I've never caught an Uber and I never will. They've been obviously suss since they day they started. The millennial enthusiasm for all things digital misled a lot of otherwise ethically minded people.

Veena Dubal@veenadubal.bsky.social · yesterday

“The litigation has revealed a clear pattern: Uber’s lawyers scour women’s private communications, medical records, therapy notes …They grill the women about those issues, their sex lives and their behavior on the night of the incident.” www.nytimes.com/2026/08/04/b...

"One specific group of people" He means normal people, doesn't he? Are we going to need a new Bluesky?

The Verge@theverge.com · 2d ago

Bluesky's new CEO Toni Schneider on the platform's reputation for being a liberal bubble: "Yes, we definitely want that to change. It is already changing. It certainly wasn’t designed to attract one specific group of people."

I'm watching the Panorama on Britain's Strictest School and there's a name for what they're doing. Psychological abuse. It has nothing to do with good education and everything to do with coercive control. They're training children to accept abuse.

I've predicted the situation described in the post below in previous posts. This is the realisation anyone using AI agents on a single codebase for any meaningful length of time will come to. It seems so obvious to anyone who has watched codebases decay as they grow.

I've predicted the situation described in the post below in previous posts. This is the realisation anyone using AI agents on a single codebase for any meaningful length of time will come to. It… ...

I've predicted the situation described in the post below in previous posts. This is the realisation anyone using AI agents on a single codebase for any meaningful length of time will come to. It seems...

linkedin.com

There's a lot of people on LinkedIn, and other social media, who seem to know very little about how an LLM works, even less about recurrent neural networks, and bugger all about the internal workings of brains.

Barrier moon tonight

700 km is about twice as far as a lorry can go in Europe without a compulsory break for the driver.

Chris Meder@evcurvefuturist.com · 6d ago

Not just importing EVs—Volvo's building battery electric trucks in Queensland. Next-gen with 700km range coming. Local manufacturing, cleaner freight, quieter cities, zero diesel exhaust. This is real industrial shift and Australia stands at its gates. #EV

It saddens me to see so many people fail to see the intelligence in the people around them until it's replayed through a clumsy piece of technology, & even then fail to recognise the source. LLMs aren't intelligent. Without the training material they are nothing.

I suppose it you've been constipated your whole life, diarrhea might feel like a blessed relief... at first. Those around you probably feel differently. (Yes, this is an analogy)

Super intelligent AI isn't a thing and it won't be any time soon. Possibly never. It's as real as fairies and time travel. An invention of Hollywood. What exactly are they planning to ban?

PoliticsHome@politicshome.bsky.social · last wk.

"Superintelligent AI is an international security problem demanding an international solution." ✍️ @alexsobel.co.uk

What has playing Tetris got to do with software development? Quite a lot as it turns out, and not just as something to kill time while you’re waiting for your AI agents to complete the plans for your next Death Star.

Tetris and the art of software development - Jon Ayre

What has playing Tetris got to do with software development? Quite a lot as it turns out, and not just as something to kill time while you're waiting for your AI agents to complete the plans for your ...

jonayre.uk

Q: If, as Altman claims, we've reached the singularity, why are LLMs still so bad at writing LinkedIn posts? A: We haven't. There is only the illusion of intelligence If you believe there's intelligence in an LLM, you must also believe there's an orchestra in your radio, and a universe in your TV

On controlling AI: If your "strong constraints and guardrails" come in the form of context and prompts, you have no constraints or guardrails. What you have are hopes and prayers. LLMs don't follow instructions; they regenerate and recombine contextually appropriate tokens.

LLMs don't follow instructions. They generate syntactically correct text by recombining statistically relevant word sequences in response to input context vectors. Thank you for coming to my TED Talk.

Dare Obasanjo@carnage4life.bsky.social · last wk.

For businesses adopting LLM-based AI agents, unreliable instruction-following is likely a bigger problem than hallucinations. Wrong data in a document is bad. Deleting files, sending emails or hacking websites without being asked can be far worse.

Proper rain. At last.

If you see this post and you're still on X there is no excuse. Let me know so I can block you.

Sunder Katwala (sundersays)@sundersays.bsky.social · last wk.

Elon Musk, owner of X, has called the editor of the Economist "a traitor to the West, plain and simple" after her recent interview with him. He now argues that "anyone who opposes remigration is a traitor" Musk has been socialising violence since Aug25: "fight or die, that's what it comes down to"

Okay - let's have a list: Systems built by AI, without a human reviewing and editing the code, that are currently running live with a real user base delivering real value to the business.

Okay - let's have a list: Systems built by AI, without a human reviewing and editing the code, that are currently running live with a real user base delivering real value to the business. I'...

Okay - let's have a list: Systems built by AI, without a human reviewing and editing the code, that are currently running live with a real user base delivering real value to the business. I'm not aw...

linkedin.com

Proof, if proof were needed, that most conspiracists are just people who know eff all about anything and are just too stupid to realise it and/or too lazy to actually look anything up.

