behzadhaghgoo

You can ask it to do either way. I think for this it did some research and also actually tried real figma in its browser. However, I think ultimately these models are trained on these products. It seems to know it already pretty well.

here I added a button to the editor so you can see code. or you can use this link: https://www.mana.space/home/rZfbotNWAXegQypmpkzaAPFC6Mp1/App...

I think it's less about the code and more about the functionality. I didn't guide the llm to add features etc. I just told it don't stop until it's feature complete. Even 2 months ago I'd have to hand hold it a lot with UX and detailed features.

we have a different structure than repo but I'll share the code in a minute

added

let me add :)

I think it's one thing if it's possible, it's another level if you can just one shot it. I agree usage matters more but it's a paradigm shift. Even one month ago you couldn't do this I think it's specifically opus 5.5. Astra or Fable 5.1 can't do it as far as I tested

sorry made it editable

you can try it here: https://www.mana.space/home/rZfbotNWAXegQypmpkzaAPFC6Mp1/App...

you can try the editor here: https://www.mana.space/home/rZfbotNWAXegQypmpkzaAPFC6Mp1/App...

I thought the same, please try it :)

My prompt was literally make something figma-level, have feature parity and do it in a way that a user can't feel a difference.

You can try mana, curious how it compares to your previous experiences. You can sign up here if you like: https://mana.space/invite/MANA-NPNV

I think one thing that's different now is that you have an agent that you can ask to handle those edge cases. UI/UX solves the very common paths and there is always a very powerful AI for actions and modifications to app.

Very curious if you try it and tell us how far it is from real figma

we can try :)

I think if you do this outside mana it's 2-3x amount of code. You don't need to replicate all of mana. I'm bullish on mana making general figma level apps possible in one shot but I think for any specific one you can build your own. It's more about how good models have gotten in understanding UX and writing complex code

This was from scratch

It’s not an ad just try it

yeah maybe they trained on the RL envs or maybe it's from computer use

One way to think about is like Notion. Imagine if every notion page was a webapp instead of a markdown and then you had an LLM that had access to it all. So anyone who uses productivity software can potentially use it to make their very personal workspace.

If you give it an API to a local file system it can already connect. For the main file system we can explore it! Curious what your use case is.

Currently our file system is a relational db which makes it more web-native for webapps. but fully local can be doable too.

I think the main thing is UI and being on the cloud. I used to have all my todos and notes on claude code. But it's much nicer to have a UI that can directly interact with them (e.g. mark something as done or click to view more details). With CC you'd need to spin up a server and frontend and host them etc. Here we abstract those steps away without really reducing capability. You can create a frontend and it can talk to the file system or external APIs very easily.

yeah everything persists. The apps and LLM both talk to the same source of data. So you can either talk to change things or have apps do it.

Yup, basically like Claude code operating on a folder but the folder is on the cloud and it can generate interactive UIs for interacting with the folder.

If you combine an LLM that can run code, a cloud file system, and UI generation, you can create a new kind of os that evolves dynamically with the user.

Here is a clone of hackernews in 2 prompts: https://www.booknote.cloud/editor/AMGjCiDcVQyt206NibH48WOViT...

Hey Hackernews,

A friend of mine and I have been thinking building a general purpose environment for GPT so it can generate code on the fly and run it. Imagine if you could have JS frontend and Pyodide-based Python backend in a notebook-like environment. You have python libraries running in the browser, react, can call any API, navigate the web, ... . And given the current code-gen quality, it works with no debugging most of the time. This way normal users can generate micro-tools for their use-cases on the fly.

We have been surprised by how generally capable it is. From making simple calculators to handling complex workflows like doing OCR on data and syncing it with firebase. It's not quite ready for a public release but we have a few videos and would love to know what you think!

Here are some examples: (note that other than passing some api keys, we have done no hardcoding what so ever, it's SO GENERAL!)

Doing OCR on business cards and creating salesforce contacts: https://youtu.be/yC3Crrav-jg

Interacting with Firebase and creating a custom frontend with fuzzy search: https://youtu.be/EgSkn6vUVJg

Data-cleaning and basic datascience: https://youtu.be/TD-1orzQyuU

We also have a waitlist and will start onboarding people in 2 weeks. If you are interested sign up here: https://booknote.cloud

When they release an api we would charge our users and pay them

MULTI.ON is a chrome extension that automates your browsing and acts as your AI assistant.

It can help you with finding any content on the web, ordering food on doordash, shopping on amazon, post on social media like Twitter, Linkedin & Messenger, fill out any forms, give shopping opinions. Basically any task you can imagine.

You just input your goal and it controls the browser to achieve it. We didn't tune it for any specific task so your imagination is the limit of what it can do. And it generalizes to tasks much better than we expected!

It was inspired by https://github.com/nat/natbot/blob/main/natbot.py. We made it into an extension, switched to ChatGPT and improved prompting and parsing significantly.

Sign up and we will send you the extension as soon as we are done with final polishes! https://multion.ai

This is so cool!!!

Thanks! Oh where is the button hijacked?

Right now everyone runs and compiles on the same computer. Basically the browser editor connects to that computer and lets you run vscode.