Victoria Wells

@caro_milanesi @AndyAZ Because it’s not secure and would not meet strict data governance rules that most companies have to abide by

@AndyAZ Sure for the integration part but why would they if it drives higher productivity?

Claude plugging into Microsoft 365 is a real threat for Copilot. We have seen this before. When COVID hit, Zoom won on deployment speed and ease of use. IT loved Teams. Users did not follow. Copilot has the same problem. Being bundled into a subscription companies already pay for gets you in the door. Getting users to actually open it is a different challenge. If Claude is more useful and now lives inside the same apps, the license argument stops carrying weight.

@KirstieMAllsopp Offer accepted last July - only exchanged last week!

This is nuts, absolutely crazy, our system is broken.

@DrHoenderkamp @wesstreeting Blood tests are also being rationed almost impossible to get one where I live within three months. Used to be able to just turn up next day and wait

I said it a few weeks ago. Over the last 6 months I have had daily rejections of scans and referrals that as a doctor I feel are clinically needed. It’s a disgrace ⁦@wesstreeting⁩ NHS rations hospital referrals to cut waiting list telegraph.co.uk/news/2026/03/2…

@The_Ambassador_ I read this in the Times years ago and due to my love of satsumas, it’s really stayed in my head!

@The_Ambassador_ Apparently, uk is pretty much the only market for satsumas. The trees have a life span of about 20 years. It’s not that profitable because we’re the only ones interested (clearly not enough other countries have tried them) so they’re dying out

I don't want fucking "easy peelers", I want SATSUMAS.

@silkiecarlo @Apple I’d actually rather Apple did the age check on my device than handing details over to loads of different apps

🚨 Apple's new iOS update has just put an ID check on all Britons' phones. It is beyond outrageous! Unless you give an ID/credit card, you now have a child's device with restricted access to information. @Apple MUST U-turn on this! My comments on BBC: bbc.co.uk/news/articles/…

@PhilipDBunn I hate the grammar checks too. They always want to take the words out of my writing that makes it my style.

A lot of people responding things like "lol, do you have the same criticisms of the internet and spellcheck?" are surprised to learn the answer is yes, obviously.

@soniasodha It will give them the chance to try and push through lesser charges to try and get them under the 3 year prison sentence and emotionally blackmail the victims into accepting this.

I cannot bear the government’s “if you don’t support restricting access to jury trial you don’t care about rape victims” schtick. We’re not that stupid.

The government using “women’s rights” as a cover to drastically restrict access to jury trial is one of the most cynical tactics I’ve ever seen. Glad there are some in the VAWG sector calling it out.

@Office @joshuachuminski I feel like they could have put some effort into making this sound less like a bot

Does anyone know how to turn off @Copilot in @Office? The latest update seems to have either removed the option or hidden it entirely. No, @Microsoft, I do not want your Clippy on steroids drafting this content now, or ever.

@sukh_saroy @AudreySuffolk I take this post more seriously if it hadn’t come across as if written by AI.

🚨Nobody is ready for this paper. Every LLM you use GPT-4.1, Claude, Gemini, DeepSeek, Llama-4, Grok, Qwen has a flaw that no amount of scaling has fixed. They cannot tell old information from new information. A patient's blood pressure: 120 at triage. 128 ten minutes later. 125 at discharge. "What's the latest reading?" Any human: "125, obviously." Every LLM, once enough updates pile up: wrong. Not sometimes wrong. 100% wrong. Zero accuracy. Complete hallucination. Every model. No exceptions. The answer sits at the very end of the input. Right before the question. No searching needed. The model just can't let go of the old values. 35 models tested by researchers from UVA and NYU. All 35 follow the exact same mathematical death curve. Accuracy drops log-linearly to zero as outdated information accumulates. No plateau. No recovery. Just a straight line to total failure. They borrowed a concept from cognitive psychology called proactive interference old memories blocking recall of new ones. In humans, this effect plateaus. Our brains learn to suppress the noise and focus on what's current. LLMs never plateau. They decline until they break completely. The researchers tried everything: "Forget the old values"- barely moved the needle Chain-of-thought- same collapse Reasoning models- same collapse Prompt engineering- marginal improvement at best But here's the finding that should reshape how you think about AI infrastructure: Resistance to this interference has zero correlation with context window length. Zero. It only correlates with parameter count. Your 128K context window is not memory. It's a junk drawer that the model can't sort through. The entire AI industry is charging you for longer context. This paper says context length was never the problem. If you're building agents, memory systems, financial tools, healthcare pipelines, or anything that tracks changing data over time you are building on top of this flaw. And almost nobody is talking about it.

