While working on my Podline.fm, I’ve thought a lot about technology choices and the challenges of creating a new product. NoCode was always a top contender for the initial prototype.
When I started building Podline, I explored NoCode tools to see if they could help me create what I wanted. My project needed a highly customizable audio input widget to embed on other websites and unique ways to transcribe and summarize audio data into valuable insights for customers. I had to figure out if NoCode tools could make this possible.
I found that popular NoCode systems could do most of the job, but not to my exact specifications. For example, I could record audio but not embed it how I wanted or achieve the high quality needed. My target audience is audio professionals who require top-notch audio recordings with large files and complex encoding and processing. That’s something NoCode platforms don’t offer or make very hard to establish.
Experience this article as a podcast or as a newsletter:
On the server side, transcription and summarization solutions exist, of course, like the OpenAI APIs. But these are paid services. My “business plan” requires all AI systems to be able to run locally. Ultimately, Podline will be a self-contained system with all necessary components run internally without extra cost, except for external public storage due to the sizable storage needs that audio data requires.
So, what does this mean for NoCode?
Well, while NoCode tools make this extremely easy for software entrepreneurs, they don’t always meet the specific requirements for certain projects.
The example for Podline is dealing with transcription and summarization. I am literally running two language models on the web app server podline runs on.
These are download-once-run-forever backend systems, and they handle all of this for free. No no-code system would let me do this.
Many no-code solutions are for early adopters and innovators. They can cross the chasm, but I don’t see my initial experiments being valuable when built on someone else’s platform with its limitations. I’d rather be flexible and build my own thing so I can experiment with the features I want, not just what’s available on a platform. My customers are audio professionals who aren’t usually early adopters. Their expectations are pretty high and specific.
My goal is to create something that works, looks good, and can cross the chasm once released to its target audience. This product has an initial complexity; it’s not just connecting some APIs. It’s about treating audio data with respect and fitting into an existing workflow.
I chose to code, and I’m glad I did. Because it feels like I’m not coding alone anymore. The AI-assisted coding revolution is mind-blowing – my software-building speed has increased tenfold compared to a few years ago. The whole idea behind NoCode is appealing, but for now, coding gives me the flexibility and control I need.
Because for a new small software bet, speeding up development and prototyping is the goal. AI-assisted coding is reaching that potential, often matching what NoCode promises. As an example, I use the Laravel framework with its easy-to-integrate services. Laravel has many plugins and extensions, similar to NoCode tools. I installed Laravel, Jetstream, and Spark with just a few clicks (or terminal commands, in this case). After that, I had a fully functional login and billing system. Just like I would have had with a NoCode tool, but mine is completely customizable and under my control.
The fact that I could use an AI chat assistant in my editor was amazing. This assistant saved me so much time by suggesting solutions even before I knew there was a problem. AI-assisted coding is like NoCode because the machine does the coding part. The developer’s job is to judge the quality of the result and how it fits into the existing code base.
This technology would have been wonderful to have 10-15 years ago when I started coding. It could have saved me years of trial and error in “really getting” programming languages. With AI-assisted coding, learning from an experienced AI makes it so much easier and way less scary to experiment with code to see and understand how it works.
Developers today don’t need to go through the same challenges as they did 10-20 years ago. They now have these tools at their disposal, making them more like low-coders. This isn’t meant as an insult; it shows that understanding every intricacy of code is no longer necessary for success. It helps, and it will definitely make debugging easier, but ChatGPT and the Jetbrains’ new AI assistant do a pretty good job at that, too.
Ultimately, when you want to create a fast prototype just to see if it works, no-code is a good choice, even if it’s limited. But using AI-assisted coding is even better, in my opinion.
So, as I build Podline in public, I trust that with the help of a strong foundation, a smartly built ecosystem of tools, and AI assistance, I’ll be doing great. I’ll keep coding and building this as a code-based project.
In fact, I’d love to know what you think about this convergence between AI-assisted coding and building with no-code. A few years ago, it felt like things were drifting apart, coding here, no-code there, and now they’re coming back together. If you want to share your opinion or have any other question you’d like me to answer on the show next week, go to podline.fm/arvid and record a question!
