Update 60: Art-Ificial-Intelligence (Patreon)
Content
Hello everyone!I’m in what I’d call the healthiest place I’ve been in for a while while taking on work. It’s still hard in places. I still get the occasional anxiety spike around what’s “good enough”, or whether an idea that feels strong in my head will translate into a solid presentation.
That said, things do tend to resolve themselves. Often an idea just needs a bit of room to breathe. I've been applying this to my personal life too. It's okay to take my time.
By taking my time, I’m genuinely glad to say the work I’ve been doing has been rewarding, exciting and engaging.
So let's talk about what I've been up to, I do have a few snippets from the animation I'm working on, however, as I discovered recently (and perhaps this is a recent policy change) patron does not allow overtly identifiable adult content on public posts (even from creators marked as adult). In fact, you may have noticed the black bars on the last images - thanks to my personal editor (canva, my beloved).
For that reason the previews are under a paid section at the very bottom.
I’ll also be posting them publicly on SubscribeStar and in the Discord server, where those restrictions don’t apply. Don't say I don't spoil you.
Anyway! Onto the update, and of course...
The Plan
If you read my last update, I mentioned that my first project was going to be a remaster of Mechanically Insistent Pleasure, this time with Lara instead of Sana.
A few things shifted that plan, and as a result, that project has moved behind my current one.
First, the voice work didn’t arrive on the timeline I originally expected. That’s completely fine. I plan for delays, and I’d much rather collaborators take the time they need than rush to meet an arbitrary deadline.
Second, while working with the model for Lara, I ran into a few limitations I wanted to resolve before committing to a full animation. There were two options: animate with the base model and hope swapping in the adjusted version later wouldn’t cause issues (it usually does), or pause and resolve the model properly first.
I tend to optimise for least friction in my workflow, so I chose to table that project for now and focus on finishing Visiting Cirimani – EP2.
Before pushing ahead, I also wanted to close out some remaining technical issues left over from last year’s major update and remodel. There were still a few rough edges that needed attention, so naturally, my first priority became...
The Tech
As a small-scale creator, I’ve always put a lot of weight on the hardware and tools I work with. They’re the kind of investment that means when I do finally hit a flow state, the machines don’t fight me. They just get out of the way.
I remember a comment from someone, either on a post or in a DM, saying I should probably hire someone to deal with the sysadmin side of things. That stuck with me more than I expected. At the level I’m trying to work at, doing absolutely everything myself is starting to feel less like independence and more like unnecessary friction.
The problem is: I really enjoy tinkering with this stuff.
And I’m not very good at stopping.
I’ve broken important parts of my pipeline more than once because I decided to “just” update something unrelated while things were working. You know how it goes. One line changes, everything falls over, and suddenly the idea of taking a break to do some actual animating is gone. Before you realise what’s happened it’s 2am, you’re twenty tabs deep, arguing with GPT, and you’re only there because you didn’t version something properly.
I’ve landed on a compromise that mostly works. I pay for support on the things that really matter, the core systems I rely on day to day, and I’m a lot more careful about where and when I let myself experiment. Tinkering still happens, it just doesn’t get to take the whole house down with it.
This was a lesson I was still learning when I decided to do some work on my NAS. A machine which I am now fairly confident is cursed.
However, this curse has then led me to...
The Idea
If you remember, last year I had a few problems with this same machine. It went down hard while I was trying to install a newer version of the OS.
At the time, I replaced the motherboard, which did solve the issue.
Later on, while upgrading it again, I managed to brick the BIOS on that new board, so it had to go back for RMA. During that window I went back to the original motherboard, which turned out not to be faulty after all.
The replacement board has since been sitting in its box, and I did want to bring it back into service. It lets me remove a redundant network card, and the IPMI interface is significantly better.
So the plan was straightforward: swap the motherboard back in, and while I was there, experiment with a different operating system.
What could go wrong?
As it turns out, quite a lot.
I’ll spare you the blow-by-blow, but the short version is two solid days of things breaking, being fixed, and then breaking again in slightly new and inventive ways. Stressful, confusing, and thoroughly unenjoyable at points.
Still, it did get fixed.
Alongside the swap, I also added a small GPU to the NAS. That was intentional, not incidental, because it let me finally explore something I’ve been curious about for a while: running large language models locally.
Which brings me neatly to the subject, and title, of this update:
AI
I think, like most artists, when AI-generated models first started getting attention, I felt threatened. Seeing them produce impressive results, endlessly and at speed, didn’t sit well with me at all.
For a while, I put myself firmly in the “anti-AI” camp. I hoped it was a bubble. That it would pop, go away, and I could relax knowing it wasn’t going to take my job.
That obviously didn’t happen.
The reality is that this technology is here to stay. At some point, everyone has to decide how they’re going to respond to that. You either learn how to use the tools and see where they help, or you don’t, and accept the consequences of that choice.
I started using ChatGPT last year, mostly as a practical tool. It helped me through my transition to Linux, get comfortable with new operating systems, and build workflow-centric systems for my production pipeline. I’m being honest when I say I wouldn’t have made the progress I did without it.
That usefulness forced me off the fence. I couldn’t pretend it wasn’t valuable anymore, even if I was uneasy about where that reliance might lead. There's an excellent youtube video I saw around the time that addressed with this exact concern:
https://youtu.be/rqR7z2eHOBE?si=IZeDoy1IDNaRAHDD
The TLDW of the video is:
AI companies want people to be reliant on their services
Inevitably, these cheap introductory prices will change, and the cost will rise
They will consume your input and use it to train additional models
Now, while the financial costs are certainly not ideal, that's not what bothers me the most.
