AI Tools for Mobile Development: The Personal Journey of a Flutter Developer

AI helpers, chats, and AI-generated content fill the Internet. Millions of people use AI to solve tasks, help with decision making, and much more. So what about Mobile Development and what tools are the most useful for Flutter in particular? Let’s find out. And yes, the preview picture is AI-generated, but my opinion is not.

In my current practice, I was able to work with a lot of different AI tools, including the most popular Chat GPT, Copilot/Copilot Chat, Claude, Gemini, and the most exotic of them all, Kapa AI and Ollama.

In this small article, I will give my personal experience with using all of them and some small tricks.

I think there is no need to talk much about OpenAI’s Chat GPT. Nowadays, everyone uses Chat to resolve difficult errors that suddenly appear while developing a new feature, to generate some custom animation, or just to talk with. Chat GPT is most commonly used and was one of my first AI models that I used to work with, and I was pleasantly surprised. But it doesn’t become my favorite for one reason – it was not built into an IDE, and code needs to be passed as a separate file or just beside the prompt.

So next, I tried Copilot and Copilot Chat, and this was truly revolutionary for me. Because this AI assistant can be added to Android Studio as a plugin, it can give you smart suggestions that actually use your codestyle. Also, later they introduce Copilot Chat that gives developers access to Chat GPT or Claude, but, plot twist, it allows you to send actual files with a prompt, not just copy-paste snippets to an OpenAI web page.

It helps a lot if you’re a solo developer who needs code review, can’t find a problem in your code, but a bug is still there, or just needs to add something to already existing code. It’s great if you use architecture and need to stick to it – Copilot Chat can use your code as an example and generate code with the same structure as you already have.

Also, Copilot relieves a lot of pain if you have Flutter code and need to convert it to Native Android/IOS and vice versa. But sometimes it makes some stupid mistakes and deletes needed code in a refactoring process.

So always check twice what AI tools have created for you!

Even more exciting are the newest AI assistants with Agent Mode, like Claude Code or Copilot Agent Mode. I was able to have a small glimpse of its power, and it is amazing – it can create files for you itself and make a lot more context-right decisions. But for now, it is very costly, and I believe Agent Mode AI will be the next evolutionary step for AI assistants.  Can’t wait for it to be cheaper.

Regular Claude and Gemini are GPT-like chatbots. I use Claude in Copilot Chat, and from my perspective, it has more context than Chat GPT and works a lot faster. Gemini I used on my phone as a personal assistant in messages and ask a lot of silly questions – looks like it’s good and works fast, but I don’t have a chance to try it with code.

The most exotic things that I tried are locally hosted AI with Ollama – a toolkit that allows you to use small, purpose-curated AI models but fully local on your computer or laptop. My favorite for coding was CodeGemma by Google and Granite-Code by IBM. I was impressed with how fine its suggestions are that are made at my gaming PC. This model of work with AI will be very good for businesses that are extra careful with sharing their codebase and data to a public AI because this can work fully locally and doesn’t share anything with outside services. But also, it is slow and needs a lot of electricity.

Also, last but not least, I was able to work with library documentation using the Kapa AI product. Recently, I needed to work with a complicated and closed-source code library that doesn’t have great documentation. So when I need to do anything with this library, all regular AI bots are nearly useless – the code that they generate is outdated, doesn’t include what I need, or does everything wrong. But, luckily for me, they have an AI assistant built by Kapa AI that uses their closed knowledge base, searches Stack Overflow and other sources for answers. And this AI assistant greatly helps me with understanding the library, generating views, and even migrating to the newest version. AI assistants like Kapa AI are great to transform your local Wiki or other type of documentation into a helpful AI assistant, and I think this is the future for documentation on large, long-lasting projects.

I have to say that AI transformed and improved my personal Developer routines.

AI helps me a lot of times to get unstuck with some convoluted problem, or to give a boost of productivity on a not-so-productive day, or just casually to roast my code. Still, AI can’t work without us, humans, who will control what is done and check if AI doesn’t add a bitcoin miner to the homepage or other suspicious AI activity. But I am excited to see what can be done in the next several years.

If you’d like to discover more about how we can help your company incorporate AI or if you’re in need of other technology services, contact us.