Latent space (Scrollable Section)

“Ancestors say, there’s 4 things in life that are the hardest to crack: Live or die, is or not, win or lose, honor or shame. I say, it’s just one word: Me” — The Grandmaster (2013)

Everything is ME

Bird is ME

Wind is ME

Celebration is ME

Desperation is ME

Life is ME

Death is ME

Everything in life is in our brain, is in my brain. I think, therefore I exist. I am, therefore I feel. I hope, therefore I believe.

When I was in high school, I developed this understanding of body mind separation. My body isn’t me, it is merely a shell that host my mind.  

This greatly separated my brain induced emotions and body induced emotions. Normally when you feel fatigue in the body, it is often reflected in the brain, therefore it also fatigues the mind.

With this "Body Mind Separation" philosophy, I started to learn and realized the "sweetness" of controlling my mind

(For example, when my body feel fatigue, since body and mind are separated, my brain never feels the same fatigue that normally would've channeled and manifested from body to my brain)

Now looking back, "The value function for human is emotion" by Ilya Sutskever cannot be more correct.

After joining COLLINS as Creative Technologist (which is the title that I hate, I prefer AI specialist/Design Generalist), I've been exploring the realms of building full-stack solutions (apps) for creative industry and as I'm writing this I've built 48 apps (Including the flagship product "Aspect OS"), 8 fine-tuned LLMs, and 27 LoRAs for COLLINS so far (5 months).

On top of all that, I started my own adventure on learning to become an actual AI researcher by designing my own Architecture and Neural Network layers for a Typography generation model.

This is an actual model pre-trained from scratch all by myself using all of the open-source fonts available on the internet with a total of 8000+ fonts and half a million glyphs. It is a 136 million parameter few-shot model designed, trained, debugged, post-trained by me alone.

And inspired by my beloved Sergio (my colleague at COLLINS), we named it "Photon".

And the first model finished training is called "Photon 0.8 GT". 0.8 not 1 because of pending post-training and lack of proper UI, G stands for Graph Neural Network architecture and T stands for Transformer architecture which comes after GNN.

As I'm writing this 2025/12/22, I'm training Photon 2, this is the third model I'm fully training with completely different architecture and Neural Network layers setup.

I purposefully trained the models using my own money to rent the H100 Cloud GPUs and on my own laptop during weekends (so outside of COLLINS work hours, although I work all weekends and average 16 hours a day for COLLINS anyway) (Seriously never took a day off since I joined, not even one weekend, well actually there was one night for celebrating my friend's birthday but that's it, and there's a few nights for events but those events are for work as well).

The reason I wanted to separate this model from COLLINS or just design agency or anybody in general is that I understand how much of an effect this model will have to the type foundries, and I wish the final decision of whether I open source this model will be solely my decision and my decision only.

(No matter what you say about me training this model, I trained this model fair and square using legal and licensed data, If you don't like how this piece of technology I build is gonna change your life, do something about it, train a model yourself and see how hard it is to pull this off, and how much craftsmanship and hours, and blood and sweats are need to design, build, and create a model especially as a designer, not someone who went to Harvard and studied Deep Learning or Machine Learning. AI didn't learn how to generate typography, I designed this sophisticated machine, I designed it, I made it, I built it. Just like how type designers with digital apps like "glyph" took over the jobs of hand drawing type setters, I'm the next version, and I won it fair and square.)

(Sorry about the slight rant above, I really needed to get it off my chest, sometimes people hate me for what I'm building, but they don't understand how hard it is to pull off what I built, and the fact I didn't cheat at all. All of the hours I put into this, was never (rarely) appreciated by anyone in the creative industry, and it does chill my heart)

Now back to the topic of Latent Space

In order to build AI models and build apps at a pace faster than a software engineering team. One must be tunnel visioned, focused, wired in. And my version for that is my "Latent Space".

Latent Space means a compressed space for complex data, and to be able to build an actual app even with the help of AI, one must juggle between, AI code slop prevention, multiple API providers, Frontend, Backend, SQL database, Authentication system, Storage Buckets, Security for RLS and API rate limiting, tenant isolation, etc…., all sorts of complex app features and most importantly, the speed of making all of these happen.

As you can probably imagine, that's a lot of shit going on per second, so the amount of focus a person need, and the amount of mental capacity to compartmentalize and quantize information to be able to process them simultaneously in the latent space so you can work on multiple things at the same time, is enormous.

This is also why I believe it is very hard for someone to replace what I do, not just the technical difficulties and the speed of building, but more on the mental side, even you can pull off what I mentioned above in a short period of time, do you have the mentality to do it? I doubt a lot of people have it.

When I'm in my latent space to process all complex data and tasks, the exact position is somewhere above my head, like a few inches above my head there's a imaginative eye that's overseeing and placing tasks and data in imaginative space/area, a 3D area.

And when I'm in my latent space, my entire brain is not in my body anymore, it is in that imaginative (latent) space above the head doing calculation and planning, the physical result is my facial expression is totally dull looking like I'm angry or something. I was told multiple times when I relax I have a shit face, and relaxed is exactly what my body is doing, moving all of the brain cells up to the imaginative 3D space where I can place all of the tasks and overview things and orchestrate things.

This is how I work, and I have a pretty good track record of claiming efficiency and effectiveness of my Latent Space Mental Model.

I gotta give a huge shout out to the GOATs who introduced me and nourished me: Nick Ace (Ex-CCO) and Leland Maschmeyer (CEO / Co-founder) at COLLINS. I not a person that expresses my feelings a lot, but deep down in my heart I'm truly grateful for the environment they gave me, the freedom they gave me to do all sorts of crazy stuff and stay in my Latent Space for long hours even days without interruption. I know I won't get that in other workspaces.

Special love to Nick for his trust in me.