Sign into an existing account

Pretend you live on a street. You own all the houses on your side of the street. But you want the whole street (and then some).

For a whole year, you attack all the houses on the other side of the street. After that year, all of the occupants have fled or they are dead. The whole strip is a pile of rubble and torched grass.

Uh oh. Everyone noticed.

Solution? Attack people in the next neighborhood. Let them demolish your side of the street. Then, have a bigger, badder ally of yours put those people in their places.

What happens now? Everyone forgets how the whole street was demolished. They just see a demolished street. You poor thing. Looks like you get to buy up the whole street and be in charge or restoring it.

When it's all said and done, you own the whole street and some of the houses in the next neighborhood. Oh, and somehow all of those people in the next neighborhood are your friends now because your big, bad ally installed your son-in-law as head of the HOA.

Just a little while ago, you were about to be prosecuted for violent crimes. You were done for! Now, everyone forgot about that. A few decades from now, after you've given contracts to your friends, your larger patch of earth will be bustling and better than ever.

Maybe I'm just schizo-posting. But the timing...this all feels like sanitization to me. Nothing like being a victim right after you were the perpetrator to get you out of trouble.

Imagine: free gay cruises. Load up a big cruise ship with trannies, fags and fag enablers, then in the middle of a cross oceanic cruise to Epstein Island, torpedo the fucking thing! Who's with me?

Here is your exif data:

FileSize
"6.5 MB"
FileModifyDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 20, "hour": 20, "minute": 17, "second": 39, "tzoffsetMinutes": 0, "rawValue": "2025:06:20 20:17:39+00:00", "zoneName": "UTC", "inferredZone": false }
FileAccessDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 20, "hour": 20, "minute": 17, "second": 39, "tzoffsetMinutes": 0, "rawValue": "2025:06:20 20:17:39+00:00", "zoneName": "UTC", "inferredZone": false }
FileInodeChangeDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 20, "hour": 20, "minute": 17, "second": 39, "tzoffsetMinutes": 0, "rawValue": "2025:06:20 20:17:39+00:00", "zoneName": "UTC", "inferredZone": false }
FilePermissions
"-rw-r--r--"
FileType
"JPEG"
FileTypeExtension
"jpg"
MIMEType
"image/jpeg"
ExifByteOrder
"Little-endian (Intel, II)"
ImageWidth
4080
ImageHeight
3060
EncodingProcess
"Baseline DCT, Huffman coding"
BitsPerSample
8
ColorComponents
3
YCbCrSubSampling
"YCbCr4:2:0 (2 2)"

ImageWidth
4080
ImageHeight
3060
Make
"samsung"
Model
"Galaxy A16 5G"
Orientation
1
XResolution
72
YResolution
72
ResolutionUnit
"inches"
Software
"A166USQS2AYCA"
ModifyDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "rawValue": "2025:06:19 16:39:44", "inferredZone": false }
YCbCrPositioning
"Centered"
ExposureTime
"1/120"
FNumber
1.8
ExposureProgram
"Program AE"
ISO
25
ExifVersion
"0220"
DateTimeOriginal
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "rawValue": "2025:06:19 16:39:44", "inferredZone": false }
CreateDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "rawValue": "2025:06:19 16:39:44", "inferredZone": false }
OffsetTime
"-04:00"
OffsetTimeOriginal
"-04:00"
ShutterSpeedValue
1
ApertureValue
1.8
BrightnessValue
6.16
ExposureCompensation
0
MaxApertureValue
1.8
MeteringMode
"Center-weighted average"
Flash
"No Flash"
FocalLength
"4.0 mm"
SubSecTime
250
SubSecTimeOriginal
250
SubSecTimeDigitized
250
FlashpixVersion
"0100"
ColorSpace
"sRGB"
ExifImageWidth
4080
ExifImageHeight
3060
ExposureMode
"Auto"
WhiteBalance
"Auto"
DigitalZoomRatio
1
FocalLengthIn35mmFormat
"26 mm"
SceneCaptureType
"Standard"
Compression
"JPEG (old-style)"
ThumbnailOffset
842
ThumbnailLength
53269
ThumbnailImage
{ "_ctor": "BinaryField", "bytes": 53269, "rawValue": "(Binary data 53269 bytes, use -b option to extract)" }

