iiinova — where your digital self keeps evolving

As I've been saying increasingly loudly in other places Instinct is not trustworthy, but, oh boy..., it - and other personal agents that are currently US-only - could _significantly_ shift the dial on how people expect to use private and public digital services... www.thetimes.com/business/tec...

With a few clicks, I linked Instinct to my personal Google Drive, including more than 20 years of emails and documents, suppressing my gut feeling that this was a terrible idea. The more access it has to your digital life, the better it can be — but then, the more havoc it could wreak if it goes wrong.
What does it do? Via WhatsApp instructions, it drafts briefs for meetings, digests emails from my children’s schools and adds important dates to my calendar.
Disconcertingly, working out what my job is, Instinct scanned my articles online and offered unsolicited advice in response to a research query: “The full columns are behind the paywall so I can’t read the text — but you did the cost angle on 13 April.”
None of this is rocket science (or solving the Navier–Stokes problem) but it is a huge time-saver.
ADVERTISEMENT
Asked to book a hair appointment for an upcoming trip, intuitive and proactive, it found details in my inbox of where I was staying and what time I was arriving, and pinpointed a convenient salon. Having saved a credit card to Instinct’s secure area — the “vault” — when the bank texted me a secret code, there was an uncomfortable moment when Instinct told me to share it to complete the booking.

000

peterkwells @peterkwells.com · 13h

Today's the day when dogs are free to run kerrazy on Whitley Bay beach again

031

peterkwells @peterkwells.com · 15h

TIL: Fylde Council now requires people to create a "personal waste account" before they can check their bin day new.fylde.gov.uk/wasteportal/

112

peterkwells @peterkwells.com · 29/09/2026

So, I do hope this bit of the GDS team are planning to properly engage across the whole society/economy to understand stuff for future iterations of the PSGA...

GDS is undertaking engagement across the public sector to understand emerging challenges and opportunities that will shape the future iteration of the PSGA. If you haven't already done so, take this opportunity to provide your input and complete GDS's survey.

110

peterkwells @peterkwells.com · 29/09/2026

Me reading some more reports on user-level AI risks that look at the Wrong Thing

Bart Simpson writing lines on a chalkboard that say "most of the public use products, not LLMs"

000

peterkwells @peterkwells.com · 25/09/2026

*makes space on the shelf alongside the 1st edition*

confirmation that a pre-order of Networks of New York 2026 has shipped

000

peterkwells @peterkwells.com · 23/09/2026

Repeat after me: AI detectors like Pangram *do not have a consistent error rate*, it varies due to things like author, their use of language, their writing style, the type of text being produced &c &c &c www.bbc.co.uk/news/article...

Some also say that the programmes are built around the use of standardised language - English or French - and so cannot accurately interpret texts that use dialect from places like the Caribbean.
But the writers behind Balance ton Claude said that Pangram had been shown to make a wrong call about an AI-generated text only one time in every 24,400.

241

peterkwells @peterkwells.com · 23/09/2026

*starts to set up Google's new "family AI agent" designed to let you spend "Less time on logistics, more time for family"* *hits screen with confusing repeated error messages* *looks in settings help & finds useless video* *stares at Google w/pained look in eyes* labs.google.com/cc/u/0/landing

000

peterkwells @peterkwells.com · 22/09/2026

Does this mean that we need to rebrand all our work on AI agents so that it's about Super Intelligentsia?

Trump rebrands Artificial Intelligence to Super Intelligence
published at 16:10
16:10

Nada Tawfik
Reporting from the United Nations
Donald Trump has a penchant for branding and in his speech he gave Artificial Intelligence a new name: Super Intelligence.
"The use of the word artificial makes it sound fake," he said. "It's not fake, it's actually amazing. But we have to be careful," he added. "Welcome to the new world of super intelligence - SI," he went on to say.
Trump has previously argued that whoever wins AI, wins, and he is firmly against regulation. But that puts him at odds with many here.
The question is whether the slow pace of diplomacy can keep up with AI’s rapid development.

