What @The_CodingJesus haven't understood about LLMs:
1. Nature of the Tool: LLMs are not just natural language code builders; they aren't a new "programming language." While used for this, it is far from their ultimate purpose. You don't need to be a coder to master an LLM; it is far more useful to understand linguistics, logic, and philosophy. Your critique doesn't stem from underestimating the AI, but from a misunderstanding of human nature that leads you to overestimate human capabilities. When we communicate, we are essentially performing the same operation as the machine: we are organic text-autocompletes. The difference is merely topological.
2. The Nature of Promises: The "unmet promises" you mention aren't inherent failures of LLMs. They are the same unfulfilled promises found in politics (ending hunger) or medicine (curing cancer). Achieving these doesn't depend on the tech itself, but on the real interests of those investing in it. The market logic remains unchanged: it’s not about the end goal, but profitability.
3. Economic Vitality: Claiming the business model is "unsustainable" is inconsistent with how markets work. Do you truly believe that the massive investments from the world's most powerful entities are unjustified? Are they "betting it all" by mistake? The fact that they sustain something seemingly in the red should be proof enough that this isn't a mere "business venture"—it is a vital infrastructure. To see it as a failure of logic is, frankly, naive.
4. Hallucinations and Errors: This is the most telling point. If you hire the best architect and engineer to build a skyscraper, is that a 100% guarantee the building will never fall or stay exactly on budget? AI "hallucinations" shouldn't be seen as a flaw, but as a reflection of their proximity to us. "To err is human." Excel spreadsheets don't err; humans do. Using an LLM as a pocket watch or a calculator shows a lack of understanding of both physics and basic philosophy. If you think "time" is simply what a clock shows, you are missing the point entirely.
5. Communication and the "Intern" Myth: Your inability to communicate a task to an AI is not necessarily a failure of the AI. If you want it to act like you, you must teach it to be you. The claim that an "intern" wouldn't make these mistakes is hyperbolic. If you gave an intern the same vague guidelines, would they deliver exactly what you wanted without training? Of course not. You would have to train them to follow your specific protocol—just like a model.
Conclusion:
Your objections are valid and shared by many, but they restrict the user to a mediocre experience. I am not a programmer, yet by applying philosophy, linguistics, and logic through Claude, I developed a randomness algorithm superior to standard pseudo-random number generators. I am currently developing a new theory of motion.
Philosophy is programming. Communication is programming. These spheres were only separate because philosophers didn't understand code and developers didn't understand philosophy. AI has finally broken that barrier. Today, you can "code" philosophy and test the applicability of your theses in real-time.
I hope this adds a different perspective to the debate. Cheers.
https://t.co/FiO50vReI0
Two Lorenz attractors. Same equations. Same chaos. Only the noise differs.
(It's the first time in history the philosophy creates code, thanks Claude - @AnthropicAI )
Left: Gaussian (PRNG). Right: Contingent (CRNG).
The numbers tell the story:
Gaussian creates 20 artificial regime changes.
CRNG — only 11.
Gaussian vol clustering collapses to 0.07.
CRNG holds at 0.98.
The overlay reveals it all: Gaussian diffuses the butterfly into noise. CRNG preserves the deterministic structure.
Chaos is deterministic. The noise you add should respect that.
Try it yourself: https://t.co/MXpTxKmPum
pip install crng
#chaos #lorenz #python #opensource #crng #datascience #anthropic #claudecode #codephilosophy
The multi-scale convergence chart is the most revealing.
Every asset shows the same pattern: kurtosis INCREASES with scale. Short windows look calm. Long windows reveal the fat tails hiding underneath.
S&P 500: K=2.8 at 60 days → K=26 at 252 days.
This is the phase transition from the spinning coins experiment, happening in real markets, in real time. Supercritical cascade amplification retards convergence toward the Gaussian — and the longer you look, the more you see it.
Your VaR model assumes K=3 at every scale. It's wrong.
I built a real-time market regime detector using CRNG — the Contingency RNG I published last week.
The idea is simple: if CRNG reproduces 86% of real market statistical signatures, then its calibration parameters become a fingerprint of the current regime. When the fingerprint changes, the market changed.
The detector calibrates CRNG via from_data() on sliding windows of returns, extracting three parameters:
Kurtosis (K) — how fat the tails are
Vol clustering (ACF) — how much volatility begets volatility
CRNG Amplification — how deep into the supercritical regime the asset operates
Four regimes emerge from kurtosis alone: CALM (K<5) — Gaussian territory, no fat tails NORMAL (K 5-12) — typical market, moderate fat tails STRESSED (K 12-30) — elevated tail risk CRISIS (K>30) — extreme regime, cascades dominating
Here's what the markets look like RIGHT NOW (March 28, 2026):
S&P 500: CALM (K=2.8) — practically Gaussian Gold: NORMAL (K=6.3) — recently shifted from CALM Bitcoin: NORMAL (K=8.2) — deep in supercritical regime (Amp=3.98) Ethereum: CALM (K=4.9) Oil: NORMAL (K=6.0) Apple: CALM (K=4.7)
But the most striking finding is the MULTI-SCALE CONVERGENCE.
The same asset measured at different time windows reveals the phase transition in action:
S&P 500: 20d → K=2.8 (CALM) 60d → K=2.8 (CALM) 120d → K=3.4 (CALM) 252d → K=26.0 (STRESSED)
The recent crash is invisible at short scales but screaming at the yearly scale. Fat tails dissipate as you zoom in — exactly as the spinning coins phase transition model predicted.
This is not a trading signal. It's a regime classifier. Every Monte Carlo simulation using numpy.random.normal() assumes K=3 always. This detector shows K ranges from 2.8 to 26 depending on the asset and the scale. Your risk model is wrong if it doesn't account for this.
The detector runs on a single command: python regime_detector.py --multi --live
Open source. pip install crng. https://t.co/rpSQZthV4z