. Crenshaw 的头像
Brandon M. Crenshaw7 个月前

@Replit Making it open source once it’s done tonight. Encrypted agentic communication device, agents can text - call - email - run commands.

Rick Delashmit 的头像
Rick Delashmit7 个月前

@Replit This is incredible, but can it run Crysis? Seriously though … insane build!! 👏 🫡

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit I’m sure it could 🙃

Danilo Brizola 的头像
Danilo Brizola7 个月前

@Replit That’s AMAZING!!! So far the most advanced vibe-coded solution I’ve seen! 🤩🚀

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Thank you 🧡

Bik Nandy 👽 的头像
Bik Nandy 👽7 个月前

@Replit This is nuts

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Thank you!

Theo 的头像
Theo7 个月前

@Replit Do you have a github for it? @agi2asi

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit I’ll post it soon

Michael Grinich 的头像
Michael Grinich7 个月前

@Replit totally wild

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit 🙏

Brad Wilde 的头像
Brad Wilde7 个月前

@Replit Amazing! I’d love to hear a little bit more about what agent mode you used most (e.g., lite, economy, power, or max), and what was the overall cost to develop?

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit This was done in ~3 weekends with today’s equivalent of Power autonomy. I do not usually recommend going into Max mode unless you’ve spent a considerable amount of time setting up your docs, architecture, and design systems

Ruth Heasman 🌳🌷🦚🐉😃 的头像
Ruth Heasman 🌳🌷🦚🐉😃7 个月前

@Replit Incredible! 🤯 Build X Code as well and I’d pay to use it online!

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Challenge accepted. Xcode 26.3 with copilot or prior version?

Ruth Heasman 🌳🌷🦚🐉😃 的头像
Ruth Heasman 🌳🌷🦚🐉😃7 个月前

@Replit Dunno? As a PC user I currently don’t have either! You decide.

Somi 的头像
Somi7 个月前

@Replit okay the custom MCP servers wired to PostgreSQL and Stripe is what takes this from cool demo to actual infrastructure. how's the Siri voice control working under the hood?

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

I’ll admit that I had to run numerous tests when I got it to work the first time cause I had a hard time wrapping my head around it. Gave agent the AwesomeMCP GitHub link and watched it plug most of them into it. Felt magical. Siri is powered by Realtime via WebRTC to help with latency. Going to upgrade it to 1.5 tomorrow and give it another spin

Fares 🇵🇸 的头像
Fares 🇵🇸7 个月前

@somi_ai @Replit congrats you've already built a better Siri than Apple

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

Tbf, it doesn’t take much lol. I love the replica Siri so much that I’m now building a swift native replacement that’ll be powered by Realtime 1.5. All i really wanted was an assistant that could understand my intent instead of randomly calling the plumber when being asked what the weather will be like this afternoon

Faris Masad 的头像
Faris Masad7 个月前

@Replit Wow

Samuel Spitz 的头像
Samuel Spitz7 个月前

@Replit This is wild

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Grazie friend

ATusernameisATjim 的头像
ATusernameisATjim6 个月前

@samuel_spitz @Replit Well done that's epic

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ6 个月前

@samuel_spitz @Replit Thank you

David Riggs 的头像
David Riggs7 个月前

@Replit Replit is the best coding agents I’ve use in my oinion

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Couldn’t agree more

Neeraj Khandelwal 的头像
Neeraj Khandelwal7 个月前

@Replit Tim Cook after looking at this demo:

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit 😂

Blair Rorani 的头像
Blair Rorani7 个月前

@AndrewJDavison @Replit 👏🏽

Manny Bernabe 的头像
Manny Bernabe7 个月前

@Replit Cool!

Kevin @ swordslive.ie 的头像
Kevin @ swordslive.ie7 个月前

@Replit looks great - Mac os is a highly described system - I'm finding limitations is systems that are being invented.i.e. Are not highly described already. In this case I'm still having to think through the logic first myself and get the agent to iterate through it.

Benjo 的头像
Benjo7 个月前

@Replit @grok what screensharing/screen recording software is he using?

Sadi Moodi 的头像
Sadi Moodi7 个月前

@Replit This is wild. The real flex isn't the macOS replica - it's the MCP servers with Stripe/PostgreSQL/GitHub hooks. Most "rebuilt X" projects are visual clones. You actually wired infrastructure. Agent 3 + MCP is the combo that matters here.

Nav Toor 的头像
Nav Toor7 个月前

@Replit the "no templates" part is the real flex here. building a full os environment just by describing it proves we're moving from syntax to pure intent.

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Still amazes me to look back at my work from 2020-2021 and see how far LLMs have come. So much more on the horizon too.

TD 的头像
TD7 个月前

@Replit 👏🏼👏🏼👏🏼🚀

Jonny D 的头像
Jonny D7 个月前

@Replit DAMN! how did that cost bro. Opus???

