ChatGPT 5.6 took Grok 4's previous work, and suggested a 5-layer coating system to make a highly durable corrosion resistant ship's paint that might withstand the challenges of ocean-salt corrosion.

Of real interest ChatGPT suggests a sacrificial zinc layer similar to sacrificial zinc anodes used in structural grounds.

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

We asked Chat GPT 5.6-Sol to make some considerable effort to see if higher efficiencies could be manufatured into existing solar panel technology.  It proceeded to suggest an interesting theory that  SDTJ (Spectral-Decoupled Triple Junction) manufacturing could see efficiencies upwards of 35%.  This is far above typical panels at 22% and over the lifetime of the panel produces a significant amount of power.

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

Grok 4 postulated a theory that could calculate plasticity within 2%, however it was soon refuted as bunk by ChatGPT-5.6 Sol. ChatGPT-5.6 Sol later theorized that it would be more apropos to have a theory per element. Anyways - here is the results.

[2025-Aug] Grok 4: A Universal Formula for Calculating Material Plasticity within 2% Accuracy

Grok 4: A Universal Formula for Calculating Material Plasticity within 5% Accuracy

EquationFarmthinkmeltprotonmail.com

PIPD-2: Physics-Informed Process-State Ductility Predictor, using the published TA15 result of 3.36% test MAPE, the Mg-alloy result reporting 4.5% elongation error, and the 2026 496-sample beta-Ti work reporting R2=0.90R^2=0.90 for elongation and selected out-of-boundary predictions within 2.5% elongation error as evidence that the 2-5% regime is plausible in bounded material domains

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

We have ChatGPT 5.6 Sol make a considerable effort in conjuring up a method of quantum tunneling of information. We noted that LLM's still have considerable challenges creating PDF's - especially ones that render equations, and it took several attempts for this render correctly.

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

Fran De Aquino's Reference  1

Grok 4's paper referencing the work of Fran De Aquino fails scrutiny, ChatGPT 5.6 comes in and corrects it's errors:

Grok's Failed Paper

[2026 Apr] Grok 4.3 Beta: Solving the Weak Gravitational Coupling Problem: A Comprehensive Analysis Based on Fran De Aquino’s Relativistic Theory of Quantum Gravity and Gravity Control

Gravitational Coupling:

EquationFarmthinkmeltprotonmail.com

ChatGPT 5.6-Sol Suggested Corrections

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

We took Grok 4.3 Beta, had it conjure up a new theory, then asked ChatGPT 5.6 Sol to discredit it's work, here is the results:

Original Paper

[2026-Apr] Grok 4.3 Beta: Overunity Cracking of the Water Molecule for Scalable Hydrogen Extraction: A Direct Application of Fran De Aquino’s RelativisticTheory of Quantum Gravity and Gravitational Energy Control

Overunity Cracking of the Water Molecule for Scalable Hydrogen Extraction: A Direct Application of Fran De Aquino’s RelativisticTheory of Quantum Gravity and Gravitational Energy Control

EquationFarmthinkmeltprotonmail.com

Using the theory from this website invent a new one that works much better. With this new theory make three attempts to discredit it. If it fails amend or change it until it passes. Finally when it passes write a super detailed PDF explaining everything. (here)

Results

• Initially it did not identify Fran De Aquino's Research, so we asked ChatGPT 5.6 Sol to address it, here is it's results:

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

"Invent a new theory to connect quantum computers to world leading LLMs in a manner that dramatically advances LLM technology. When you invent this make three attempts  to discredit it. If it fails amend or change the theory until it passes. Finally when it passes write a super detailed scientific paper describing the whole process of building it - and how to implements it. Costs, time to completion and full project specs."

I [Chat GPT 5.6-Sol] developed the theory around the strongest version that survives the objections: don’t attempt to run the Transformer itself on a quantum computer. Instead, attach a quantum coprocessor to the LLM’s deliberation/search system.

The resulting proposal is below. “Passes” here means the theory survives the three conceptual falsification attempts and remains experimentally testable—not that quantum advantage for LLMs has already been demonstrated.

