Really? all the media reports contradict this statement.

9% is huge, that is a complete failure, this isn’t some exam, it is people’s lives.

Shoplift get 5 years, defraud investors of $1.8b, get just over a year, justice at its best.

You are absolutely right...:P

I don't know mind people using AI to create open source projects, I use it extensively, but have a rule that I am responsible and accountable for the code.

Social media have become hellscapes of AI Slop of "Influencers" trying to make quick money by overhyping slop to sell courses.

Maybe where you are from the em dash is not used, but in Queen's English speaking countries the em dash is quite common to represent a break of thought from the main idea of a sentence.

While I am fairly skeptical of LLMs and the current version of AI that are being peddled by companies like OpenAI, the main problem is, once again, social media.

That someone is vibe-coding their application, running the symptoms of their illness through ChatGPT, or using it to write the high school essay isn't really a problem. It's also not a massive issue that you can generate random videos, the problem is that social media not only allows propaganda to be spread at the speed of light without any filter and verification, that it actually encourage it in the name of profit.

Remove the engagement retaining algorithms from social media, stop measuring success in terms of engagement. Do that an AI will become much less of a problem. LLMs are a tool for generation, social media is transport mechanism that makes the generated content actually dangerous.

If it is going to be accurate it is PHP.

Some questions I have been thinking about: 1. Do I make it agnostic to Claude?

2. Do I move each of the agents into their own repos?

3. Is it useful?

Strange, I didn’t install an extension or enable it.

It happens even when I go to Google.com. Every search get hijacked.

This is wrong.

All Safari extensions are bundled with apps and require users to manually enable them to function.

I tried installing the ChatGPT app, and noticed that there’s a "ChatGPT Search" extension that provides the functionality you’re looking for. However, enabling it involves a couple of steps: You can either go to Settings > Apps > Safari > Extensions > ChatGPT Search > Allow Extension, or you can tap the left side of the Safari address bar, select Manage Extensions, and enable ChatGPT Search. Hilariously, the extension didn't work when I had DuckDuckGo as my default search engine so I had to change it to Google when I was testing this.

Additionally, to allow the extension to redirect the page from Google, you must navigate to google.com, tap the left side of the Safari address bar, select ChatGPT Search, and grant it permission to access the page content.

Search hijacking just can't happen on Safari without the user going through multiple hoops.

I certainly didn’t do all that.

Is this a new protocol for us to put our remote MCP and remote resources behind?

No. It is the same model without reasoning.

So is maybe gpt-5.2 with reasoning set to 'none' identical to gpt-5.2-chat-latest in capabilities but perhaps with a different system (system) prompt? I notice chat-latest doesn't accept temperature or reasoning (which makes sense) parameters, so something is certainly different underneath?

My guess is they develop multiple models in parallel.