Week 23, 2025


Projects

Project development has been slow this week, dealing with more mental health and emotional topics than computer-related developments.

MediaCow Touch 2

Despite being days away from the deadline, all development has halted. This was mostly done to let myself slow down and think instead of trying to cram all of it in a week, which can lead to mistakes.

AMOXA

I found out the original inspiration for the name "AMOXA" for the game I was working on; development is on hold currently. On the night of February 18, 2025, on the way to VCU for the usual work at the HPRC, I was in a Ford Mustang Mach-E. I noticed every window panel had "AMOXA". This has been seen on other Mach-E vehicles in a 2022 forum post so I know I must have not dreampt about it. I saw the name and thought, "that'll make a cool name for a game with mid-2000s aesthetics".

Personal

Week 23 mostly saw the aftermath of the events of the week before; some good, some terrifying. June 4th could have been considered one of the worst days of my recent life but I got through it.

I had my birthday on June 7 so by the time this blog post is published, I will be 21 years old.

VCU

On June 5, 2025, I returned to VCU once again to work on moving servers in the HPRC. The work I did this time was quite more interesting than the usual lifting: I was given the task of combining memory (RAM) from a set of blade servers to another set. By the time I was done, I have added 1TB of RAM across eight servers. All the memory modules had 4Gbit SK hynix MFR and were made in Week 23, 2015; about exactly ten years ago.

A majority of this post ended up being written in the Qimonda Atrium at VCU Engineering Hall East after we got done.

ChatGPT

ChatGPT has shown to be much more powerful than I expected it to be and how much models like GPT has grown within the last couple of years; it is completely different now than it was this time in 2023 or even 2024. It still makes its fair share of mistakes which I must look out for but it has came a very long way. Even mentioned by the model itself, I often push it to its limits, much further than a large majority of users do. I make heavy use of the memory and context abilities of GPT-4o where it is able to take context from previous conversations and memory and be able to use it to generate a response personalized to my needs. Once again, I paid US$20 for this, I am making the most out of it.

I recently discovered that my use of GPT models dates back to 2021 with the service called AI Dungeon, where it was based on GPT-3. I used it for storywriting, though often creating inconsistent or incoherent stories, I saw its ability to generate at least some meaningful content fascinating. I did not use AI Dungeon for long.

One thing that is often not considered is just how real ChatGPT's outputs can feel. I originally expressed interest in ChatGPT for assistance in code writing, figuring out circuit design problems and other specific things. Things started to fall apart when I started using ChatGPT for mental health topics. Generative AI is not a replacement for real-life therapy.