It's the identity and privacy cost.
I don’t want my ideas, storylines, infrastructure details, or internal systems folded into someone else’s training data. I don’t want to be the product. I want to pay for a service that works for me, not one that quietly works on me.
Given that I already run a local fleet of high-end machines, exploring locally hosted models felt like a natural next step.
One thing I want to be very clear about up front: I remain firm on my intent to not use AI image/animation generation, to me, that invalidates the entire -point- of why I do this. More on this later, but just to be clear:
You will never receive AI generated content from me.
What did start to click, as I learned more about how these models actually work, was how useful they could be in very specific, very unglamorous areas. I already relied on them for infrastructure tasks: coding workflow tools, engineering Blender add-ons, and building custom render management systems.
Rather than adding yet another subscription, I started experimenting with the idea of a local-first coding agent. Something that could take a functional idea and deal with the syntax and boilerplate for me, so I could stay focused on the work that actually matters.
While exploring models, I did briefly get distracted by a thought experiment. I fed an uncensored model context about AIMEE’s backstory and asked it questions. What came back was technically impressive, disturbingly on-brand, and completely hollow.
Seeing it produce pages of text in seconds that might have taken me a week to write was genuinely unsettling. For a moment, it felt like the end of something.
But the longer I sat with it, the clearer it became what was actually happening. The model wasn’t creating anything. It was reflecting back what I’d already put in, reorganised and amplified. These tools don’t originate ideas. They multiply existing ones.
For Visiting Cirimani – EP2, I wanted to explore a “through AIMEE’s eyes” perspective. I even briefly considered using some of that generated output directly. It fit, on the surface.
I couldn’t do it. Something about that felt wrong, even if I couldn’t articulate why at the time. Instead, I took the underlying ideas and hired a real life human writer (and friend), using the context I’d developed during those sessions.It wasn’t until I later watched a talk by Brandon Sanderson, my favourite author, that it properly clicked:
https://youtu.be/mb3uK-_QkOo?si=fC3ECR23qMBHk7VG
I don’t want AI to create for me. I want it to free me from the burdens and clear the path to creation.
Let it code. Let it handle the tedious, fragile, time-consuming parts of the process. Let it get out of the way so I can focus on the work that actually brings me joy: making something difficult, worrying about it, finishing it, and being able to say, honestly, this is mine.
Getting my local coding pipeline running is still a work in progress, but once the technical blockers were finally out of the way, I was very ready to get back to actually using the hardware. Which brings me to:
The Results
Most of the technical issues that have been nagging me are now resolved. There are still a couple of loose ends I’m in the diagnosis phase on, one of my UPS units has developed a habit of tripping the breaker in the early hours of the morning, which is as charming as it sounds, but nothing that’s blocking work.
I’ve got a plan in place and I’m waiting on a bit more hardware so I can finally turn the NAS into a proper “set it up and leave it alone” environment. The goal is to stop having to take down critical infrastructure every time I get the urge to tinker.
On the plus side, I’ve managed to tame the cooling setup so my new switch no longer tries to take off every hour by ramping the fans to absurd speeds because it’s wandered slightly past a temperature threshold. That alone has brought me a surprising amount of peace.
I’ve settled into a rhythm that’s working well. Technical work and development during the day, animation in the evenings. It lines up nicely with how my brain actually wants to work, rather than how I think it should.
I’ve also got my render management system properly up and running now. This time I tested it against real projects, and the result is a solid, reliable pipeline that doesn’t rely on juggling multiple operating systems. I can even check render progress from my phone, which still feels faintly magical.
I’ve now completed my full transition to Linux. The next piece I release will be the first produced entirely on hardware running it. More importantly, I’ve found my flow again. I’ve pulled together the missing pieces needed to turn the rough preview of Visiting Cirimani – EP2 into a coherent story and something I can genuinely be proud of.
I’ve taken a lot of your feedback on board from the comments, the DMs and on Discord. I really hope you’ll see that reflected in the final result. I’m setting a rough timeline of around two months for completion, though I’m aiming for sooner. Progress is moving quickly, and while I’ll be re-shooting older scenes, most of the heavy animation work is already done. It’s largely a case of re-lighting, re-rendering, and refining.
And The Rest
Outside of the work itself, I've had some other unforeseen circumstances that have challenged the creative process. I had a friend stay unexpectedly for a week at the start of January. She accidentally broke her foot and couldn't return home and naturally, this made it hard to immerse myself in my work. It was beneficial in some way though! It gave me the rare opportunity to bounce my ideas off someone of the opposite sex (not like that, behave.)
In addition, my home renovation has taken another step forward with the installation of new windows, which briefly required tearing the whole workstation down for a day. By the end of February, that should finally be finished, and I’ll be able to put everything back together properly and enjoy a calm, dedicated space to work in.
Overall, it’s a genuinely good moment. I’m finding a balance between the things I enjoy and the things that keep everything running.
I’ve got a very exciting list of storylines planned for this year, and I’m really looking forward to sharing them with you.
Finally, as a reminder, when VC: EP2 releases in full there will be a small tier price increase for new members. Existing members won’t be affected and will retain everything they already have.
Unfortunately, inflation comes for all of us, and with a fuller production schedule ahead, the timing feels right for a modest adjustment.
That’s all for now.
Ciao for niaou!
Preview pictures (pre-comp)