The most capable AI models on the planet now run on a phone that predates the iPhone App Store. That’s the pitch of Aikon, a hobby project that turns the Nokia 6300 into a chatbot terminal. It’s a stunt, sure. But it’s a clever one, and it shows how thin the client side of an AI product can get.
Some background. Nokia released the 6300 in early 2007, the same month Steve Jobs introduced the iPhone. It became one of Nokia’s best sellers on the back of a clean design, a low price and simple software. Nearly two decades later, it’s a stranded device. It runs on 2G, and its browser never got support past TLS 1.0, a standard Apple, Google, Microsoft and Mozilla agreed to deprecate in 2020. By Engadget’s reporting, an estimated 25 percent or fewer of websites still accept those connections, so the 6300 can’t really load the modern web.
Turkish entrepreneur Emir Karşıyakalı found a way around that. In late September he showed off Claude S40, an app he vibe coded to talk to Claude from his old 6300. He has since widened it to work with ChatGPT, Gemini and Grok, and renamed it Aikon, which is Nokia spelled backwards.
The architecture is the interesting part. The phone, running Nokia’s S40 or S60 operating system, never talks to an AI company directly. Instead, you run your own Go server that holds your API key from Anthropic, Google, OpenAI or SpaceXSI. The server relays the request, then chops each response into 8kB blocks the phone can handle. Because you’re using each vendor’s API, you pay per token, not through a chatbot subscription.
For a gimmick, the feature list is generous:
- A reading mode with a dedicated keypad shortcut, built for the 240×320 screen
- Offline saving of responses, if you have a memory card
- Photo uploads for chatbot analysis
- Voice messages up to 30 seconds, transcribed by an OpenAI model
- Server-side message storage for up to 30 days
- Support for English, Spanish, Portuguese, French, German, Russian, Indonesian and Turkish
The project is open source, so anyone can fork it.
Does this matter? Not commercially. Nobody is building a business on 2G Nokias, and Aikon has a hard limit: if you’re in the US, you can’t run it. T-Mobile shut down the country’s last 2G network, its old GSM service, in August. Anywhere that still has 2G coverage is fair game, and the installation instructions are on GitHub.
But the project is a useful reminder. Every major chatbot is, at heart, an API call. Compare that with the push for on-device AI from Apple, Google and Samsung, where new models often justify new hardware. Aikon argues the opposite case: with a thin relay in the middle, a very old device can reach frontier models. That’s the same logic behind SMS and WhatsApp-based AI assistants aimed at low-connectivity markets, and it’s a more practical idea than this particular phone suggests.
It’s also a neat example of what vibe coding is good for. A solo developer built a multi-provider, multilingual client for a 2007 platform with a tiny memory budget. A few years ago that would have been a niche weekend of J2ME pain. Now it’s a polished side project. Don’t buy a 6300 for this. Do take note of how low the floor for AI access has become.