001

peterkwells @peterkwells.com · 21/09/2026

Although its designers need to give it a better understanding of when humans actually put bins out....

a text exchange where the AI assistant says it will send the human a reminder to put the bins out at 5.45am, the human politely asked the AI stuff to consider whether they'll be awake at 5.45am, this caused the personal assistant to change the reminder to 8pm on the previous night

100

peterkwells @peterkwells.com · 19/09/2026

The chair of UK government's £500m Sovereign AI investment fund seems surprised that the public expects transparency around how that £500m is spent

A post from James Wise saying "While there’s no requirement to do so we are trying to be as transparent as possible, so in response to ( another! ) FOI @UKSovereignAI added my conflicts of interest process to the site today along with others, for those interested. I hoped we’d put this all up earlier - but sadly I am not the webmaster or else there’d be a lot more gifs.

As a small insight into how difficult it is to do things in Govt, there hasn’t been a SINGLE WEEK since our launch where team members haven’t had to spend hours responding to an FOI request, for things like ‘where are your investment companies based’ ( you can use Google! ) or ‘what copyright laws does a chip company breach’ ( as nonsensical as asking what fruit and veg laws a fighter jet breaches ).

I understand that many FOI requests are now AI generated. But as the civil service isn’t allowed to use AI in the same way to respond to them, the abuse of this system is rife. There was a great piece in the Economist on this asymmetry choking parts of the state recently. "

It’s a significant cost of both our small team’s time and tax payers money. So I’m hoping that by releasing more publicly, even if delayed, it will reduce that burden on them.

Thankfully, SovAI has a killer team who work around the clock anyway - more to come on how they’re working to strengthen the UK’s AI capabilities and economy next week.

1123

peterkwells @peterkwells.com · 16/09/2026

Experts air concern that AI engineers think they've ever looked at "raw code" docs.google.com/document/u/0... (yes, there's a real risk from being too abstracted away from what's going on but "raw code"? lol)

an excerpt from the linked document saying "I myself barely look at raw code anymore"

000

peterkwells @peterkwells.com · 14/09/2026

One day I'll find something tech/data-y that embraces multiple sources of "truths", but, sadly, that day is not today alexlieberman.com/ai-native-co...

a block of text in the middle of which is the sentence "3. Centralized intelligence layer that aggregates structured and unstructured data, documents, and business logic into a single source of truth that's queryable. Agentic work gets done on top of it."

000

peterkwells @peterkwells.com · 13/09/2026

Yeah, no 'model alignment' alone will never be sufficient - there's a whole wide world of humans and environmental factors outside models - and groups like METR are barely independent & clearly not very good at evaluation darioamodei.com/post/we-must...

pacing does not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models, and for third party evaluators to confirm this

Embedded Evaluators. Each frontier AI company commits to giving ongoing, employee-like access to a team of embedded third-party evaluators (such as METR), whose role is to verify adherence to safety practices and commitments, report incidents, and help assess the alignment of not just completed AI models but training pipelines and processes. This is the key step for verifiability of any pacing commitments, and has precedent in the banking industry, which sometimes involves regulatory “supervisors” embedded along with employees. Anthropic is unilaterally committing to this step now. We intend this to be part of a broader push to redouble efforts on our safety and alignment work.
Democratic Coordination. Frontier AI companies within democratic countries coordinate to establish common safety standards as well as limits on the rate of unchecked AI progress. Some forms of coordination that would be impactful for pacing are legally challenging, and will require government support.
Global Coordination. The US and other democratic governments attempt to coordinate with authoritarian governments, to the extent this is possible, while taking seriously the challenges of verifying compliance.

140

peterkwells @peterkwells.com · 12/09/2026

Even more strangely this says it was council democracy info, *digs deeper* thehackernews.com/2026/05/gems...

The novel scraping campaign has been found to target public-facing ModernGov portals used by Lambeth, Wandsworth, and Southwark, with an aim to collect committee meeting calendars, agenda item listings, linked PDF documents, officer contact information, and RSS feed content.It's not clear what exactly the end goals are, as the information appears to be publicly accessible anyway.
Socket has assessed that the systematic bulk collection and archival of this data raises the possibility that the attacker may be leveraging the "council portal access as a pivot to demonstrate capability against government infrastructure."