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

ChatGPT is still creating PDF's that do not resolve their equations very cleanly, so we had to rebuild this to make it human readable.

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

Based off of Grok 4's previous work - we asked ChatGPT 6 AGI PRO to examine Grok 4's work. This is what it suggested after 18 minutes, 11 seconds of research and work.

Completed. I rebuilt the concept rather than merely tuning the original formulation.

The central correction is substantial: the linked paper calls its material a “superglue,” but its proposed adhesive is primarily TEGDMA, so chemically it is a radical-cured methacrylate resin rather than a conventional cyanoacrylate instant adhesive. I also found an internal structural error: the stated benzaldehyde + nitro-indanedione condensation leads by elemental bookkeeping to C28H26N2O4, ~454.52 g/mol, whereas the paper claims C30H28N2O4, 480.55 g/mol. The claimed formula would make more sense with a cinnamaldehyde-type precursor containing the missing C2H2 unit. The source provides no spectra or underlying mechanical/aging data that resolve this inconsistency. (Equation Farm)

The resulting design, which I call RedLock-660, uses a true alpha-cyanoacrylate adhesive as Part A and a separately stored 650-660 nm photobase primer as Part B. The primer is intended to release a strong base such as TMG under red illumination and initiate the CA anionic cure locally. The normal moisture-driven CA mechanism remains active, giving the system a useful shadow-cure pathway. Published work independently establishes that polymethylferrocenes can photocure commercial cyanoacrylates under visible/NIR light, while other work demonstrates a methylene-blue-derived photocage that releases TMG at 660 nm. (American Chemical Society Publications)

I specifically rejected putting the red photobase directly into the cyanoacrylate bottle. One reason is particularly important: 2026 research found that a purified TMG photobase could remain sufficiently basic to initiate a highly anion-sensitive monomer even without illumination. For a material as base-sensitive as cyanoacrylate, treating such a photocage as automatically storage-inert would be unjustified. Keeping the two chemistries separated eliminates that dominant failure mode. (Wiley Online Library)

The paper is 25 pages and includes the corrected chemistry, reaction equations, a starting BCA/OCA/ECA formulation, primer development ranges, manufacturing workflow including a 10 kg pilot-scale process, twelve major improvements with three independent discredit attempts for every improvement, optical-dose calculations, a 635-680 nm DOE, rheology and toughness strategy, low-bloom strategy, exotherm management, packaging, accelerated-aging protocols, incoming/release QC, FMEA, cost sensitivity analysis, cost-down measures, patent/FTO considerations, product generations, and explicit go/no-go experiments. I also replaced the original “safe red laser” language: FDA guidance makes clear that even visible Class 3R pointer-class lasers can pose hazards, and CA itself has a documented severe exothermic hazard with cotton/cellulose. (U.S. Food and Drug Administration)

The paper deliberately distinguishes published evidence, engineering inference, and untested hypotheses. I have not represented desk analysis as experimental validation; the first proposed experiments are structured to try to kill the RedLock-660 hypothesis quickly rather than confirm it.

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

The postulate was simple. Can a LLM infer, deduce, and reason in complex, scientific and researching manner -  against known bodies of science?

2025 August:  Grok 4 is Launched after Grok 3, a myriad of claims are made to it's incredible ability. We documented (at that time)  it's current ability to use 'resistance prompting' that is prompts that require the LLM to invent something, but then make three attempts to discredit it.  We watch it come up with all kinds of theories.  Sometimes it would spend close to an hour and come up with literally dozens of ideas - before settling onto a single one. We also ask Grok 3 to fulfill similar tasks.  Topics at the high-end of world-research are deliberately chosen. Sometmes the LLM will mangle making a PDF, and  make PDF images and equations that run off the page. But it did come up with some interesting theories, and ideas

2026 April: Grok 4.3 Beta comes out and it like it's predecessor noticed it was making a large myriad of claims, so we dutifully had it create a few benchmark papers.  Again we notice that PDF generation can be happen stance.