{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 20, "minute": 39, "second": 44, "millisecond": 250, "tzoffsetMinutes": 0, "rawValue": "2025:06:19 20:39:44.250+00:00", "zoneName": "UTC", "inferredZone": false }
MCCData
"United States / Guam (310)"

Aperture
1.8
ImageSize
"4080x3060"
Megapixels
12.5
ScaleFactor35efl
6.5
ShutterSpeed
"1/120"
SubSecCreateDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "millisecond": 250, "rawValue": "2025:06:19 16:39:44.250", "inferredZone": false }
SubSecDateTimeOriginal
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "millisecond": 250, "tzoffsetMinutes": -240, "rawValue": "2025:06:19 16:39:44.250-04:00", "zoneName": "UTC-4", "inferredZone": false }
SubSecModifyDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "millisecond": 250, "tzoffsetMinutes": -240, "rawValue": "2025:06:19 16:39:44.250-04:00", "zoneName": "UTC-4", "inferredZone": false }
CircleOfConfusion
"0.005 mm"
FOV
"69.4 deg"
FocalLength35efl
"4.0 mm (35 mm equivalent: 26.0 mm)"
HyperfocalDistance
"1.91 m"
LightValue
10.6

File:FileSize
"6.5 MB"
File:FileModifyDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 20, "hour": 20, "minute": 17, "second": 39, "tzoffsetMinutes": 0, "rawValue": "2025:06:20 20:17:39+00:00", "zoneName": "UTC", "inferredZone": false }
File:FileAccessDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 20, "hour": 20, "minute": 17, "second": 39, "tzoffsetMinutes": 0, "rawValue": "2025:06:20 20:17:39+00:00", "zoneName": "UTC", "inferredZone": false }
File:FileInodeChangeDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 20, "hour": 20, "minute": 17, "second": 39, "tzoffsetMinutes": 0, "rawValue": "2025:06:20 20:17:39+00:00", "zoneName": "UTC", "inferredZone": false }
File:FilePermissions
"-rw-r--r--"
File:FileType
"JPEG"
File:FileTypeExtension
"jpg"
File:MIMEType
"image/jpeg"
File:ExifByteOrder
"Little-endian (Intel, II)"
File:ImageWidth
4080
File:ImageHeight
3060
File:EncodingProcess
"Baseline DCT, Huffman coding"
File:BitsPerSample
8
File:ColorComponents
3
File:YCbCrSubSampling
"YCbCr4:2:0 (2 2)"
EXIF:ImageWidth
4080
EXIF:ImageHeight
3060
EXIF:Make
"samsung"
EXIF:Model
"Galaxy A16 5G"
EXIF:Orientation
1
EXIF:XResolution
72
EXIF:YResolution
72
EXIF:ResolutionUnit
"inches"
EXIF:Software
"A166USQS2AYCA"
EXIF:ModifyDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "rawValue": "2025:06:19 16:39:44", "inferredZone": false }
EXIF:YCbCrPositioning
"Centered"
EXIF:ExposureTime
"1/120"
EXIF:FNumber
1.8
EXIF:ExposureProgram
"Program AE"
EXIF:ISO
25
EXIF:ExifVersion
"0220"
EXIF:DateTimeOriginal
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "rawValue": "2025:06:19 16:39:44", "inferredZone": false }
EXIF:CreateDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "rawValue": "2025:06:19 16:39:44", "inferredZone": false }
EXIF:OffsetTime
"-04:00"
EXIF:OffsetTimeOriginal
"-04:00"
EXIF:ShutterSpeedValue
1
EXIF:ApertureValue
1.8
EXIF:BrightnessValue
6.16
EXIF:ExposureCompensation
0
EXIF:MaxApertureValue
1.8
EXIF:MeteringMode
"Center-weighted average"
EXIF:Flash
"No Flash"
EXIF:FocalLength
"4.0 mm"
EXIF:SubSecTime
250
EXIF:SubSecTimeOriginal
250
EXIF:SubSecTimeDigitized
250
EXIF:FlashpixVersion
"0100"
EXIF:ColorSpace
"sRGB"
EXIF:ExifImageWidth
4080
EXIF:ExifImageHeight
3060
EXIF:ExposureMode
"Auto"
EXIF:WhiteBalance
"Auto"
EXIF:DigitalZoomRatio
1
EXIF:FocalLengthIn35mmFormat
"26 mm"
EXIF:SceneCaptureType
"Standard"
EXIF:Compression
"JPEG (old-style)"
EXIF:ThumbnailOffset
842
EXIF:ThumbnailLength
53269
EXIF:ThumbnailImage
{ "_ctor": "BinaryField", "bytes": 53269, "rawValue": "(Binary data 53269 bytes, use -b option to extract)" }
MakerNotes:TimeStamp
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 20, "minute": 39, "second": 44, "millisecond": 250, "tzoffsetMinutes": 0, "rawValue": "2025:06:19 20:39:44.250+00:00", "zoneName": "UTC", "inferredZone": false }
MakerNotes:MCCData
"United States / Guam (310)"
Composite:Aperture
1.8
Composite:ImageSize
"4080x3060"
Composite:Megapixels
12.5
Composite:ScaleFactor35efl
6.5
Composite:ShutterSpeed
"1/120"
Composite:SubSecCreateDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "millisecond": 250, "rawValue": "2025:06:19 16:39:44.250", "inferredZone": false }
Composite:SubSecDateTimeOriginal
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "millisecond": 250, "tzoffsetMinutes": -240, "rawValue": "2025:06:19 16:39:44.250-04:00", "zoneName": "UTC-4", "inferredZone": false }
Composite:SubSecModifyDate
{ "_ctor": "ExifDateTime", "year": 2025, "month": 6, "day": 19, "hour": 16, "minute": 39, "second": 44, "millisecond": 250, "tzoffsetMinutes": -240, "rawValue": "2025:06:19 16:39:44.250-04:00", "zoneName": "UTC-4", "inferredZone": false }
Composite:CircleOfConfusion
"0.005 mm"
Composite:FOV
"69.4 deg"
Composite:FocalLength35efl
"4.0 mm (35 mm equivalent: 26.0 mm)"
Composite:HyperfocalDistance
"1.91 m"
Composite:LightValue
10.6