@SamCoatesSky It seems mad no one can see how dangerous they are. The media needs to do a better job of exposing them.

The Greens are now the number 1 pick for all voting age groups under 50, in today’s YouGov poll

@formularacers_ @Planet_F1 Most of the European teams have to fly through or stop in UAE airports to get to Australia.

🚨 | BREAKING: FOM Statement. "Our next three races are in Australia, China and Japan, not in the Middle East - those races are not for a number of weeks. "As always, we closely monitor any situation like this and work closely with relevant authorities." [@Planet_F1]

@MrGMYT I’ve been trying to walk and use the ground mounts as much as possible to enjoy the game play a bit more. I’ve found some spots I can’t get to on foot but it’s nicer than just whizzing up and down all the time

This might be a hot take. But I actually wouldn't have been against locking flying until you've finished the campaign

@_TheSmartAlec1 They have 500 million users…they aren’t sunsetting anything

I FUCKIN KNEW IT LMAOOOOOOOOOO EVERY COMPANY WILL BE CANNIBALIZED BY AI TO AVOID THE BURSTING BUBBLE

@AnthropicAI Shouldn’t you sort your own security out rather than expect everyone else to do it for you?

These attacks are growing in intensity and sophistication. Addressing them will require rapid, coordinated action among industry players, policymakers, and the broader AI community. Read more: anthropic.com/news/detecting…

We’ve identified industrial-scale distillation attacks on our models by DeepSeek, Moonshot AI, and MiniMax. These labs created over 24,000 fraudulent accounts and generated over 16 million exchanges with Claude, extracting its capabilities to train and improve their own models.

@hannahsbee @Transgendertrd I believe Wes Streeting knew this would happen and this is why the trial plans got as far as it did. I think he knew full well that when the tried to plan the trial that they wouldn’t be able to make it safe, it he had to let it play out

🚨BREAKING: NHS-backed puberty blocker trial PAUSED as regulator raises safety concerns. The Medicines & Healthcare products Regulatory Agency (MHRA) has requested the King’s College trial team amend study protocol, to better reflect risks to children 🧵 assets.publishing.service.gov.uk/media/6998b06d…

@edzitron Definitely send me an app instead of a gif 👍

??? What? What are you on about lol

@edzitron So the Ai Infra version of WeWork?

@mickeyfromtheuk I think it will go back to looking elegant! I’ve always disliked the pink version!

🏰 Some people are a little concerned that when Cinderella Castle gets repainted from its current pink back to grey, it's going to look bland. I personally don't think it'll look bland. While it is going back to the old (and in my opinion, superior,) colour scheme, looking at the concept art I think it won't go back to exactly how it was. The new colours, while still being greys and blues, I think will pop and look really nice and majestic. I can't wait to see it.

@BBCBreakfast Yes - and the faded road markings don’t help in the dark with the extra bright headlights everyone has. At least if you have strong white lines with reflective paint, you can still just about see what’s going on

'Everyone's safety depends on road markings' Simon Williams from the RAC spoke to #BBCBreakfast about a survey suggesting one in six UK drivers have had a near-miss with another vehicle because of faded road markings. bbc.co.uk/news/topics/cx…