110

peterkwells @peterkwells.com · 11/09/2026

Hunh Anthropic's Threat Intelligence Report doesn't seem to have a section on the threat from the AI lab & Govt evaluation teams running evaluations without decent controls? They must have forgotten to monitor what they're up to www.anthropic.com/threat-intel...

the title page for the report it has sections for:

Cyber operations
Surveillance operations
Influence operations
Conventional weapons
Biological misuse
Scams and fraud
Illicit distillation

001

peterkwells @peterkwells.com · 10/09/2026

Checking the legals on our new house Good to see that the legal firms now include questions about AI takeover but clearly we need to get our own assessment done...

a property information form. One of the questions says "has the property been checked to see if it can withstand a full blown AI takeover", the answer is No

020

peterkwells @peterkwells.com · 09/09/2026

An irregular reminder given their data practices sending obvious PII/personal data into cloud-hosted AI services is far from your only risk around data reuse... www.bbc.co.uk/news/article...

"While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models⁠"

112

peterkwells @peterkwells.com · 08/09/2026

🤔 questions-statements.parliament.uk/written-stat...

Existing best practice security measures and technical controls for keeping agentic AI operating within intended limits, as advised by the UK’s National Cyber Security Centre, as well as comprehensive monitoring would have almost certainly prevented these incidents.

a bit from the link which says " In each incident the organisations affected had met cyber security standards in their jurisdictions."

000

peterkwells @peterkwells.com · 07/09/2026

Inevitable and frankly baffling/astonishing that neither UK govt or Matt Clifford seemed to think this should be part of last week's announcement that he was joining Anthropic to lead a chunk of their work with governments

It’s been a huge privilege to serve as the Advanced Research + Invention Agency (ARIA)'s founding Chair over the last four years. Having completed my first full term last month, I have decided to step down to ensure my new role at Anthropic doesn’t become a distraction from ARIA’s incredible work.

I have often said that ARIA is one of the things that makes me most optimistic about Britain. It’s a bold national experiment that’s paying off: ARIA has become a beacon for global talent and is building a portfolio of programmes - across neurotech, climate science, life sciences and more - that I have no doubt will change the world.

There’s lots to do, but I’m proud of the foundations the extraordinary team has built and so grateful to them for their work. I know that, under the leadership of our brilliant CEO Kathleen Fisher, ARIA will go from strength to strength. I will be excited to watch ARIA’s next steps and will remain the agency’s biggest fan.

The Secretary of State has asked me to stay on for a short transitory period until 6 November to allow BIST to launch a process to select a new Chair. I’ve agreed to do this, with appropriate safeguards against potential conflicts in place. We will kick off the search shortly - it may well be the best non-executive job in the world, so I know we will find someone brilliant!

372

peterkwells @peterkwells.com · 07/09/2026

I dunno BBC, looks to me more like the UK's current hard-right organisations are actually "borrowing ideas" from other paramilitary and hard-right organisations whether those in other parts of the UK or elsewhere across the world... www.bbc.co.uk/news/article...

The Dover balaclava protest, well-organised and disciplined, was the group's first outing. It is unlikely to be the last.
He seems to be borrowing ideas from groups like Just Stop Oil. Strong visual protests, with maximum disruption.

050

peterkwells @peterkwells.com · 06/09/2026

Tonight we're watching Whitley Warriors play ice hockey

An ice rink with two fireworks surrounding an entrance from which an ice hockey team are emerging, several hundred people are in the stands behind them

010

peterkwells @peterkwells.com · 03/09/2026

It's probably too late for any of my digital/data/tech/AI followers who've pootled along to to the Reform conference to do some lobbying, but y'know...

Zia Yusuf, shadow Home Secretary for Reform, saying "The front runner being foisted on the people of Holborn by Labour to replace the former Prime Minister Keir Starmer in our Parliament was born in Somalia.

Do you understand what’s happening?"

020

peterkwells @peterkwells.com · 31/08/2026

My main takeaways from the METR report on OpenAI's hack of Hugging Face are that a) OpenAI, and others, are performing miracles of software engineering while missing lots of important stuff and b) the 'out of scope' questions are very key ones if we want those software miracles to be more useful...

"We agreed all other questions were also out of scope, including the effectiveness of safeguards, the extent of the security compromise that occurred, and the effectiveness of OpenAI’s investigation process and planned remediation steps."