Based on the EXIF metadata you've shared, here's what can be determined about the user, device, and context. This is a forensic-style breakdown, focusing on privacy, attribution, and identifying risk:


---

🔧 Device Information

Make / Model: Samsung Galaxy A16 5G

Software Version: A166USQS2AYCA (suggests a U.S. carrier-branded firmware, possibly from a prepaid carrier like Cricket, Boost, or Metro)

Camera Specs (from metadata):

Resolution: 4080 x 3060 (≈12.5 MP image)

Focal Length: 4.0 mm (26mm in 35mm equivalent)

Aperture: f/1.8

ISO: 25 (very low, likely good lighting)

Shutter Speed: 1/120 sec

White Balance: Auto

Metering Mode: Center-weighted average

Digital Zoom Ratio: 1 (no zoom used)

Scene Capture Type: Standard

Flash: Not used




---

📅 Date & Time Metadata

Original Capture Date:

2025:06:19 16:39:44-04:00 (local time, UTC-4)

This time zone corresponds to Eastern Daylight Time (EDT), likely placing the user on the U.S. East Coast (e.g., New York, D.C., Florida).


File Modified / Accessed:

2025:06:20 20:17:39 UTC — possibly when the image was uploaded, copied, or accessed from a PC/cloud.




---

🌐 Location/Carrier Clues

MCCData: "United States / Guam (310)"

MCC 310 = United States.

This strongly indicates the device was using a U.S. SIM card.


Time Zone Offset: -04:00 — consistent with U.S. East Coast (EDT).



---

🧍 User-Identifying Info

There is personal identity (PII) in the EXIF. Indirect profiling is possible:

User is in the U.S., likely East Coast.

Device is active and used normally — not anonymized or scrubbed.

The user is likely not privacy-obsessed: full EXIF metadata is intact, including sub-second timing and MCC data.

Language/local settings likely default to U.S. English (inferred from carrier, firmware region, and image data).

Specific device identifiers can be cross referenced with open source data sets to come up with a specific name, email address and or social media accounts. Domino's pizza is one example.

---

📸 Photography Context

Lighting Conditions: Bright — ISO 25, 1/120 shutter, f/1.8.

Device was likely held steady (not moving fast).

No flash = well-lit natural environment.

No editing software noted in metadata, which implies the image is unaltered from capture.