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Around $100 but I’ll get the exact figure

Matt Beebe 的头像
Matt Beebe7 个月前

@Replit this is some inception level shiii

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit I’ll take some time to share a video of the copilot replica within VS Code replica building out a mini MacOS replica within this system. It’s very trippy to watch

Christiano 🦾🌱 的头像
Christiano 🦾🌱7 个月前

@Replit 2 bad it lies & the backend is always incomplete & after stating everything is done correctly and charging every time @agi2asi @amasad @ReplitSupport due to ur agents fuckup, you guys made me lose a client that was going to pay me to make websites for all their co. costing$1000s

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit @amasad @ReplitSupport Have you submitted a ticket? If not, please do and let @ReplitSupport know so we can investigate

Christiano 🦾🌱 的头像
Christiano 🦾🌱7 个月前

@Replit @amasad @ReplitSupport Along with your “agent” completely locking us out. Also your site said that we had to contact support and then I was told support could not help us?? This makes clients not want to use my services or you guys ever @agi2asi

Christiano 🦾🌱 的头像
Christiano 🦾🌱7 个月前

@Replit @amasad @ReplitSupport @amasad @ReplitSupport

Bala Sista 的头像
Bala Sista7 个月前

@Replit BLOWN AWAY

CryptoKong 的头像
CryptoKong7 个月前

@Replit I am positive that this code is not exportable, it can’t scale and outside of the specific backend Replit runs on—this would crash or fail to run on any other system. OS-level architecture is complex and difficult for a reason.

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit It actually is. I’ve exported it and run it on a very old Lenovo T450, a modded Pi 5 and an older MacBook Air. Happy to demo that if the community is interested. Total size is <500MB with all the modules. Considerably smaller prior to installation.

CryptoKong 的头像
CryptoKong7 个月前

@Replit Absolutely, you would need to show the hardware and then you going through the fresh install process.

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit No problem

David Hrubý 的头像
David Hrubý7 个月前

@Replit These fucking regards won’t sto pushing their tools ahahah

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Don’t mean to push another tool here but I hear Grammarly is pretty lit

iambenslater 的头像
iambenslater7 个月前

@Replit This is why Apple released the new MacBook Pro this morning with such excessive power. Agentic control baked into the OS controlling EVERYTHING, with a custom Apple LLM running locally. With deep Siri control.

Waseem 的头像
Waseem7 个月前

@Replit this is impressive

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Thank you

Alan Wunsche 的头像
Alan Wunsche7 个月前

@Replit 🤯 @Jason

Crazyox🌶️(微微辣版) 的头像
Crazyox🌶️(微微辣版)7 个月前

@Replit 这才是AI该有的样子

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Thank you 🙏 This is my first attempt at posting a full fledged demo and the encouraging words mean the world

inz 的头像
inz7 个月前

@Replit Meh

luca. ∆ИƉЯƐΛ 的头像
luca. ∆ИƉЯƐΛ7 个月前

@Replit Appreciate you

相关视频

🚨 this chinese guy makes over $1,000,000 a year… by building AI agents. no employees. no massive startup. he just keeps building. while most people are still asking ChatGPT random questions, he’s using Claude to build software that solves real problems. this is what people call vibe coding. he opens Claude and says: “build me an AI agent for real estate businesses that creates property videos.” Claude writes the code. builds the interface. adds subscriptions. helps deploy the app. within a day, he has a working product. then he starts building the next one. that’s the part most people don’t understand. he isn’t trying to build one billion-dollar company. he’s building dozens of AI agents, each solving one problem for one industry. → an AI agent for dentists → an AI agent for ecommerce brands → an AI agent for podcasters → an AI agent for real estate businesses each one automates work that people normally do by hand. each one is built with simple prompts. each one can become a real business. the crazy part? you don’t need to be a software engineer anymore. you need to know how to think like a builder. how to spot problems. how to explain solutions to AI. and how to ship. that’s exactly why i’m reading this article: “How to Actually Build Your First AI Agent.” because this is the skill that’s creating the next generation of builders. the people who learn to build AI agents today won’t just use AI. they’ll own the tools everyone else ends up paying for.

MIKE

39,041 次观看 • 3 个月前

New Course: ACP: Agent Communication Protocol Learn to build agents that communicate and collaborate across different frameworks using ACP in this short course built with IBM Research's BeeAI, and taught by Sandi Besen, AI Research Engineer & Ecosystem Lead at IBM, and Nicholas Renotte, Head of AI Developer Advocacy at IBM. Building a multi-agent system with agents built or used by different teams and organizations can become challenging. You may need to write custom integrations each time a team updates their agent design or changes their choice of agentic orchestration framework. The Agent Communication Protocol (ACP) is an open protocol that addresses this challenge by standardizing how agents communicate, using a unified RESTful interface that works across frameworks. In this protocol, you host an agent inside an ACP server, which handles requests from an ACP client and passes them to the appropriate agent. Using a standardized client-server interface allows multiple teams to reuse agents across projects. It also makes it easier to switch between frameworks, replace an agent with a new version, or update a multi-agent system without refactoring the entire system. In this course, you’ll learn to connect agents through ACP. You’ll understand the lifecycle of an ACP Agent and how it compares to other protocols, such as MCP (Model Context Protocol) and A2A (Agent-to-Agent). You’ll build ACP-compliant agents and implement both sequential and hierarchical workflows of multiple agents collaborating using ACP. Through hands-on exercises, you’ll build: - A RAG agent with CrewAI and wrap it inside an ACP server. - An ACP Client to make calls to the ACP server you created. - A sequential workflow that chains an ACP server, created with Smolagents, to the RAG agent. - A hierarchical workflow using a router agent that transforms user queries into tasks, delegated to agents available through ACP servers. - An agent that uses MCP to access tools and ACP to communicate with other agents. You’ll finish up by importing your ACP agents into the BeeAI platform, an open-source registry for discovering and sharing agents. ACP enables collaboration between agents across teams and organizations. By the end of this course, you’ll be able to build ACP agents and workflows that communicate and collaborate regardless of framework. Please sign up here:

Andrew Ng

105,758 次观看 • 1 年前