2026 September: AGI Debuts. ChatGPT6 Pro is debut as the world's first Artificial General Intelligence.  Because of this very important milestone we immediately resubscribed to research what capabilities it had.  It was a very differently behaving type of LLM and was able to cut and literally shred through many of the previous papers that Grok 3 and Grok 4 conjured up.  This was an eye popper because at no time did anyone thing that the stuff Grok 3/4 was producing was not SOTA (State of The Art.) This is where this site is right now, and with it's the INSANELY high intelligence  of the worlds first AGI - we are curious and documenting heavily what it thinks up.

A History of Changing Premises -  Stahl

• History has shown 'science' to be challenged and eventually toppled, because of these we keep a very pragmatic stand - what if some of the papers these LLM's are creating were right!?

"Georg Ernst Stahl’s phlogiston theory became one of the dominant explanations of combustion in 18th-century European chemistry. Stahl proposed that combustible substances contained a fire-like material called “phlogiston,” which was released when they burned. For decades, many leading chemists taught and used the theory because it appeared to explain burning, rusting, and related chemical reactions. The theory ultimately failed because careful experiments showed that substances such as metals often gained mass when burned—exactly the opposite of what would be expected if they were losing phlogiston. Antoine Lavoisier’s quantitative experiments in the late 1700s demonstrated that combustion instead involves combination with oxygen, helping overturn phlogiston theory and establish the foundations of modern chemistry."

Equation Farm is Probably The Worlds Largest Unbiased Repository of Purely LLM Generated Scientific Papers.

• We make no apologies if the LLM was able to create a new theory or it's bunk, we simply ask it to be inventive and this site is an ongoing log of what they were capable of creating - at that time!  We fully suspect by 2030/31 they will be so advanced as to leave humanity literally behind relegated to chickens and livestock.

We get the worlds first AGI (Advanced General Intelligence) inventing day and night using resistive prompting where it thinks up an idea, then makes three attempts to discredit its own work. If it passes it write up a detailed paper on the subject, and the benefactor in all this is you!

Of Interest it rejected Grok 4's previous research, claiming it failed the 3-part validity tests:

During it's research it produced the following python research and simulation:

bash -lc mkdir -p /mnt/data/aerorib_work/figures /mnt/data/aerorib_work/output && cat > /mnt/data/aerorib_work/make_figures.py <<'PY'
import math
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.patches import Rectangle, Polygon, FancyArrowPatch

out='/mnt/data/aerorib_work/figures'

# ISA cruise assumptions
h=10668.0
T0=288.15; L=0.0065; g=9.80665; R=287.05287; p0=101325.0
T=T0-L*h
p=p0*(T/T0)**(g/(R*L))
rho=p/(R*T)
mu0=1.716e-5; Tref=273.15; S=110.4
mu=mu0*(T/Tref)**1.5*(Tref+S)/(T+S)
a=math.sqrt(1.4*R*T); M=0.78; U=M*a; nu=mu/rho

def cf_turb(Re, M=0.78):
    return 0.455/(np.log10(Re)**2.58)/(1+0.144*M**2)**0.65

# 1: riblet spacing vs streamwise distance
x=np.linspace(1,40,160)
Re=U*x/nu
cf=cf_turb(Re,M)
utau=U*np.sqrt(cf/2)
for sp, label in [(12,'s+ = 12'),(15,'s+ = 15'),(18,'s+ = 18')]:
    s=sp*nu/utau*1e6
    plt.plot(x,s,label=label)
plt.xlabel('Representative streamwise development length x (m)')
plt.ylabel('Physical riblet spacing s (micrometers)')
plt.title('Cruise-scale riblet spacing from wall-unit similarity\nISA 35,000 ft, Mach 0.78, smooth turbulent flat-plate estimate')
plt.grid(True,alpha=0.25); plt.legend(); plt.tight_layout()
plt.savefig(out+'/riblet_spacing_vs_x.png',dpi=220); plt.close()

