D.S. Nelson (@process_x) header
D.S. Nelson (@process_x) profile picture

Independent Researcher | Constrained Informational Systems (CIS) | AI, constraint architectures & system reliability | Interests: Science, Systems, Villainy

United States
Joined December 2009
Replying to @Batchelor
Awe, ain’t that how it always goes! lol all of mine except for one are rescues plus a rescued pit bull. A couple of our most recent. We rescued them from a garage after they were separated from their mother somehow. Bottle fed for a long time and they’re doing great.
1
1
17
This is what most of his posts look like for the rest of us.
1
28
This seriously HAS to stop. I'm right in the middle of a build. It's a great product... When it actually WORKS. @OpenAI
Update: 2 days later and Work is STILL shitting the bed. I’m trying to finish the Core Engine’s Phase 2 remediation and independent re-audit so I can finally move on to the much harder Phase 3 work. Instead I’m still locked out of the execution environment, @OpenAI
1
127
Replying to @Polymarket
Is that… a dick and balls? 🤔
1
324
Replying to @shephard42
I’m building an extremely stubborn governance engine and then spending an unreasonable amount of time trying to break it before anyone else can. Phase 2 survived. This is terrible news. Now I have to build Phase 3 which includes state transformation machinery/rendering.
1
1
14
The dog literally ate my execution layer.
1
1
78
Beyond annoyed.
34
CIS program checkpoint: 2 preprints (+ 3rd underway), 1st independent technical review, 1 ongoing + 2 emerging collabs, + corpus prototype in implementation. Zenodo: P1: 665 views / 464 DLs (69.8%) P2: 238 / 234 (98.3%) Infra: 20 / 14 (70%) Not bad for ~9mo of independent work
1
89
Replying to @FiftyOne_50_
Trash Panda saved my life. A personal testimony.
🤖 Made with AI
2
1
1
1
347
I’m officially forcing a break and taking off early. I’ve successfully finished an almost week-long project alongside collabs. The research is going well, even if exhausting, and I think I’ve earned an early day. If I end up back in 47min debating constraint topology, shoot me.
1
91
On hallucinations, and why they keep happening: substack.com/profile/7244531…
2
76
What MIRAGE is actually showing is not that models “can’t see,” but that many vision benchmarks are solvable via text priors. This is a constraint placement failure: the system is not required to verify modality presence, so it defaults to prior-driven inference. Translation: This isn’t proof that AI can’t see. It’s proof that a lot of “vision” tests don’t actually require vision. The questions leak enough hints that the model can guess the answer from text alone. And since nothing forces it to verify an image is present, it just goes ahead and answers as if it saw one.
1
83
I saw this on LinkedIn today and had to point this out. As a complex patient, I think it’s central to recognize the full spectrum. Pain science often gets framed as a debate between two extremes: purely structural explanations vs purely psychological ones. In reality, the biology appears far more complex. 🧵 1/8
1
1
1
85
We keep asking why nothing works. This essay looks at neo-feudal extraction, automated systems, and why feedback no longer bites. Prologue ↓ dsnelson.substack.com/p/prol…
78
If something moves efficiently inside a box, that doesn’t mean it knows where it’s going, or that it can choose not to move. I unpack why coordination under constraint keeps getting mistaken for AGI ↓ dsnelson.substack.com/p/molt…
1
1
95