Turn Markdown vaults, GitHub repos, and AI-assisted docs into publishable websites.
Obsidify brings GitHub Sync, MCP Server, AI Assistant, and publishing status into one focused Mac workspace.
I've been harassed for months by #scam callers claiming to be from "PCP Claims", offering a "refund" of ยฃ1400 for my vehicle claim.
They had all my personal information already.
After weeks of simply hanging up, they upped their game and started calling 4/5 times a day.
I finally gave in and played along. They sent a fake contract link asking for a signature, which I tracked to an @amazon #AWS account; filled with everyone's personal information - including their signatures in plain text.
I reported it to Amazon.
Their response? "We can't help! Try working with the scammers to address any remaining concerns."
Then, they notified the #scammer that I'm on to them - who promptly deleted everything and registered a new domain. It's a UK solicitor... who think their tactics keep them hidden.
Expect a call from the SRA.
#pcpclaims #disbarred
The answer can differ depending on your perspective. For instance, if you are a one man dev team in a small company it is likely that company will have you using AI rather than hiring other devs.
However, if you are in an established company that has had great success with various teams. They are unlikely to move away from that model.
Unfortunately, it still makes.things hard for juniors as AI does raise the bar since companies are aware many candidates are likely using AI to land the job.
At least, this has been mt experience. Until you speak with someone from enough companies to discover a real trend there is no real conclusion that can be drawn at this time until we see some stats.
@UK_Daniel_Card Think I've been wrong in my career more times than right ๐. It's the getting things wrong that you learn from and improve from though.
@kaaaash____ Depends what you mean by best ๐. Arch is my daily driver, I like customisability, but boy is it hell if you screw it up and don't set up proper btrfs partitions with automatic backups on every "yay" or "pacman" install.
@WarrenPeace_75@SpergCapital Interesting, I did see this as one of the original posters. But unfortunately can't gather any further details as X deletes metadata. For all I know Sperg gathered it not knowing or it was sent by someone else.
I'll keep an eye out. Thanks, Appreciate the assistance ๐.
I try to keep my profile tech-related. However, I noticed images circulating on here regarding the Edinburgh attack.
I did a small investigation, and some details caught my attention in relation to AI.
Thread Below ๐งต:
@Dhruvam987 We are standing on the shoulders of giants. Just as our ancestors have in other areas. It is reminiscent of how infrastructure has evoled over time, things become so complex it is neigh impossible for a single average person to understand every part of the system
@Zinny_Edmund As a person who is reasonable and lived through a period if my life with no technology in the house, No.
As someone who knows there is a non-zero chance Rokos Basilisk comes to be in the next 100 years:
Yes, absolutely we do ๐.
I try to keep my profile tech-related. However, I noticed images circulating on here regarding the Edinburgh attack.
I did a small investigation, and some details caught my attention in relation to AI.
Thread Below ๐งต:
Another circulated image joins the collection.
This one is especially interesting because the face has been blurred, which appears to be an attempt to make it look more like a legitimate anonymised image. But blurring a face does not solve the provenance problem.
I compared three images:
A = a frame taken from the actual video footage
B = the circulated blurred/annotated high-res-style image
C = the circulated unblurred high-detail image
When B is compared against A, the match is weak:
- 1,092 SIFT keypoints in A
- 1,534 SIFT keypoints in B
- 38 good feature matches
- 5 RANSAC geometric inliers
- -0.042 masked grayscale correlation after alignment
In plain English, the blurred circulated image does not align well with the known footage frame.
But when B is compared against C, the match is stronger:
- 1,534 SIFT keypoints in B
- 6,729 SIFT keypoints in C
- 37 good feature matches
- 8 RANSAC geometric inliers
- 0.533 masked grayscale correlation after alignment
So the blurred circulated image is a weak match to the actual footage frame, but a stronger match to the other circulated high-detail image.
That matters because the real footage is low-quality, blurred, and compressed. These circulated images contain cleaner body detail, tattoo detail, clothing detail, facial structure, and overall composition than the available footage appears capable of supporting.
OpenCV cannot prove that an image is AI-generated. What it can do is test structure, alignment, and similarity. In this case, the comparison supports the provenance issue: the circulated high-detail images do not behave like clean evidence from the known footage.
My conclusion remains that these images are AI-enhanced or reconstructed. If someone disagrees and has the genuine high-resolution source, they should provide it. Until then, these should not be treated as verified evidence.
For those trying to keep people informed as the story develops: use the actual footage frames if you want credibility. I would advise people not to share suspicious images as fact.
It is rather disturbing, especially for anyone in or around the AI space, how many people seem to have accepted these images without checking them.
I am considering putting together a short guide for ordinary users on how to spot AI-enhanced content, not just fully AI-generated content.
For anyone wanting to replicate this comparison: use Python and OpenCV, mask the annotations, run SIFT feature matching, and compare.
I try to keep my profile tech-related. However, I noticed images circulating on here regarding the Edinburgh attack.
I did a small investigation, and some details caught my attention in relation to AI.
Thread Below ๐งต:
You are moving the goalposts ๐
When it suited you, the circulated images โmatchedโ my real footage frames. Then they were apparently the โexact same pictureโ. Now suddenly it is โnot the picture I postedโ.
That is exactly why provenance matters.
I am not arguing that every circulated image is the exact same picture. I am saying the high-detail circulated images are not behaving like clean evidence from the real footage. And frankly, I agree: this is not exactly the same as the blurred-face image you posted. But it is close enough to warrant comparison, especially when both are high-detail circulated images claiming to represent the same scene.
So I compared three images:
A = frame I took from the actual footage
B = blurred/annotated circulated image
C = unblurred high-detail circulated image
B vs A is a weak match:
- 1,092 SIFT keypoints in A
- 1,534 SIFT keypoints in B
- 38 good matches
- 5 RANSAC inliers
- -0.042 masked grayscale correlation after alignment
B vs C is a stronger match:
- 1,534 SIFT keypoints in B
- 6,729 SIFT keypoints in C
- 37 good matches
- 8 RANSAC inliers
- 0.533 masked grayscale correlation after alignment
The funny part? ๐คญ
The blurred circulated image is a weak match to the known footage frame taken directly from the video, but a stronger match to the other high-detail circulated image. In plain English, the image you are defending aligns worse with the real footage than it does with another suspicious high-detail circulated image.
The real footage is low-quality, blurred, and compressed. The circulated high-detail images contain cleaner facial structure, tattoo detail, body detail, clothing detail, and composition than the real footage appears capable of supporting. Unless someone can provide the genuine high-resolution source, these images are not verified evidence.
You do not have to take my word for it either. Run the comparison yourself with Python and OpenCV: mask the annotations, use SIFT feature matching, then compare B against A and B against C.
Use the actual footage frames if you are trying to draw conclusions. I am pointing this out because these images are clearly not clean evidence from the footage, and people are going to damage their own credibility by using suspicious AI-enhanced or reconstructed images as proof ๐.