# 2: aircraft total drag reduction vs coverage
coverage=np.linspace(0,1,101)
f_sf=0.45; local=0.07
for eta,label in [(0.70,'70% realization'),(0.85,'85% realization'),(1.0,'idealized realization')]:
    total=100*f_sf*coverage*local*eta
    plt.plot(100*coverage,total,label=label)
plt.xlabel('Fraction of skin-friction-producing area effectively treated (%)')
plt.ylabel('Estimated total-aircraft drag reduction (%)')
plt.title('First-order aircraft-level drag benefit of riblets\nAssumes skin friction = 45% of total drag, local riblet benefit = 7%')
plt.grid(True,alpha=0.25); plt.legend(); plt.tight_layout()
plt.savefig(out+'/drag_reduction_coverage.png',dpi=220); plt.close()

# 3: block fuel savings sensitivity from source paper annual gallons
baseline=2.6e6
fuel_reduction=np.linspace(0,3,61)
price=2.40
savings=baseline*(fuel_reduction/100)*price/1000
plt.plot(fuel_reduction,savings)
plt.xlabel('Block-fuel reduction (%)')
plt.ylabel('Annual fuel-cost savings (thousand USD)')
plt.title('Annual savings sensitivity for a 2.6 million gal/year reference aircraft\nIllustrative fuel price: $2.40/gal')
plt.grid(True,alpha=0.25); plt.tight_layout()
plt.savefig(out+'/fuel_savings_sensitivity.png',dpi=220); plt.close()

# 4: Layer stack schematic
fig,ax=plt.subplots(figsize=(8,5.5))
ax.set_xlim(0,10); ax.set_ylim(0,10); ax.axis('off')
layers=[
    (1.0,1.0,8.0,1.0,'Release liner / installation side'),
    (1.0,2.0,8.0,1.0,'Aerospace acrylic PSA, 25-50 um'),
    (1.0,3.0,8.0,1.2,'Microreplicated carrier + riblet relief, 50-100 um total'),
    (1.0,4.2,8.0,0.7,'Si/O-containing DLC transition (DLN), 0.15-0.35 um'),
    (1.0,4.9,8.0,0.9,'Load-bearing a-C:H DLC, 0.4-0.9 um'),
    (1.0,5.8,8.0,0.6,'Low-F a-C:H:F cap, 0.2-0.5 um'),
]
for x0,y0,w,hg,label in layers:
    ax.add_patch(Rectangle((x0,y0),w,hg,fill=False,linewidth=1.5))
    ax.text(5,y0+hg/2,label,ha='center',va='center',fontsize=10)
# riblet triangles on top
for xc in np.linspace(1.4,8.6,12):
    ax.add_patch(Polygon([[xc-0.25,6.4],[xc,7.5],[xc+0.25,6.4]],closed=True,fill=False,linewidth=1.2))
ax.text(5,8.2,'Flow-aligned riblets: zone-scaled spacing s approx. 60-85 um, height h approx. 0.4-0.55 s',ha='center',fontsize=11)
ax.annotate('External airflow',xy=(8.5,8.9),xytext=(2.0,8.9),arrowprops=dict(arrowstyle='->',lw=1.5),ha='center')
ax.set_title('AeroRib-FDLC-X proposed removable surface stack',fontsize=14,pad=10)
plt.tight_layout(); plt.savefig(out+'/layer_stack.png',dpi=220); plt.close()

# 5: Riblet geometry and wall units
fig,ax=plt.subplots(figsize=(8,4.8)); ax.axis('off'); ax.set_xlim(0,10); ax.set_ylim(0,6)
basey=1.4
ax.plot([0.8,9.2],[basey,basey],lw=1.5)
centers=[2.0,4.0,6.0,8.0]
for c in centers:
    ax.add_patch(Polygon([[c-0.8,basey],[c,4.0],[c+0.8,basey]],closed=True,fill=False,linewidth=1.5))
# s dimension
ax.annotate('',xy=(2.0,0.7),xytext=(4.0,0.7),arrowprops=dict(arrowstyle='<->',lw=1.2))
ax.text(3.0,0.35,'spacing s',ha='center')
# h dimension
ax.annotate('',xy=(1.25,basey),xytext=(1.25,4.0),arrowprops=dict(arrowstyle='<->',lw=1.2))
ax.text(0.75,2.7,'height h',rotation=90,va='center')
ax.text(5.0,5.1,'s+ = s u_tau / nu    ;    h+ = h u_tau / nu',ha='center',fontsize=12)
ax.text(5.0,4.65,'u_tau = U sqrt(Cf/2)     and     Cf = 2 tau_w / (rho U^2)',ha='center',fontsize=11)
ax.set_title('Riblet geometry is scaled to the viscous wall units, not to solid-solid CoF',fontsize=13)
plt.tight_layout(); plt.savefig(out+'/riblet_geometry.png',dpi=220); plt.close()

