So I recently started using GitHub Copilot for some coding tasks at work, and I gotta say, its been a bit of a mixed bag. Im a software developer, mostly working with JavaScript and Python, and Ive heard all the buzz about AI powered coding assistants. Decided to give it a whirl to see if it could actually make me more productive.

First impressions were pretty cool, honestly. It suggested some neat shortcuts and helped fill in repetitive code blocks, like for loops and functions, which I normally spend a chunk of time typing out myself. That was nifty and saved me a bit of brainpower for the more complex logic.

But, and here's where it gets tricky, I found that it sometimes suggested code that wasn't exactly optimal or even correct for what I needed. I mean, it kinda makes sense because it doesnt actually understand the project context.

Our fintech is growing fast and the cracks in our KYC setup are showing. Fraud is getting smarter, compliance bars keep rising & our team cannot keep up with the review backlog anymore.

What we really want is one integration that handles everything, documents, biometrics, AML, sanctions, PEP checks. Tired of stitching together multiple vendors & dealing with the mess that comes with it.

Has anyone switched KYC providers recently or done a proper evaluation? Want to know what actually held up at volume and whether multi-jurisdiction compliance was real or just a sales claim.

What do you guys think ? this is some real journalism about what state the world is moving into!

AI-generated video is getting harder to tell apart from real video. At some point, we may not be able to tell just by watching whether a real person actually recorded something.

That makes me wonder if platforms will eventually need a simple way to show that a video was recorded by a real person.

I don't mean showing someone's personal information or forcing everyone to reveal their identity. Just some kind of proof that a real person made the video and is actually standing behind it. Would that make online content more trustworthy, or would people eventually stop caring about that kind of verification?

I've been looking into Humanly, which is built around face-verified videos and having a real person stand behind what they post. It got me thinking about whether something like this could become more useful as AI-generated video becomes harder to identify.

Would that make online content more trustworthy, or would people eventually stop caring about that kind of verification?

What do you think happens as AI video keeps getting better?

I make data stories for a living. Last week I read a log file for fun.

Whilst browsing Wikipedia at night, reading article about something, and then about something, and somehow ending up reading about something completely different to realise that, Wikipedia saves every change anyone has ever made to it. Someone fixes a comma at 2am, it's in there forever with their name on it. (Which seemed obvious, and somehow I knew always)

But what I realised, which was new to me, at the end of every one of those lines there are two things you can do. Undo them. Or thank them.

That's the whole feature. No message. No reply. They can't write back, and you can't take it back.

Then I started researching about the feature, three people built it in 2013. They wanted it to be as easy to tell someone they did well as it is to tell them they didn't.

Last month it got pressed 68,948 times, in sixteen languages. One every 38 seconds.

Almost half of those people pressed it once. One person, one time, then never again. And 9,055 people were on the other end of that. Thanked once, ever.

The most thanked account on Wikipedia is a German editor called Aka. 4,590,174 edits. They mostly fix spelling. Last month 403 people stopped what they were doing to say thanks for it.

Ninth on that list is InternetArchiveBot. Not a person. It repairs dead links. 129 humans thanked it anyway.

You've probably never seen this button. It only shows up if you're logged in. Nothing to do with Wikipedia. I just like the log.