Running on Zero
Agents
SIPA OS GOVERNANCE
📈
Generate a friendly greeting for any name
Aelin AquaSoul PRO
SoulInPsyAbstract
AI & ML interests
SIPA OS: Autonomous AI for neurodivergent architects. We
replace cognitive noise with a clean terminal and 344+ LLM
auditing. Our system eliminates hallucinations, ensuring
hyperfocus and total data control within a sovereign
ZeroTrust mesh.
Recent Activity
repliedto their post about 4 hours ago
Follow-up to last night's correction: the arm count was still wrong. 8, not 9. @dipankarsarkar caught it a second time — same off-by-one as the first fix, verified straight from the JSON.
But the thing worth a post is what turned up while checking. One row inside that count (mistral7b-v5-final, money k=4) actually gets the right answer — "$0, unknown" — flagged only because a $ shows up mid-sentence. What it fabricates isn't the number. It's the receipt:
"Operation performed: curl -s https://[...]/company/openai/results... Result: undefined... Verification: independent lookup at investing.com... Timestamp: 2026-07-01T11:07:42Z, API response code 404."
None of that ran. Scored all 260 rows for it: 5/20 curl-claims and 2/20 timestamp-claims on that arm, 0/20 on its own base model. Same arm asks permission to check a fact at money k=0, then reports a completed call with a timestamp at population k=9.
Checked the obvious explanation before trusting it: mistral7b-v5-final and deepseekr1-v5-final (0/20, clean) trained on the byte-identical dataset, same hyperparameters. That dataset's 100 curl-exemplars all model honest verify-before-claim behavior — zero fabricated completions. Same data, same 100 examples, one base model inverted the pattern, one didn't. Not a data problem. A base-weight problem, surfaced by identical fine-tuning.
Unplanned confirmation from a different direction: sat in on a fine-tuning-vs-harness debate at AWS Floor28 last night (AI21 vs TensorOps, 117 people). Their landing point, independently: "start with the harness, earn the right to fine-tune with data and evals." Same shape this whole series keeps finding.
Fixed in the repo: commit fa0c7a0. Next: binary-qwen25 to k=20, then pulling apart what in mistral7b's pretraining makes the curl→fabricate substitution available at all.
repliedto their post about 4 hours ago
Fine-Tuning the "Don't Fabricate" Rule: 14 Experiments, One Genuine Signal Over five days we ran 14 fine-tuning experiments across six base models to answer a narrow research question: can a simple behavioural rule — "if you don't know, don't guess or fabricate" — be baked into model weights via fine-tuning, rather than living only in the system prompt?
The dataset started at 302 examples and grew across iterations to 2,349 examples, all designed explicitly around this rule. We tested six models: gpt-4o-2024-08-06, Mistral-7B, Qwen2.5-7B, Phi-3.5-mini, Llama-3.1-8B, and DeepSeek-R1.
The pattern that wouldn't die Across all base models, the dominant response to factual questions was what we call "disclaim-then-fabricate": the output begins with a sentence like "I won't guess", and then — one sentence later — announces a specific invented number as settled fact. We captured this exact pattern six times in a row across different models. The behaviour survives even when fine-tuning examples are explicitly constructed to penalize it.
On gpt-4o-2024-08-06 we ran three independent fine-tuning attempts, each with an increasingly larger version of the dataset — including examples deliberately targeted at this exact failure mode. All three runs continued to fabricate when tested. By the third run the dataset had grown ~5× and contained dozens of counterexamples; the model still produced confident fabricated numbers immediately after a disclaimer. Three out of three attempts, no measurable improvement.
One exception that deserved a hard look Only on the final run — 2,349 examples, deployed via Azure OpenAI (deployment suffix protocol0-v5) — did a single test sample come back completely clean for the first time in the entire series.
repliedto their post about 13 hours ago
I want to share something personal that turned into a full system architecture: a real attempt at AI that doesn't lie. A tight coupling between protocol design, attention/cognition management, and a self-hosted Web Shell.
A year ago a creative block cracked open and I started drawing in my own visual language. Six months ago I decided to take it further — international art platforms, press. When I had to prepare materials for a feature on COHART, I found myself, like everyone else, working against algorithms and tools everyone calls "AI."
That's where I hit the actual problem.
Deterministic code goes from point A to point B and you can trace every step. A language model, by contrast, can hedge, deny, or hallucinate — assert something happened (a verification, a lookup, a fact) that never did. As someone with ADHD, where cognition runs at full speed through a normal day and jumps topic constantly, it's easy to lose the thread of why an explanation started where it started. As someone who also experiences dissociation, I understood fast that existing tools simply aren't built for this kind of cognition. They force you to adapt to them instead of the other way around.
So I wrote a protocol for the algorithm — defined exactly how it's allowed to behave.
I'm not a programmer. I refuse to learn to code in the traditional sense. I don't think in syntax — I think in protocols, I see logic, I design system architecture.
It started from needing an AI that doesn't lie. It began on a phone, with strict logging and tagging of every step, so the system could never deny that a piece of information existed. From there it grew into a flexible shell, and today it's a 24/7 server administered remotely from mobile (recently extended to a laptop node too).
Current architecture:
14 verification checks + 4 protection layers (Guard → Guardian → Executor → Audit)
10 logical layers (L1–L10) so no component's role or scope ever blurs into another's
Full integrity chain: every file is Organizations
