Janice Sample

📚 PDF ⏳ 6 min 📖 6

Infinite Slop

AI Slop. Essentially bad code written by AI, or a bad article written by AI, or bad AI art or music. Slop! That is until it improves out of sight roughly every 90 days. The code was slop 18 months ago, now nobody writes code, the machine does it. The whole Bitcoin sector has just spent the last month red-teaming its defences against the excellent insertion skills of the new models. 

If it’s slop, how did it hack your email? How did it compromise ColdCard? How did it exfiltrate OpenAI and attack servers at a rival? It’s slop precisely until it isn’t. Perhaps what we need to contend with is not the ability of the machine but the rate of its improvement.

Slop went to the next level this week with the release of infiniteslop.ai. The brainchild of Pieter Levels. On Friday fal released a new model that can generate video faster than its playback speed. Pause for a moment on that. Earlier this year 5 second videos took 2 minutes. Now, 5 seconds takes less than 5 seconds. 

The idea of infiniteslop is simple. Type what you want to see on the screen, it joins the queue and it is then woven into the permanently running video. As I am writing this a chicken is surfing a grand piano which is collapsing down the stairs. It’s utterly mindless, and yet it isn’t. 15,000 people are watching with me. 

He wrote this on Friday. 12 hours later it was built and sponsored by fal (because producing video is still very expensive $4,300 per day for a continual 24 hour stream.) 

Have a look at infiniteslop.ai. Opinions differ, from “it's completely stupid” to “we have crossed the rubicon”. I’m in the latter camp (obviously) because the machine is generating realities quicker than reality. Contend with that thought for the weekend and note the speed of improvement of the machine, not just the machine. 

If you want to use the original model itself for video generation you need an account with fal.ai. Enjoying it will be difficult because you will be producing at a speed you will never have the time to consume.

KYCeeeeeeeeee

We did rather predict this last week. An ongoing hack at a “major identity verification service”. Nothing confirmed as yet of course but no doubt it will come to light shortly. In the meantime you can buy any of these items at your local dark web provider. Yours sincerely, Janice Sample. 

More of us

My way. It's the only way. 

I would think there is ample evidence to the contrary in the last 100 years demonstrating that Britain's Prime Minister is precisely wrong. It matters not, because Labour MPs elected him for a reason and he is living up to it. They wanted a left wing Prime Minister, not a “poncey lawyer from London” and they got one. 

Burnham has emphasised a philosophy of "aspirational socialism". He is a collectivist, he believes in redistribution and his budget later this year will be very interesting as a result. 

He has already allocated £10 billion to the building of council houses across Britain which will be available for ‘social rent’. They will be the first 70,000 homes of a planned 300,000 new social homes across the UK.

….and is he popular? Yes he is! 46% approval rate with a net favourability rating of +13. The highest of any PM since the early Covid days. We shouldn’t complain either because he will spend like the money is pretend. Which it is. 

It was all going wonderfully well and then this happened.  

To avoid what is commonly known as a “Liz Truss Moment” he then pivoted, promising fiscal responsibility. How though? He pledged 3% of GDP to defence, houses for everyone who doesn’t work and can’t pay, and cheaper everything for everyone. 

The key dates for Andy’s slow motion car crash are as follows:

Labour Party Conference: 27th - 30th September

Budget Date 2026: 28th October 

The conference itself will be ‘peak Andy’. It's in Liverpool, his hometown. He will be lauded as the King of the North. It will be a nauseating lovefest of spending other people’s money, but it will reveal the direction of travel and the bond market will respond to any misstep.

Then October. Taxes have to go up for him to be credible in any way. The bond market also knows taxes have to go up. If they don't, it's Liz Truss time. If they do (they will), it will be full blown communism time because Andy’s idea of taxation is “what’s yours is also mine”. He and his new chancellor have already committed to no increase in income tax (although Andy wants the 50% rate back), National Insurance or VAT, which leaves…..

• CGT equalisation. Basically equalises CGT rates with income tax. Roughly, it would double the rate to 45%.

• The Mansion Tax. Lowering the threshold to £1.5 million, so people in London shoeboxes can also pay it.

• Flat inheritance tax of 10% across the board.

But it will be all right, everything will be all right. The struggle is finished. We have won a victory ourselves. I love Andy, and you love Andy too.

Nobody uses it

There was disbelief at these statistics when they were discussed on X this week. For most people in the world though it is a case of “What’s the S&P500?” If you are reading this you probably know, because you are a westerner with some proximity to finance. Most of the world doesn’t know and does not care.

The distribution is most revealing though. Look how many weak countries are in the top 10. South American basket case economies. War ravaged countries, Ukraine, Iran.  This thing actually works as intended. Look out when those Western currencies really start to wobble. 

Euro-Trash

I thought this was a joke at first. Surely the EU would have learned a lesson after they legislated against AI before it was even born? None of it, ChatGPT is now a ‘Very Large Online Search Engine’.

That classification comes with a number of monitoring obligations that OpenAI will more than likely be able to live up to. More importantly though, they also open themselves up to the following commitment.

“Algorithmic Transparency & Moderation: OpenAI must disclose how its recommendation/search algorithms work, explain content moderation decisions”

Slightly more difficult than that one. Essentially the keys to the OpenAI kingdom. How does your model generate its responses? How does the algorithm work? The answer is partly how the model is trained and partly the weights, and entirely the IP of their business. Are they really going to share this with the European Union, an entity known to be trying to build competitor products?

Why is ChatGPT a Very Large Online Search Engine and DeepSeek is not? What is the point of designating closed models while totally ignoring the far more dangerous and unfettered open models that I can just download from the internet? This one (Qwen3.8-27B-OBLITERATED) refuses nothing. A 27 billion parameter model that will help you do pretty much anything. It will never be regulated by the EU because it is completely open source. So really, what is the point of what they are doing? It serves only to demonstrate they have precisely zero clue what they are actually doing.  

How OpenAI responds here will be fascinating. They may simply issue “Euro-GPT” to European users which will do exactly what the DSA wants and nothing that the end-user wants. We are already at the stage where the Europeans get every new release last, weeks after most of the rest of the world. How long until they don’t get them at all? 

Further information

Our August 2026 report to investors can be found here.