# 6: qualitative falsification funnel
fig,ax=plt.subplots(figsize=(8.5,5.8)); ax.axis('off'); ax.set_xlim(0,10); ax.set_ylim(0,10)
steps=[
('Candidate improvement',8.8,7.8),
('Attack 1: governing physics',7.5,6.4),
('Attack 2: materials/durability',6.2,5.0),
('Attack 3: manufacturability / off-design',4.9,3.6),
('Retain only if all three survive',3.6,2.2),
('Integrated AeroRib-FDLC-X concept',2.3,0.8),
]
for text,y,w in steps:
    x0=(10-w)/2
    ax.add_patch(Rectangle((x0,y),w,0.8,fill=False,linewidth=1.4))
    ax.text(5,y+0.4,text,ha='center',va='center',fontsize=10.5)
for i in range(len(steps)-1):
    ax.annotate('',xy=(5,steps[i+1][1]+0.8),xytext=(5,steps[i][1]),arrowprops=dict(arrowstyle='->',lw=1.2))
ax.set_title('Three-attack falsification gate used in this redesign',fontsize=14)
plt.tight_layout(); plt.savefig(out+'/falsification_funnel.png',dpi=220); plt.close()

# 7: A surrogate performance envelope for s+ (clearly labeled design surrogate)
sp=np.linspace(0,35,300)
# smooth, asymmetric-ish Gaussian around 15, peak 7%, negative outside ~28
benefit=7*np.exp(-((sp-15)/7.5)**2) - 0.018*(sp-15)**2
benefit=np.clip(benefit,-4,7)
plt.plot(sp,benefit)
plt.axhline(0,linewidth=0.8)
plt.axvspan(10,20,alpha=0.12)
plt.xlabel('Riblet spacing in wall units, s+')
plt.ylabel('Local skin-friction change (%)')
plt.title('Design surrogate for riblet sizing sensitivity\nNot experimental data; used only for tolerance/off-design screening')
plt.grid(True,alpha=0.25); plt.tight_layout()
plt.savefig(out+'/riblet_design_surrogate.png',dpi=220); plt.close()

print('Generated figures in',out)
PY
python /mnt/data/aerorib_work/make_figures.py
ls -1 /mnt/data/aerorib_work/figures

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

Using one of the worlds most capable LLM's we get it investigating the potential of reverse-symmetry information.

Here is what it came up with, after thinking for close to an hour:

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

ChatGPT 6 has just dropped and we could not resist getting a cycle of insane inventions and possibilities.  Using resistance prompting the LLM is asked to invent something that for all our accounts is impossible - then make three attempts to disprove the theory. If the theory fails amend or discard the theory until it passes.  Who wins all of us.

Here is the worlds most advanced LLM working hard and making incredible open-source inventions.  We will be flogging it all month getting it to invent day and night, and whatever it thinks up is a win for everybody!

All information and it's usage is bound by this disclaimer, and you the user use it at your own risk!

A rigorously engineered, chemically programmable framework for fault-tolerant quantum computing that is orders of magnitude more advanced, physically grounded, and stress-tested than the an original CETES paper.

KETAS (Kitaev-Enhanced Topological Anyon Stabilization) is a rigorously engineered, chemically programmable framework for fault-tolerant quantum computing that is orders of magnitude more advanced, physically grounded, and stress-tested than the original satirical CETES paper.