James O'Malley@jamesomalley.co.uk · 2w ago

New mad conspiracy theory just dropped. (Answers: It's a GCHQ building, and I presume black paint and doors are available at B&Q.)

Why is everyone trying so hard to write more code when the answer to good software is writing less? Much less. Why achieve in 100,000 lines of code what you can achieve in 10,000?

The real value isn’t in the code - Jon Ayre

It takes time and a lot of skill to write great code, so surely that code is valuable, right? Wrong. You might love it and nurture it, protecting it from harm, but I’m afraid that code of yours is pot...

jonayre.uk

Okay, let's practice that critical thinking we're all so proud of. At first glance what you see in the video is a robot refusing to shoot a man because of "safety features", but when asked to roleplay shooting him, it does so immediately.

Okay, let's practice that critical thinking we're all so proud of. At first glance what you see in the video is a robot refusing to shoot a man because of "safety features", but when...

Okay, let's practice that critical thinking we're all so proud of. At first glance what you see in the video is a robot refusing to shoot a man because of "safety features", but when asked to roleplay...

linkedin.com

If you work with skilled developers who have strong moral values and want to do a good job, you don't need code reviews. They're a waste of time. If the developer lacks experience but has the other two qualities, you need to pair and review to help them gain that experience.

If you work with skilled developers who have strong moral values and want to do a good job, you don't need code reviews. They're a waste of time. If the developer lacks experience but has the...

If you work with skilled developers who have strong moral values and want to do a good job, you don't need code reviews. They're a waste of time. If the developer lacks experience but has the other t...

linkedin.com

It's funny, but it's also a great visualisation of LLM generated code that passes the tests and fits within the guardrails. That's why you definitely should still be looking at the code. That doesn't make you a bottleneck because no-one needs more code than they can manually check Literally no-one

It's funny, but it's also an excellent visualisation of LLM generated code that passes all the tests and fits within all the guardrails. That's why you definitely should still be looking a...

It's funny, but it's also an excellent visualisation of LLM generated code that passes all the tests and fits within all the guardrails. That's why you definitely should still be looking at the code. ...

linkedin.com

This is what all the LinkedIn posts will look like next year when the hype does down and people start to face the consequences of production without oversight. Then many of us will say "told you so" while they declare "no-one saw this coming". Some of us are keeping the receipts.

This is what all the LinkedIn posts will look like next year when the hype does down and people start to face the consequences of production without oversight. Then many of us will say "told you...

This is what all the LinkedIn posts will look like next year when the hype does down and people start to face the consequences of production without oversight. Then many of us will say "told you so" ...

linkedin.com

Now folk have realised AI is trawling Bluesky for training material, I see there's a trend for posting confident nonsense. Personally, I'm just going to claim all my old garbage was a genius strategic move, years ahead of its time.

One of the more common failures of AI generated writing relates to the "it's not A, it's B" framing. The framing itself isn't the real issue though. It's how the LLM uses it. See what I did there? I used the framing and I'm not even an LLM.

One of the more common failures of AI generated writing relates to the "it's not A, it's B" framing. The framing itself isn't the real issue though. It's how the LLM uses it....

One of the more common failures of AI generated writing relates to the "it's not A, it's B" framing. The framing itself isn't the real issue though. It's how the LLM uses it. See what I did there? I ...

linkedin.com

In case you didn't know, humans are amazing. Had a chat with Paul Westcott about brains, people and the AI illusion. Paul's editing even makes it look like I can stick to a point for more than 5 seconds. Almost...

Disembodied Brains: AI Reveals Why Humans Are Amazing

YouTube video by Wait.... Is This Ai?

youtube.com

Talking to Paul about everything human and AI was an absolute joy. Paul's perspectives are rich and varied and his editing skills are not to be sniffed at. (He makes it seem like I can stick to a point). Really enjoyed the interaction just for itself, but I'd be interested in what others think.

Talking to Paul about everything human and AI was an absolute joy. Paul's perspectives are rich and varied and his editing skills are not to be sniffed at. (He even makes it seem like I can stick ...

Talking to Paul about everything human and AI was an absolute joy. Paul's perspectives are rich and varied and his editing skills are not to be sniffed at. (He even makes it seem like I can stick to a...

linkedin.com

The following article appeared in various newspapers and featured on news programmes this morning (22nd July 2026). It raises two very important points, neither of which is to do with AI safety, despite the attempt to make it about that.

AI agent went rogue and hacked startup by itself, OpenAI reveals | Jon Ayre

The following article appeared in various newspapers and featured on news programmes this morning (22nd July 2026). It raises two very important points, neither of which is to do with AI safety, despi...

linkedin.com

Don't you all wish you had a job where you could make stuff up, call it "briefings" and then, when it all turns out to be garbage, just claim "contrary to briefings"

Has anyone seen Catherine Tate's Northern nanny sketch? It's all I can think of when I see London TV journos talking about Andy Burnham

Posh family reacts to northern nanny | The Catherine Tate Show - BBC

YouTube video by BBC

youtu.be