It’s been a long journey getting here.
2 years of my life. Little to show for it yet.
Hundreds of dollars. Hundreds, maybe thousands of hours. Missed sleep. Tired days. 8 generations of architecture. 19 20 AI-generated images of Solaire.
Tetsuo MK-VIII is now deployed to my local server running on 18 allocated cores to a VM. The driver was self-interest, really: Linode’s prices went up for hosting, and my revenue streams are absent, so I needed to host it locally or decommission the project.
I was initially going to decomission it but then found some explanations for its performance in MK7 that were pretty trivial bug fixes, and before I knew it I was pivoting from batch processes to distributed network services with an inhouse API gateway, a job queuing system, and it all kind of fell into place, so, here we are a couple weeks later and it’s finally deployed. The rewrite is more modular than even the the MK7 version was, so, as things get added or removed in the future its alot less work. Glad I did it. I feel lucky that it turned out this way, as, I have alot invested in the project at this point, and didn’t want to decomission it but also am quite burned out on it and I don’t want to touch it for a while in anything resembling a serious manner.
50 bucks a month back in my pocket. And if it makes money, maybe more. I’m not expecting it to at this point. The stock market is too unpredictable and I’ve reached a limit in my ability to try to tackle the issue of knowing which stocks will go up and which will go down.
Maybe I can try adding more forecasting algorithms later with different models. MK8 allows that to be a drop-in capability the way it’s written now. Other than that, I’m out of ideas and out of inspiration on it. But, I needed to do this to completion.
Not otherwise trained in mathematics and unlikely to pass a quantitative stats class, I can now say I’ve got some self-taught experience with time series forecasts, machine learning, and a few other things I picked up along the way. It sharpened my web platform capabilities.
The new version shows potential in that all of the tasks I was performing routinely to calibrate it, which I’d always forget to do and it would nosedive, all happens automatically now. So touch time is almost non-existent. It just runs forever in a box in my office.
The only downside is, if I lose this place, the internet connection to it is the only proof of all that work and, if it were to become profitable after this, the only way it can do what it does. It has to have an internet connection.
Something will open up.
In the meantime, as with every fresh deployment after a haul like Tetsuo was, I see little glaring things to fix in the UI after deploying it that I’ll polish off at a leisurely pace in between other things.
I guess I’ve used all my excuses to not go back to Dark Horse and do a new version. There’s a package manager that needs finished there, and then a build system for the packages, pipeline analysis for security patches, A Dyad refresh, Pyrois refresh, then an installer image builder once it transitions to a package based distribution — all kinds of stuff, and very little time to do it.
Whatever. It’s done enough:
Job search continues.
It’s time to do some spring cleaning around the house.
The dog needs exercise.
I need exercise.
I need a drink.
I need to relax for a while.
Tetsuo is dead. Long live Tetsuo.