Build Mode
A climbing wall app I’m currently in the testing stage on.
According to my GitHub I have been trying to make things on the internet since at least 2011, but I know I started earlier. A designer by training having worked on the web since the early aughts, I’ve always wanted to make my own products. But I'm no programmer.
I've tried, my GitHub is a wasteland of failed attempts with different programming languages, Ruby, Javascript, PHP, whatever, it didn't matter, past HTML and CSS, it never clicked. But a few years ago no-code tools became a real thing, and I started trying them all. I have prototypes with Bubble, Webflow, Thunkable, AppGyver, and of course the newer ones Glide, Framer, Bubble, FlutterFlow and many others. These were ok. You could create apps that were better than spreadsheets, and could even do some complex stuff. But pretty much everything was limited by the constraints of the builder. You could get an MVP out there but, it would never be a real performant, designed product.
“my needs are different than the enterprise aligned roadmaps of no-code”
This is probably where my needs are different from the target market for these tools. These tools are often aimed at enterprise building better spreadsheets. No knock on this it’s needed (the world really runs on excel after all). But I want to make real standalone products. My desire is to reach initial user adoption with a product that meets the MLP state on a shoestring budget. This means having the features you need (different from the features you can create with a platform), a reasonably well designed interface (don't need perfection at this point), good UX, reasonably performant, and able to be on the platform of your choice. These no-code tools can hit a few of these but none can do it all.
But with the rise of LLMs there is a new crop of no-code tools that at least initially feel like magic. Lovable, Bolt, V0 and others let you combine the ease of no-code drag and drop builders with the power of LLM coding to make more capable apps. You can start with a text box and have a working app hooked up to a database, with a real login system, in an afternoon. At first it feels like it's making good on the no-code promise from a decade ago, if this is where you stop you're good. You can get passable design, iffy to ok UX, and more features than previously possible. But throw in a context window and a random hallucination or two and you quickly render your prototype unworkable. At this point you’re out of credits. No-code tools suffer from an enshitification problem. They all start off affordably, but a year in, start raising prices with the inevitable enterprise pivot. They tack on features for big orgs; single sign-on becomes a higher roadmap priority than improving features. Want more functionality in the map widget? Not until we build in admin controls. And don’t get me started on required version updates that break everything.
Like I said my needs are different than the enterprise aligned roadmaps of no-code. One frustrating aspect of the web world is people who get to make things are primarily developers. This is because you can get by without a designer, and many developers have design skills. If not, they can grab one of many design systems out there. In addition there are, tons of easily scalable backends, version systems, what ever you need to make a digital product quickly and cheaply, as long as you can handle the code part. But as a designer, product manager, really anyone else it's learn code (an entire other profession), or wade into the world of the limitations and ever increasing costs of no-code tools.
This is where vibe coding, tools like Cursor and Copilot come in. They are not as easy to pick up, and assume you know at least enough code to be dangerous. But these tools shine where the promise of no-code never did. I can polish design and UX to my standards, I can have any feature I need, not limited by the platform's roadmap or sketchy plugins, it can help with the things I'm bad at. Most importantly they can get me to a fully working MLP with the features I need, and take it to users for real feedback and iterate. They can take me from zero to hiring a developer to make it better. Much like my 3D printing post from a few weeks ago this just wasn't possible a couple years ago. Recently I was able to take an app off a no-code platform (who were raising their prices yet again), pull it into cursor, add a feature I had been hitting my head against a wall on for weeks, and send out a new build in less than a day. I don’t need a lot more proof than that.
I know there are downsides the code is not super clean (not that no-code was better), it is really easy to hallucinate and mess things up (commit your changes), and we will deal with the fallout of AI at some point. But I have four apps sitting in various stages in the app store, One is live, one is in testing (see the post image), and the others I may never launch. I have a handful of old products I may rebuild, and I have 2 other products in the hopper (one is huge). I imagine if any of these find success they won’t continue to be generated with an LLM, they'll have development teams, more UX people, writers, the people you need to make a real product. I know there are problems, but for now I am going to keep creating, because finally I can.
