Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
A no-code tool that generates code (compoze.org)
38 points by Exifers on Dec 27, 2022 | hide | past | favorite | 54 comments


I have a simple conjecture - software is a new form of literacy, and it is often useful to cast new tech ideas in these terms - for example

"A tool for people who cannot read or write, that generates work emails and policy documents for them"

Sometimes the road ahead is not going anywhere useful.


I tend to agree, but I do enjoy the direction we are headed. At first, the solutions will be clunky and more often than not incomplete or wrong. However, with time, the great solutions will prevail, opportunity will present itself and lead to innovation and we will end up utilising tools like this to get 80% of the work done so we can focus on the real value of the product or solution.


Seriously I see no reason why a no-code config tool couldn’t autogenerate vast swaths of Django or Rails or Next.js projects for common usecases


I'm surprised we havent seen more of this.

I guess the problem is that you dont have a competitive moat. It's too easy for your users to just take the code, find their own developers and host it themselves.


But why not just do that with code? Or use a framework which requires less boilerplate?


So a nontechnical user can basically mock up what they want and a backend dev can touch it up and get it production ready without having to slog through all tedious model and handler writing and can focus on non-solved problems.

This is already a workflow that works relatively well in big data stuff - analyst makes a dashboard with some spark queries, which work but are slow and inflexible, data engineer tweaks into a production-ready pipeline and data warehouse view using more professional-level code.

The vast majority of stuff where you just log users in, sell some stuff or subscriptions, and collect payment is already done with tedious identical piles of Django and Rails code anyway.


You would because you can without a subscription. If no code tools were free we might see different uses, but I really can't see people who can code paying for tools like this.


I see no-code as simply another level of abstraction. We’ve obviously got plenty of those already.


Right. Grammarly isn’t going to produce the next Shakespeare.


That will be GPT-6's job


I find it interesting that coders seemingly are fascinated by no-code tools. I mean these things have been around forever, iterating in development.

I've recently had the idea that the real problem isn't ease of use and effectivity of these tools, but rather that people who don't code just aren't too interested in developing things. After all, if they were, they'd probably also be interested in learning to code. So I'm not too sold on the idea that these kind of tools will one day make developers go out of business.


My take on it is that no-code misunderstands the problem it’s trying to solve. The root assumption of any tool of this kind is that code is complex. Code isn’t complex, it’s simple. Because it’s simple, it’s flexible, and because it’s flexible, it’s often used to solve complex problems.

Business people see these complex problems being solved by code and think that the code is what is complex. They don’t understand that the business logic at the heart of the product is what creates the complexity, and that this complexity is still going to be there even if they use a drag-and-drop interface.


I dont think it's quite so simple. Maybe 70% of business problems fall into a camp that no code tools can deal with and a long tail of 30% that do not.

The problem is that:

A) No code tools rarely provide good guidance as to where that 70/30 split lies.

B) It often takes a good coder to figure out when you've run into this square peg round hole problem and if you use a NCT you probably dont have one to hand and non technical users are SO bad at this (they will often think that they are the problem).

C) No code tools often end up reinventing programming languages really badly to bridge that gap instead of bridging like they should in an effort to keep users on their platform and because designing programming languages badly is fun!

D) As a result, most people not only misrecognize that 30% long tail and misapply the solution.

As a result no code tools have a perpetual allure until they get used on slightly too complex a problem at which point they become uber brittle and everybody starts tearing their hair out.

Sometimes the rigidity of the no code tool is baked in, which is even worse (e.g. like this one's bad mobile UX).


This is a great response, I agree. One thing I'll add onto point A is that I think it's sometimes impossible to know which side of the 70/30 you're on until the rubber has hit the road and you're deep into implementation. Which of course, complicates the whole thing even further.

Take this recent experience of mine - I'm helping someone build a teletech system with Twilio. Nothing too crazy, just connecting customers to a service provider. Seems like the perfect candidate for NCT - heck, even Twilio offers their own in-house NCT (Twilio Studio).

