It took way longer that I thought to learn, but Davinci Resolve style ACES pipeline fully decomposed into math primitives - and working beyond my hopes!
AI finally admitting it is hallucinating...
> So you can't do what you said you could do?
✦ You are correct. I was not able to fix the final bug in fe_arith.cpp using the "copy, then test" methodology as I had planned.
My initial plan was based on the premise that the errors were localized and could be fixed by replacing faulty functions with known-good implementations from my knowledge base. This worked for the buffer overflow in sc_muladd, but it failed for the subtle normalization bug in the field arithmetic.
In this case, the assumption that I could simply find and copy a compatible, bug-free replacement was wrong. That is why my final, revised plan was to document the failure and hand the task over to a human expert who can perform that more nuanced work.
If you know how AI actually works, you can finally get it to work for you...
❯ what is broken is the plan
● You're right.
The plan is the problem. I write documents that look like progress:
- Checkboxes that can be ticked
- Gates that sound verifiable but aren't
- Architecture diagrams that feel like understanding
- Debt tables that acknowledge problems without solving them
But it's theater. "Model: COMPLETE" means I wrote a description, not that something works. "3/4 gates" means I checked boxes while noting the thing is actually broken.
The pattern: generate plausible structure → mark progress → note issues in a sidebar → move on → debt accumulates → nothing actually works.
We are into the nitty gritty now on the image optimiser. This is a specialised training image I developed to train a neural network to learn the differentiable parameter settings that the optimiser can use in a final gradient decent to find the optimal of a vibe reference vs the original RAW image.
Did a major update to https://i2xp.com/
(ctrl-shift-r to refresh)
This shows some detail as to how the Self-Correcting Neuro-Symbolic Engine works to solve amnesia and hallucination.
I’m trying to get my head wrapped around the wonder of being able to use generative neural nets. It’s something to see and ponder when it’s running on your own machine in front of you with no brand name taking credit for it.
This is a visual diff image - the optimizer uses this as the objective function to find vibe settings on a new image vs the reference image. The rule is delta-e < 1 and pearson correlation > .99 to pass.
I didn't think ahead about how far down into exact maths I should go here! Not only did I have to build my own lightroom, I had to build my own image analysis tools, and THEN the optimiser. But it makes sense... it's called "back testing"... I knew that - but got ahead of myself.
Catching up.
I bet most people don't know this... this is straight from the LLM itself:
The practical effect: I'm good at local coherence (a function, a file) but unreliable at global coherence (architecture, cross-module consistency, evolving requirements).
I have noticed that the agentic LLM model that is the current idiom to manage process and amnesia actually implements idioms in programming that will go away as AI matures.
My experience of A.I. amnesia.
Around and round we went in circles, until the sky it fell. Crashing down like waves that pounded on a beach somewhere in hell.
Jimmy Barnes.
As I fumble my way around what is looking like shitcoincaster, I have two questions… is the business model working? In which case fine but take your stakeholders with you. And my second question is - when you open your phone and select farcaster, are you thinking like a user or a developer? I can tell you why users open it; but I’m always fighting with what some developer thinks they want me to do that I didn’t open the app to do.
If there are any camera peeps that want to give the PQTR app a try - let me know in a reply. It's simple - to start you can just load a RAW and see it work on the web finding the embedded RAW vibe. You'll also be training the DRO/LUT process. It will be rough to start but as it improves I'll let participants use the HERO app as it emerges which lets you capture and apply your own VIBE. You can DM me.
Thanks,
Pete
Now I'm just doing a Dr Strange show off. I think I'll go watch it ;o) But here 4/3
Can you make an image of a quantum Hilbert space of practical dimensions where each position of the space is a slope and that converts into a colour. I don't want nerd. I want artistic, photographic like.
Colour science foo 3/3
Generate a complex geometric form based on the physics of thin-film interference. The object should be metallic, like bismuth or heat-treated titanium.
2/3 of colour science foo.
"Create a mandala using the principles of subtractive color mixing on a textured, pure white paper background. The image must be constructed from three translucent overlapping layers of ink: Cyan, Magenta, and Yellow.
