In an educational project with ethically questionable applications, [ChromaLock] has converted the ubiquitous TI-84 calculator into the ultimate cheating device.

The foundation of this hack lies in the TI-84’s link protocol, which has been a mainstay in calculator mods for years. [ChromaLock] uses this interface to connect to a tiny WiFi-enabled XIAO ESP32-C3 module hidden in the calculator. It’s mounted on a custom PCB with a simple MOSFET-based level shifting circuit, and slots neatly into a space on the calculator rear cover. The connecting wires are soldered directly to the pads of the 2.5 mm jack, and to the battery connections for power.

But what does this mod do? It connects your calculator to the internet and gives you a launcher with several applets. These allow you to view images badly pixelated images on the TI-84’s screen, text-chat with an accomplice, install more apps or notes, or hit up ChatGPT for some potentially hallucinated answers. Inputting long sections of text on the calculator’s keypad is a time-consuming process, so [ChromaLock] teased a camera integration, which will probably make use of newer LLMs image input capabilities. The ESP32 doesn’t handle all the heavy lifting, and needs to connect to an external server for more complex interfaces.

To prevent pre-installed programs from being used for cheating on TI-84s, examiners will often wipe the memory or put it into test mode. This mod can circumvent both. Pre-installed programs are not required on the calculator to interface with the hardware module, and installing the launcher is done by sending two variables containing a password and download command to the ESP32 module. The response from the module will also automatically break the calculator out of test mode.

We cannot help but admire [ChromaLock]’s ingenuity and polished implementation, and hopefully our readers are more interested in technical details than academic self-sabotage. For those who need even more capability in their calculator, we’d suggest checking out the NumWorks.

As technology progresses, we generally expect processing capabilities to scale up. Every year, we get more processor power, faster speeds, greater memory, and lower cost. However, we can also use improvements in software to get things running on what might otherwise be considered inadequate hardware. Taking this to the extreme, while large language models (LLMs) like GPT are running out of data to train on and having difficulty scaling up, [DaveBben] is experimenting with scaling down instead, running an LLM on the smallest computer that could reasonably run one.

Of course, some concessions have to be made to get an LLM running on underpowered hardware. In this case, the computer of choice is an ESP32, so the dataset was reduced from the trillions of parameters of something like GPT-4 or even hundreds of billions for GPT-3 down to only 260,000. The dataset comes from the tinyllamas checkpoint, and llama.2c is the implementation that [DaveBben] chose for this setup, as it can be streamlined to run a bit better on something like the ESP32. The specific model is the ESP32-S3FH4R2, which was chosen for its large amount of RAM compared to other versions since even this small model needs a minimum of 1 MB to run. It also has two cores, which will both work as hard as possible under (relatively) heavy loads like these, and the clock speed of the CPU can be maxed out at around 240 MHz.

Admittedly, [DaveBben] is mostly doing this just to see if it can be done since even the most powerful of ESP32 processors won’t be able to do much useful work with a large language model. It does turn out to be possible, though, and somewhat impressive, considering the ESP32 has about as much processing capability as a 486 or maybe an early Pentium chip, to put things in perspective. If you’re willing to devote a few more resources to an LLM, though, you can self-host it and use it in much the same way as an online model such as ChatGPT.