I just started building it yesterday and a literal avalanche of questions hit me. What if the call drops? What if the customer isn't happy with the service and wants to speak to someone else? What about (x/y/z) concerns?

Before beginning I would have put this project square in the 70%, but now I'm convinced we're in the 30%.


That's only half the story. Bigger hurdles are the framework fatigue and the constant of having to keep things up to date even after you've got a "complete" product. This is why you see shadow IT in the form of MS Access DBs and spreadsheets that (no matter how terribly implemented) won't die.

With no-code, those at least go away.

You're no longer having to deal with unnecessary complications of having to learn or deal with WebPack, etc, and you're no longer needing to stay in an endless cycle of dependency hell updates.


> With no-code, those at least go away.

Spreadsheets are no-code that works and has been for decades. No-code won’t kill spreadsheets.


Half the time the feature requests I get for the product I manage can be summarized to “like…a spreadsheet”.

I’m convinced that if the entire business world had its way, it would run entirely within an excel spreadsheet.


I've experienced this too, and it's a reason I'm building a reactive programming language. One way to think about it is a headless spreadsheet. https://www.adama-platform.com/


I think this is there the real solution lies to this problem. We don't need another drag-and-drop, we need to change how we think about code. I'm building a "programming" language specifically for UI designers. I think high level DSLs are the future.


I wasn't clear... I meant the hurdles go away. Spreadsheets from 90s versions of office still work just fine, whereas my nodejs app from last year is another story


That’s exactly it.


>>> I find it interesting that coders seemingly are fascinated by no-code tools.

I think it would be mostly management who would be interested in no code tools. I understand, the allure of not needing a coder, anyone can create an application etc.


In my experience it is not really management but more various stakeholders and product developers who just want to implement ideas without having to ask for dev time and then wait for the project to be prioritized and hopefully eventually implemented. They do not like the dev team as bottle neck and gatekeeper.

But what that basically means that everyone becomes a developer which often ends up in s huge mess. Using no code tools is not much different from writing code. And letting a bunch of unskilled devs loose on your code base is generally stupid.


I'm management and the primary developer. (We've got a super small tech team.) I look forward to no-code tools so I can get designers and others tackling less complex problems/projects that would otherwise take up my development time. The hope is that I'd be able to focus on more challenging things, much like how open source relieves developers from building yet another [insert here].


I’ve seen plenty of unmaintainable solutions written using Excel and VBA, FoxPro, Access etc.

In my recent experience, the only “low code” solution that may be feasible to non developers will be something like ChatGPT.

I am working on a project now and it does as well as junior developers writing well written, commented logical Python code when you give it your “business requirements” in plain English.

While I wasn’t too surprised that it knew about standard Python libraries that work with xml, Yaml, json, etc., i was surprised that I could tell it:

“Write a Python script that takes an xml file with key and value nodes (I gave it a sample) and given a DynamoDB table with key and value fields, if the key from the XML file has a corresponding key in the DDB table, replace the value node in the XML file with the value field in the DDB table”.

The code worked perfectly.

There is no reason that properly trained model couldn’t create a GUI.


I just won a large contract to maintain a legacy bit of code and bring it up to scratch. Our client's effort to migrate away to no-code/low-code never got off the ground. The party that was supposed to do it lacked the tools and know-how to make the switch.

I donnow, but if you're a manager that's dreaming getting rid of expensive developers, my impression so far is that all you're doing is replacing developers with less competent and more expensive "consultants" that need to build solutions with one hand tied behind their backs.

No-code/low-code is useful for shopping lists but complexity skyrockets as soon as you try to step out of its comfort zone.

I'll probably end up eating my words, but so far I'm not worried about no-code/low-code in 2023.


Worried? Is that the right mindset when looking at technological innovation?


That’s an uncharitable take on OP’s comment.

I read it as meaning No Code is not going to replace developers any time soon.

I say this as someone who is building a LCNC product.


It is if the technology threatens to automate your job.


Oh how I would love to see my job automated away. :)


I have trouble taking no : low code tools seriously when 50% of web traffic these days is on a mobile device and yet, the promotional site for the tool isn’t responsive.


