I am the founder of one of the very first no-code/low-code platforms (DaDaBIK, first released in 2001). Three years ago I released appifytext.ai, an AI agent that develops DaDaBIK apps starting from textual specifications.
Low-code and LLMs can coexist: low-code can be just another layer (or, if you prefer, a more abstract programming language) that LLMs can use. You have less freedom, but more predictability and robustness, which is perfectly fine for internal tools.
At the moment I am focus on understanding what kind of applications users want to build. I thought it would be easier to communicate the type of application that appifytext can manage well (CRUDs, internal tools and similar).
Users can already export the application they create on AppifyText, import it in DaDaBIK and host it anywhere, so from a certain point of view it is already "on the market".
AppifyText (https://appifytext.ai): I took GPT-3 and a no-code/low-code platform and I made them cooperate.
The system gets in input the description of a desired application in plain English and produces in output a ready to use Web application in less than a minute. The focus is on CRUDs, simple internal tools, personal databases & similar apps.
Hi HN,
this is a proof of concept, the result of a few weeks of work.
The idea was to create a system that gets in input the description of an application in plain English (ideally, its specifications) and produces in output the actual, ready to use, Web application in reasonable time (less than a minute). The focus is on CRUDs, simple internal tools, personal databases & similar apps.
I took a generative AI model (GPT-3) and a no-code/low-code platform (DaDaBIK) and I made them cooperate.
Users provide app descriptions as input, which are then processed by the AI to generate an output that is tweaked and passed to the no-code platform to actually build the app. The resulting app includes CRUD features, automatic relationship handling, some data validation, and additional features like chart generation. The app also contains realistic examples of data to help users get started.
I know there is room for a lot of improvement and that the resulting app is not the state of the art in terms of UI/UX but considering the results I got in such a short time I think there might be potential.
It's a beta and there are some corner cases that can generate errors, but from the first tests (hundreds of apps generated) I can see that the vast majority of requests give reasonable results, if they are in target with the focus of AppifyText.ai (requests like "Build the new instagram for me!" gave obviously unexpected results).
Low-code and LLMs can coexist: low-code can be just another layer (or, if you prefer, a more abstract programming language) that LLMs can use. You have less freedom, but more predictability and robustness, which is perfectly fine for internal tools.