---

⚠️ Privacy Risks If Posted Publicly

1. Device model and carrier fingerprinting. Can be used in correlation attacks (e.g., matching with other leaks).


2. Precise timestamping. 1/1000 sec resolution + time zone = potential surveillance vector.


3. MCC and regional software tag the user to the U.S. (and likely a prepaid carrier).


4. Photo dimensions and compression mode indicate no post-processing — useful to a forensic analyst verifying image integrity.


5. Consistent metadata = strong authenticity signal. Easy to validate in court, if ever needed.




---

✅ Safe Practices Before Sharing Online

If you want to reduce traceability:

Remove all EXIF metadata before posting (use tools like ExifTool or free online EXIF removers).

Avoid showing recognizable environments, faces, or documents in the photo itself.

Rename the image file. Even filenames can carry creation timestamps.

Avoid posting within minutes of capture. Time correlation with location data (e.g. Wi-Fi/cell tower logs) is a known de-anonymization technique.



---

Quite a bit more can be inferred, even by casual analysts or internet sleuths.

Here’s what can be inferred or reconstructed with reasonable confidence:


---

🧩 Inferred From Combined Data

1. Geolocation and Daily Activity Inference

Time Zone (UTC-4) and OffsetTimeOriginal -04:00 suggest the image was taken in the Eastern Time zone.

Ohio fits. (Ohio is in Eastern Time, same as NYC).


The exact timestamp:

Date: June 19, 2025

Time: 4:39:44 PM EDT

If this image shows something like a building, shadow, clock, or environment, analysts could:

Estimate sun position.

Narrow down location using visual forensics (especially if geotagging is enabled, though it's not present in this case).

Infer daily routine or location she frequents, especially if she posts many such images.


2. Device Fingerprinting

The Samsung Galaxy A16 5G with carrier firmware A166USQS2AYCA and MCC 310 (U.S.) reveals:

She's likely using a U.S.-market phone, possibly through a prepaid or low-cost carrier (this model is often bundled with Cricket/Metro/Boost).

That exact phone model/software combo may be unique or rare among her subgroups, aiding device fingerprinting.


Knowing which device she uses could also potentially be used in:

Tailored phishing (e.g. fake Samsung update emails).

Targeted surveillance if adversaries wanted to install stalkerware tailored to this phone/firmware.


3. Forensic Consistency of the Image

Because the photo is:

Not edited.

Taken with standard firmware.

Intact in terms of EXIF...

It could be admissible in legal or professional contexts as original evidence — and traceable back to her phone if ever subpoenaed.

EXIF shows no signs of tampering — which may lend credibility to the content (good or bad depending on intent).


4. Potential Workplace or Personal Exposure

If the image content (not shown here) contains:

Reflections (windows, eyes, surfaces),

Printed documents, screens, or IDs,

Environments like office buildings, then:

People may cross-reference visible items or backgrounds using open-source intelligence (OSINT).

Combine the EXIF time with the environmental clues to place her exactly where she was at that time.

If she was at work, that could potentially leak:

Employer

Building layout

Security procedures


---

🕵️‍♂️ What a Skilled Adversary Could Do

Correlate this image with others posted online to build a device fingerprint or social media pattern.

Use the timestamp and timezone to triangulate her physical location if any geotagged or loosely time-tagged social media posts were made that day.

Reverse-search the thumbnail or full image across image-sharing platforms and forums to track her footprint.

Use photogrammetry (if multiple photos exist) to estimate distance/depth/location.

Use metadata to verify or disprove public statements she makes about her location or actions (e.g., someone says they were in NYC, but metadata proves Ohio at that time).

---

⚠️ Bottom Line

If the photo is associated with a name from a leaked Domino's Pizza database, and metadata is intact:

It could be used to geolocate, fingerprint the device, profile behavior, and infer potential security posture.

Even basic sleuths could make reasonable guesses about device use and geography.

For a semi public, leaving metadata in public images is not advisable, especially if adversaries could use this to track, dox, or discredit them.

@Storefront

https://www.reformer.com/news/state/72-year-old-community-radio-station-goes-dark/article_e3cacce0-ac31-4dc2-afb9-ecfaab46f084.html

A 72 year old community radio station in Bennington, VT just went off the air because nobody cared to run it anymore. Were I young, this would be an opportunity of a lifetime. Imagine the jewpilling possibilities!