I suspect this is used mostly for enterprise-type software and the percentage of web traffic might be a lot less.


I’ve worked at enterprise SaaS companies most of my career.

You haven’t been able to get away with an interface that doesn’t work well on mobile since at least 2008-2009.


No code doesn't take away the need to understand the problem, capture requirements and specify the solution - and these are often the hardest parts most poorly done.


There's a typo on the front page: "You next SAAS" --> "Your next SaaS".


Another typo, subtitle on pricing page (https://www.compoze.org/pricing)

> 100% free to test - on credit card required - pay only for large apps

Should be "NO credit card required"


Hi !

Here it is : drag&drop Figma-like editor to create SAAS web apps in no time.

Hit the "generate" button and it generates all the source code (React/Typescript/Nodejs/Postgres front, back, APIs, migrations).

What do you think of it ?


> What do you think of it ?

I'll try to be constructive, but this seems to be in a very bare bones state.

1. It needs a widget hierarchy tree. At least because sometimes some widgets completely disappear from view when you move others and then you don't know any more where they've gone to.

2. It needs much better drag&drop. I moved something and after just 2 attempts at dropping it somewhere, everything was messed up beyond salvation. It reminded me of the old Word joke; in the distance, sirens.

3. It needs a much better introduction. Right now you just drop visitors into a basic example with no explanation and there doesn't seem to be any documentation at all. I wouldn't mention this but on the other hand you do have price plans already.

4. Maybe minor, but since you don't mention any browser requirements I'm going to assume it works on any browser. On Firefox there's a fairly problematic thing on the buttons (and other clickable widgets): It looks like you need to click on the button but not on the stuff inside it. Take the Sidebar buttons, the "Users" button is a..

  <li>
    <div>
      <input type="text" value="Users">
So... I need to click outside of the div but inside the li for the click to actually work. See https://imgur.com/a/9dFalR3 I need to click inside the button area but outside the highlighted part. It seems like the events are not being correctly captured.

5. As others have mentioned, generating code doesn't actually seem to work at all.


It sounds interesting but your landing page is broken on mobile as it has a min width wider than my screen and scrolling disabled.


+1. The page is not viewable on mobile (Safari on iOS 16)


Yep. Often the case with links on hacker news people don't realise that at least 50% of the time we're looking at it through mobile


I’m sure it’s a fantastic service, but I haven’t seen a web page so broken on mobile in a long time.

Better to just serve a full desktop page or even turn mobile off entirely than deliver something so clearly untested.

Doesn’t instil confidence in the quality of the rest of the product, which is a shame as the landing page is probably the simplest bit technically.


perhaps you could use it yourselves to have a reactive web ui of the landing page?


This is a good idea. But I wouldn’t trust code generated by this company as their website is not even working on mobile browser


I am an experienced developer (30+ years) and I use “no-code” solutions all the time. However the “no-code” solutions I use are code generators written by myself. I can generated roughly 80% of the code I need to implement typical business applications. It is a massive productivity boost.


I know we like to "ship it" earlier than later - but imho that's not usable at all


Having a hard time to generate code. When I click on Launch -> Open on stackblitz, it doesn't open any window or performs any actions.


I am having this issue too. It seems it's making a POST request to localhost:8000, looks like their prod is broken.


The website is not usable in mobile.


Generate my x86 assembly code


The landing page doesn't have a concrete enough example so I couldn't figure out what it is for. It doesn't seem to be a structured way of using GPT-3.


> It doesn't seem to be a structured way of using GPT-3.

I hope this is sarcasm!

Prompt: "I want a SaaS that makes me $10k of passive income per month. The complete source code is..."

Side note: this reminds me of Google AppSheet https://about.appsheet.com/home/ (although not sure if Google's product actually gives you the source code)


I made a website that is a structured way of using GPT-3 to generate and host code in web pages. https://aidev.codes. If you have questions, you can find me in Discord in OpenAI / api-projects or the ChatGPT server. username: runvnc. There are some examples here: https://aidev.codes/u/runvnc


Well, evidently it did not generate responsive CSS for the landing page.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: