I think the biggest problem with this approach is at some point, they necessarily reach a level of complexity that it takes the same amount of time to learn how to use their "no-code" tools than it does to just learn how to code.
I joined a startup that didn't have any deep coding skills - but had a great idea and vision for a product. They had made a pretty decent prototype with Bubble and could quickly implement suggestions from the beta testers.
When I joined I instantly suggested re-writing it in a "real" stack. But when I dived into it, there was a lot of built in features of Bubble they were using that would be a pain to write or integrate. The rewrite would take some time.
Instead we launched with the Bubble app as it was. It targeted a niche market, and that market came flooding in - it really proved there was demand for the product and the feedback helped shape the end result - months ahead of where would have been if I re-wrote it.
But as the app grew we ended up with "spaghetti no-code", slow loading times, crazy hacks, giant bundles (that we had no control over)... but again, it was good enough to launch with and validate the company, and it was fast to try out ideas.
We eventually did a full rewrite once we were happy with the overall structure of the app, but the process changed my perspective on the value of no-code tools in the right hands.
Let me guess, you are one of the brilliant ones.
I've met some dim people in science research tracks.
That's not really proving your point about them being "smarter than any developer". Lots of developers are smart enough to build their own tools.
More, they weren't even aware that these things could be improved. So no, I'm not convinced life science is intrinsically harder or the people there smarter. I attended some biochem and genetics classes at uni and once you have the foundations the concepts are actually pretty simple, there's just a lot of new terminology and random facts to memorize.
Often there are multidisciplinary research teams and depending on how little a specialization already overlaps with CS/SE-ish topics, having at least someone who realizes which mundane stuff can be automated can be invaluable.
Our microscope acquisition code is a single 25k LOC C file.
IMO Bubble is only ms-paint+ (entry point lowcode) and Photoshop (or Figma) is yet to hit the market.
I'd argue, no.
Bubble will not provide good handmade applications in the vast majority of cases.
Sounds to me like a generally applicable thing to be honest. Even as a programmer I can't just pick up my favorite language and use it for any type of software. There is mature/user-friendly infrastructure and libraries for X/Y targeting Z, for a limited set of permutations.
His QC simulation code in matlab would write the output of each iteration to a file because he didn’t optimize the object sizes. A simulation that should take an hour now takes 20 lol. It’s probably like in the Sherlock show how Holmes says he doesn’t want to waste neurons memorizing constellations.
Personally as a coder I have workflows, tool-chains developed over several years that writing a code to create a MVP is faster than learning no-code tools, also the end product is lean and stays with me.
So I haven't used these tools but tried notion for the first time yesterday to export my kindle highlights(HN comments)[1], So I can see the value in these no-code tools in automation workflows for even coders.
[1] https://twitter.com/Abishek_Muthian/status/14199150634906050...
If you can cope with instructing work-to-rule workers, you can code. No extraordinary smarts required.
As of "much smarter than above average programmer", programmers in average are at IQ 115. If we take "above average" as +15 points (IQ's standard deviation) and "much smarter" as another +30 points, we end at 160 IQ points person. Four standard deviations make that person one in roughly 16 000, or sixteen thousands. I think you should know a lot of people, much more than 150 or so of average person, to know many of them.
And to offer you my experience, my brother seems not to have any problems with coding in Python. He is well into 40-s, most of his career he was at the C*O positions and right now he decided to change venues one more time.
Nice example that if you define the terms to suit your argument then you can make any argument. Also, the word 'smart' has much wider scope than just IQ. It is also possible that the people OP knows are not sampled randomly from general population (he might be in a chess club for example).
I'm guessing many people would get this right up until the last step and see how much juice they could squeeze out of the prototype which I would imagine could be fatal in many cases.
I still wouldn't suggest this route for anyone.
I believe no-code is incredibly underrated by many early stage startups. Rewrite only once you found your PMF and there's a real performance issue (or other technical reason), just you like you did.
One of the best thing about no-code is that you avoid painful, timewasting continuous refactoring while trying to iterate on your products the first months. It's much easier to design & structure the code to be as close as possible to the business (DDD-style) when you write it from scratch after reaching PMF.
Being able to iterate fast early on is such an important part of startups. I think code is not great at this.
Can you say more about these problems? I'm curious what they're like in practice, if/how companies using Bubble can work around them, and whether Bubble might be able to fix these kinds of issues in the future.
The hard bit I think is the "spaghetti no-code" problem that I've seen in nearly all visual-type programing environments. Just because it's no code, all of the logic still needs to exist somewhere. When we code we can break it apart in many ways, and how you break things apart and abstract things depends on the system.
But no-code tools are fixed, and you only have one way to divide up the logic. In Bubble's case, it means there's a "design" section with UI and data-binding, and a "workflow" section with all of the actions that happen: UI interactions and business logic are smooshed in together. You end up with thousands of boxes indicating actions, in a giant list. You can kind of group them, and you can colourize them - but everything is held together by convention. A single logical flow can require many UI elements, and tracking that logical flow through the various unrelated boxes is... spaghetti!
In Structure and Interpretation of Computer Programs it says "Every powerful language has three mechanisms for [combining simple ideas to form more complex ideas]: * primitive expressions, which represent the simplest entities the language is concerned with, * means of combination, by which compound elements are built from simpler ones, and * means of abstraction, by which compound elements can be named and manipulated as units."
Most of the no-code tools seem to miss (or have a too simple version of) the final mechanism: "means of abstraction". So the entire system is built of combinations, that are poorly tied together.
Switching off the no-code version was a happy day. We'd pushed it further than it was designed to go. The initial advantages of quick prototyping slowed down, and it was getting harder (very hard) to add new features without breaking old ones. Customizing the design (to our designer's high standards!) was... tough.
The coded version was better in every way - but part of that was because we learned a lot from the prototype. I personally probably wouldn't start a fresh project with it - but I would recommended it to (technical) non-programmers who want to get an idea out of their head to see if it's any good.
Repl.it and peers seem to be on the right track here
Do open source projects have an obvious "run this script to configure your complete dev environment in respect to this project" ?
I would be happy to spin up a new VM for every project on that basis.
I was waiting for something like this for years, it’s great.
Try it out https://Gitpod.io#https://GitHub.com/gitpod-io/website
PS. If you aren't impressed by a JavaScript app then take Kubernetes for a spin!
I use VS Code devcontainers a lot locally, and it's the same concept. This really ought to be something we aspire to as a standard IMO.
I've submitted several Gitpod + VS Code Devcontainer setup PR's to projects to spare others the pain of having to figure it out.
One of the latest OSS projects I checked required gcloud and some online account to run make, it wad absurd...
As a fulltime working parent, 50 hours a month is more time than I am likely to spend anyway.
Awesome. Thank you!
Sounds pretty ridiculous, doesn't it? Why is it that people think that all this computery stuff like syntax and semantics, compilers, and operating systems is superfluous fluff invented by programmers to gatekeep their profession, instead of tools we created to actually do our job?
This is the original idea behind IDEs, in that they would come with everything set-up, and you could just jump into the code. But we lost this, maybe because of the web, or the great IDE extinction of the 00's, or the proliferation of non-integrated OSS, or some other thing that happened at the time, and now IDEs are about as hard to configure as a normal build environment.
./configure
1) Install an appropriate compiler toolchain.
2) Install whatever other tools are needed (like pkgconfig).
3) Install the necessary *-dev packages for the library dependencies.
4) Figure out which options to pass to configure to include the features you want and exclude the features you don't want.
5) ./configure
6) Go back to step 3 because you had a wrong *-dev package version, or you missed a library you care about.
7) ./configure
8) make
9) Diagnose the error messages and scream in frustration, because the compiler you installed in step 1 is too old/too new/too not gcc.
10) Install gcc version x.y.z
11) ./configure
12) make
13) Get the same error messages, because you didn't deinstall the first compiler and configure helpfully selected it again.
14) Give up in disgust and let someone else fix the bug you found.
0.1) Install the autotools.
0.2) Run autoconf, no wait, automake, no wait, autoreconf, no wait, autoreconf --install.
0.3) Delete all files in the directory and checkout a fresh copy.
0.4) autoreconf --install
1) ...
_carbyau_: Simple command to configure your complete dev environment in respect to this project" ?
layoutIfNeeded: configure
adwn: No, because ...
> [...] and can be done once then saved as a VM or Docker image to be cloned
is like that infamous HN comment proclaiming that nobody needs Dropbox, because you can simply "get an FTP account, mount it locally with curlftpfs, and then use SVN or CVS on the mounted filesystem". The original question asked about a simple command to configure a complete environment.
Ironically (or perhaps obviously) the OSS projects I've used that have approachable build processes and requirements don't seem to have easily-fixed problems...
Have used Vagrant for that, worked well, but not popular these days. Spins up a VM, you check in a script to install everything you need inside it.
Docker Compose isn’t bad for that either. Can specify a bunch of components such a databases, and a Dockerfile can specify what you need in your main container such as particular versions of Java or whatever.
Not to say all projects are using these things. But they’re out there, and can be used if the developers want.
That said it's still way too hard to make web apps the "proper" way. I think there's space for something easier but I bet it will look more like Mathematica than LabVIEW.
I’ve been programming with bubble from “technical aware founder PM/founder” and bubble is liberating.
In the past i’ve built a webapp that :
- Renders desktop
- Renders mobile
- store in database
- after a conditional form
- create a complex journey of to do list with defaults based on answers of previous cond. form
- connects to an external system with cron daily
- sends email daily from backend
- uses plugins from community
- much more
The deal with bubble is : - resources cutting 10X - if you manage to understand the adequate scope (serving up to multiple thousands) with no such fancy ui, with not blazing extreme loading perf (since no focus on static)
Bubble is : - Build 10x faster a custom ish layout with custom logic to get your first 10k users. (NPM available if you code)
Does it scale to infinity? No
Will you need code after that. Most probably. But you got there with 10x less ressources
You may as well be saying that getting furniture from IKEA isn't any simpler than buying lumber, saws, wax and paint from Home Depot and building your own.
Much in the same way that Home Depot is not for me.
* Where to deploy? * Where to store the code? * Which library to send emails with? * Which 3rd service to send emails with?
A long time ago I spent 2-3 weeks building something to do all the above. After completion, I started going to customers and realised those 2/3wks were basically wasted time because I could have just put a google form and would've done just as good a job. Except now you get something a bit better than that.
The advantage of no-code comes mostly for people who don’t have that same background.
They also need to invest heavily into UX and the ease of figuring out how to do something. I've used Bubble and similar tools, and they share something in common with coding - you need to watch/read a bunch of tutorials to figure out how to do what you need.
What they save is an initial headache of devops / authentication / wrangling which is bad enough as a programmer but extra hard for a non programmer to contend with.
Here is another idea for someone - a yes-code tool that lets you write everything in code, but handles devops/authentication and some common stuff for you. Basically Heroku + some NodeJS starter kit that is meticulously kept up to date.
There is library basically for anything you can think of. Out of the box you get all the batteries you need
Most code builders don't even give the devops/authentication bits.
A vast majority of our day is spent focusing on the business end.
It's similar to Shopify, I can put up a store in ~20 mins and start selling immediately. I would be extremely misguided to try to "build an platform" when that problem has already been solved.
Same with no-code. Why code it from scratch when it's been solved already? If you want to go super deep, sure. Seems like the hate comes from a cross of not-invented-here syndrome crossed with only 10 billion dollar companies in 2 years are cool.
Here's an example of a Bubble biz. https://goodgigs.app/
“Your browser was unable to load some necessary resources, contact your IT network administrator and ask them to allow access to
dhtiece9044ep.cloudfront.net
dd7tel2830j4w.cloudfront.net/
d1muf25xaso8hp.cloudfront.net”I truly appreciate the problem these platforms are trying to solve, but they're just not there yet. I wouldn't say it's easy to build a no-code platform, but it's much easier without the constraint of needing the output to have good performance - that's the hard part.
All you do with no-code programming is outsourcing writing the actual code to some other company, who already did all the work for you, and you just plug it together. To me, that doesn't reduce complexity, or mental load, it just offloads it to someone else. There's no "no-code" here, they just let you use the code they engineered.
Math used to be written in prose. Arguably symbols are better.
I think feedback systems are a good example of where diagrams are nice.
One advantage of diagrammatic/block/flow-based programming is that it lets the programmer avoid fiddly syntax. Many experienced programmers have internalized this to the point that it barely registers as a concern except when learning a new language, but for novices doing things like correctly using tags/brackets/semicolons often lead to more failure and frustration than correctly implementing "hard" concepts like recursion.
For examples, check out the Alice project's papers from Dann, Cooper, Pausch, et al.
So if you don't know if a particular task will grow to that level of complexity, it is probably better use of resources to use a nocode/lowcode solution until it is clearly insufficient, and then take the pain of migrating to code.
I've worked with '4G' since Mimer, Mapper and all kinds of other contenders and until a few years ago (roughly until 'Mendix') low code / no code was a limiting environment.
But Mendix got it right and there are plenty of valid domains of application for the current crop. Plenty of success stories of companies that used it as RAD and validation tool. Once the money rolls in you can always do a rewrite, it is much easier to do that when you are successful then to try to find success on a platform that is slower to develop for. Like every other tool: use where appropriate, there are plenty of domains where this approach would not work.
I can't think of a reason it'd be impossible to build a no-code app that scales as well as code, and I don't have evidence that Bubble has failed.
If felt like there was a missing level of abstraction - everything was too intertwined. Making changes to anything becomes harrowing once the app becomes large - it was slow, and you don't know which UI elements are being accessed by any blocks - so it's really easy to break things without realizing. And the undo system gives you no visual clue as to what is being undo-d if it was in a different view. It was the most fragile system I've worked on.
However, many of those issues could be fixed, improved, or redesigned. If you were doing "regular" use-case stuff it was... fun! Our designer could get in and implement/test things really easily.
My gut feeling is that there could be system that scales well - but the escape hatches need to be reliable when you hit the edges.
Could you do everything with pure Python or C++? Sure, but it would be much slower. Low code is useful if your need is not covered by an available application, and on the other hand you lack the resources and/or time to develop an app from scratch.
But I fully agree that "No-code" is a marketing term only. Even the easiest "No-code" app requires a good understanding of concepts like "if/then". The term "Low-code" is more honest.
If a toolchain that allows to deliver fully functional software of appropriate complexity without writing code was even remotely feasible, Apple/Microsoft/Google would have already been pushing hard each for their own take on it.
Considering the immense upside for whomever of the big tech that manages to do it first, and that they can throw much more resources at it than the best-valued SV unicorn, presumably they’ve tried (probably more times than we know); and presumably what they’ve found is that at a level of no-code tool sophistication where it’s good enough it becomes indistinguishable from programming—so they’re trying to attract more people into software engineering instead.
Snark aside though; do whatever floats your boat. Tools that allow non technical folks to start building useful products is still a win (even if I personally wouldn’t make that tradeoff).
In the past I have found that as soon as a server push action is required things break down pretty fast.
IE display alert at a certain time but not if the user has focus in a text field. Show an message indicator instead.