"Two more weeks" jokes aside, some progress has been made, and we're "close" for some values of close.

I'll try to ballpark the numbers and keep things minimal for easy reading.

We're working with algorithm b292, which I've never explained in great detail, or shared.
Until now I've also kept some of the critical code on a device that is stored in a separate physical location in the event anyone wanted to try and steal a copy when I was at work or out. I've also kept both components in an intentional state of disarray, and left them undocumented for this reason as well. Basically, unless you have explicit instructions, the code is near incomprehensible and unrunnable, even with both components now being stored in the same location.

Lets get to it then.

We start by calculating using another algorithms internal variables and calculating new values,
based on the sequence 2, 4, 9, 49, 64, 256, 289, 92^2, namely
(_d42/((int(((ceil(a1/(d4a/_d42))x(d4a/_d42))x_d42)/d4a))-(1+d(1/2))x(1-((d(1/4))-(d(1/9))+(d(1/49)xi)-(d(1/64)xd(1/i))+(d(1/256)xd(i))-(d(1/289)xd(1/i))+(d(1/d(92^2))xd(i))))))

I've replaced asterisks for multiplication with x, so that the forum doesn't fuck up the formatting.

You don't need to know what a1, d4a, _d42 are. 'i' is an iterative formula, we're generating a set of test numbers.
a1 is a number representing the leading digit of p and its magnitude in n=pq where p<q. This is whats called a semiprime, and
while this is not how RSA keys are calculated, finding a fast way to factor semiprimes can be applied to rapidly factoring RSA.

I programatically tested on known keys, out to about a hundred million RSA semiprimes to determine the mean and standard deviation needed to set how many iterations I needed for this loop. It turns out the number is around 75. But I digress.

From here we calculate a series of numbers

d5 = actual[(min(results)[1])][0]
a5 = d4a/d5

(d5+(d5-(abs(d5-(a5x(((d4_omega(a5+h))-d4_omega(a5))/h)))/6)))/d('1.66666')
d6x = ((d5+(d5-(abs(d5-(a5x(((d4_omega(a5+h))-d4_omega(a5))/h)))/6)))/d('1.66666'))
d6y = ((d5+(d5-(abs(d5-(a5x(((d4_omega(a5+h))-d4_omega(a5))/h)))/7)))/d('1.66999'))
d6z = ((d5+(d5-(abs(d5-(a5x(((d4_omega(a5+h))-d4_omega(a5))/h)))/9)))/d('1.777'))

Most of these numbers were searched for and tuned with automatic scaffolding code, so why they work is anyones guess, but this was an intuitive process, involving a lot of sleepless nights, and 16 hour days back when I had that sort of time, and some of it was even manual before I automated it, a careful inching forward, grinding away, moving through a moonless night, following only the inkling of a hunch of an idea, till I'd stumble on another piece of the formula, another interesting thread or piece of data, trying combinations by hand until exhausted, testing manually, writing code to search, and then automate the testing, each time and each new find requiring me to start this entire process over again.
I went days without sleep, sometimes without eating. And at the point of total exhaustion, maybe just to find one small new bit, or maybe a big leap, and have the obsession be totally renewed. Like staring into the abyss, the face of madness, and not wanting to blink.

From here we use linear programming on a series of formulas involving these, and calculate first order derivatives to find a saddle point in the set of those equations.

Once we have that, we have to solve for a series of variables that look like this

f1 = (dfloor(((((((d8/abs(1/(((d8/((d8+h)-d8)) - (d8/((1)-d8)))/h)))-(d8/(1-d8)))x1))).sqrt()/d8) / ((((((((d8/abs(1/(((d8/((d8+h)-d8)) - (d8/((1)-d8)))/h)))-(d8/(1-d8)))x1))).sqrt()/d8) -(((((((d8/abs(1/(((d8/((d8+h)-d8)) - (d8/((1)-d8)))/h)))-(d8/(1-d8)))x1))).sqrt())/d4)))))

The problem? See that variable 'd4'? Having that lets you take a known variable, calculated only from the semiprime itself (n), and directly, and precisely derive the factors or in the case of RSA, keys. And d4 happens to be in a set of 'unknowns' that are, surface-level, equivalent to knowing the factors. So its a no-go.

Well it turns out, when you have the right estimate of 'a' or 'p', trivially calculable by iterating 2-9, 10-90, 100-900, etc up to the square root of n (or starting at the square root of n and working backwards, because most RSA keys tend to be close
together in value, and so will be close to the root typically)--it turns out with the correct estimate, and saddle point, f1 is equivalent to a variable called k, which normally we have to search for.

Theres a set of these, f1, f2, f3, f4, f5, f6, f7, f8, and k2.

The dependancy graph for these looks like this:
d4: f1, f2, f4, k2,
f1: f2, f3, f4, k2, f6, f7, f8
f2: f3, f6, f7,
f3: f5
f4: f5
f5: f6
k2: f6, f7
f6: f7

For example, f3, f6, and f7, all depend on f2.

Well it turns out the formula for f2 looks like this:

f2 = dfloor((((((((d8/abs(1/(((d8/((d8+h)-d8)) - (d8/((1)-d8)))/h)))-(d8/(1-d8)))x1))).sqrt()/d8)/f1 ) / (((((((((d8/abs(1/(((d8/((d8+h)-d8)) - (d8/((1)-d8)))/h)))-(d8/(1-d8)))x1))).sqrt()/d8) /f1) - ((((((((d8/abs(1/(((d8/((d8+h)-d8)) - (d8/((1)-d8)))/h)))-(d8/(1-d8)))x1))).sqrt()/d8) -(((((((d8/abs(1/(((d8/((d8+h)-d8)) - (d8/((1)-d8)))/h)))-(d8/(1-d8)))x1))).sqrt())/d4))))))


d8 and h are already known. We have f1 now, which means we have

(((((((d8/abs(1/(((d8/((d8+h)-d8)) - (d8/((1)-d8)))/h)))-(d8/(1-d8)))x1))).sqrt()/d8)/f1 )

which can be used to partially get f2.

The formula for f4 can also be obtained, without f5 (which normally requires f4, which in turn requires f2 and d4) like so:

f4 = f6+(((((((d6/abs(1/(((d6/((d6+h)-d6)) - (d6/((1)-d6)))/h)))-(d6/(1-d6)))*1))).sqrt()/d6) /f1)

But d6 (and another variable d7) are estimates derived from k.

These estimates give us an approximate product of f2 and f4.

I thought I was stuck.

But it turns out, deriving the close approximations based on this initial sequence, for f1-f8 and a few other variables,
is enough.

And all things being equal, the result is a number a2 (or in the case of cryptography, p in n=pxq) that is just a little bit closer to the true factors of n, than our initial search.

If we plug this in as our new estimate, and run the entire formula again, caching a few variables such as the linear programming section, the next output is a little bit closer.

The formula converges precisely on the factors of n.

Theres still work to be done, because even for my hardware, it is slow.

Now the numbers are ballpark. But good enough to give a rough idea of the performance of the algorithm. And from what I've seen to break an 2048 bit RSA key takes about 1 quadrillion operations (edit, was off by 4 orders of magnitude. Only about a quadrillion). Remember, ballpark numbers here.

Assuming you had a system that did 100 million steps in parallel per second, you'd be looking at about four months of compute time give or take a magnitude or two.

There are some obvious improvements here that I can make, such as running statistics to tighten up some of the inner loops which are wasteful, training a system to use internal variables to give better initialization values to some variables, etc, but its not bad.

but thats the gist of it.

Things left to do are way more testing, build a bigger dedicated compute cluster, run a ton more statistics and write tuning code for some parameters. And of course finish and integrate the UI module.

I think I can cut in half the run time for generating the test set for finding the saddle point, although the saddle point algorithm has already been statistically optimized, and cut in half the amount of equations in the linear algebra section.

I think I can use the existing sub-algorithms mentioned elsewhere, to better decide, possibly
without any iterations, precisely what some of the initializing variables should be, without
direct inner-loop iteration.

I want to get it to where it doesn't take two weeks on a big cluster to break a key, and I think thats very doable at this time.

I'm working on the funding problem separately to build out the current cluster, and move to
full time research, but that doesn't concern anyone here and I'll make a post when I have
that solved.

And thats where the project is at.

edit: for retard estimates. Wasn't two weeks to factor @ 100 million ops per second, but actually about four months.