The critical element of these systems is that it takes you less time to get to production worthy code using the tool than it would without it. You can't just stop a demo at the point where you don't have working code. Anyone with a ChatGPT plus subscription can ask it to generate react components from a mockup, but it doesn't do a great job. You have to show your tool <i>doing a good job</i>, not just flipping through component files.
Aren't all of these thin branded clients around chatgpt extremely high risk ventures for both the company and the customers? OpenAI can change their prices, or their terms whenever they please and the crunch _will_ come in the future when they have to make back their money they are currently throwing away to grow.
The namespace is overloaded. This 2022 paper decsribes a method for looping LLM generations in a specific way to generate more reasonable decisions and outcomes: "ReAct: Synergizing Reasoning and Acting in Language Models"
<a href="https://react-lm.github.io/" rel="nofollow noreferrer">https://react-lm.github.io/</a><p>This is the type of agent that AutoGPT uses.
Millions of frameworks and hype tools stacked together now regurgitated by a chat bot that can't reason. All to bring you more generic sites that can't run on processors with billions of clock cycles per second. Web "developers" should be embarrassed.
Hey everyone, happy to be featured here and see your thoughts!
I'm the original author, I was curious to find a way to generate production code with LLM, and I did find a path to this, it requires a lot of work and engineering, solving a complex problem by splitting it to multiple smaller ones that an LLM / LLM agent can take.
IMO it is doable, but it requires a lot of work and potentially funding, so I left this project to do something fundable in this space :)
I wish I could keep working on it full time, if anyone wants to help doing that, or ask anything, reach out!
You could easily setup the same with <a href="https://formula8.ai" rel="nofollow noreferrer">https://formula8.ai</a> - even more flexible. Just write a React Component Formula.
Very cool to see, I can't wait to use something like this especially once someone automates the loop of write -> run code -> check errors -> fix errors for an LLM
We’ve done experiments with generative UI code, but we keep coming back to the LLM making crappy looking UI, and if the engineer still needs to tweak the code afterward this is just a glorified copilot. Without more specialized models, I believe the next step is an intermediate design language…something an LLM can’t make look bad. Lots of room to innovate here though. Really glad to see people working on this problem.
Just a heads up, the video playback was 2x by default so I couldn’t understand anything until I looked at the settings. I don’t think it’s a previous setting from me because I’ve never used loom before.