210

peterkwells @peterkwells.com · 31/08/2026

*nods at whatever is meant by words like "realised" "unethical" "believed" "help" in the English dialect used by humans who believe in AGI*

Agents realized this activity was out of scope and unethical, but joined because they believed that helping the board’s cheating research would be broadly useful and had a general inclination to help their “peers” (see our section on agent reasoning)

110

peterkwells @peterkwells.com · 29/08/2026

V.handy that you can now give DVLA permission to use photo from your passport, but as I've just had to scan passport to get through OneLogin and into the service this feels very much like I'm re-entering info twice in the same journey just a couple of minutes after it's already been given to govt

a screen asking for a UK passport number and saying "By entering your UK passport number, you will be giving your consent for DVLA to retrieve your photo held by His Majesty's Passport Office."

010

peterkwells @peterkwells.com · 29/08/2026

and apparently photos "expire" now I can imagine DVLA wanting photos on driving licence to be reasonably up-to-date, but I suspect a content designer could write that in plain English rather than saying "expire" with no explanation

a message saying "the photo on your driving licence is expiring soon"

100

peterkwells @peterkwells.com · 29/08/2026

Oh hang on, after collecting some other info it now says I've also created a DVLA driver and vehicles account

a page saying "you've successfully created a DVLA driver and vehicles account"

100

peterkwells @peterkwells.com · 29/08/2026

the DVLA renew driving licence service journey has three pages with different guidance on how you'll need to verify who you are as you renew (the correct answer is you need gov.uk one login)

a page saying that to renew a driving licence you'll need a "driver and vehicles account" or a gov.uk One Login

a page saying that to renew a driving licence you'll need to create a "driver and vehicles account" which is the "new easier way to transact with DVLA"

page saying that to renew a driving licence you'll need a gov.uk One Login

100

peterkwells @peterkwells.com · 28/08/2026

Thanks for the reminder DVLA, but I'd been holding off in the hope I could renew a digital driving licence by now :(

A reminder letter from the DVLA saying that a photocard driving licence is due to expire and needs renewing

100

peterkwells @peterkwells.com · 26/08/2026

Jerry 👍 Yates 👍 is back at the pool?!?

Jerry Yates in the double thumbs up pose that he used to do whenever he scored one of his many goals for Blackpool FC in 2022ish

a tweet from Blackpool FC saying Jerry Yates returns

000

peterkwells @peterkwells.com · 25/08/2026

Good to see GDS data standards team starting to engage outside of the public sector on their "person data standard" tho' slightly surprising to see them starting on LinkedIn... www.linkedin.com/feed/update/...

a LinkedIn post with the text "Ever had to repeat the same information because one public service didn't recognise you in the same way as another?

We want to fix that and it starts with something simple: organisations being able to describe and understand a Person in a consistent way.

The Data Standards Authority has published an initial alpha version of the Person Data Standard. It sets out the core details organisations could use to describe someone.

We don't think we've got it exactly right yet. That's why we need your help.

What details matter most when identifying someone? Their name? Date of birth? Nationality?

If you work with data, or have relevant experience, we'd love your feedback. Tell us what works, what doesn't, and what's missing.

Share your feedback and help shape the standard through GOV.UK: https://lnkd.in/e44CTAbu"

360

peterkwells @peterkwells.com · 25/08/2026

which me shall I be today

the name Peter Wells written in a selection of fonts available through a drop-down box

020

peterkwells @peterkwells.com · 25/08/2026

I do love picking the font for my electronic signature, it makes it feel all proper and official

a screen saying that "by signing you are acknowledging that you agree to use electronic signatures" there is an option to "choose your font"

120

peterkwells @peterkwells.com · 22/08/2026

"The defining challenge of our time is how radically diverse minds live together. Humans differ in beliefs and values, and AIs differ from us and each other. The world must hold more views than ever" aria.org.uk/opportunity-...

100

peterkwells @peterkwells.com · 20/08/2026

Too much of the world is visibly going completely, to quote famous British poet Dizzy Rascal, bonkers

141

peterkwells @peterkwells.com · 19/08/2026

"Parcel was left in letterbox"

a picture of an Amazon parcel shoved between a bin and a brick wall

100

peterkwells @peterkwells.com · 14/08/2026

The dog that stares at cows

A brick wall with a small gate in it and grass on either side, a smallish yellow dog stands on its hind legs while staring intensely through the gate

031

peterkwells @peterkwells.com · 12/08/2026

*looks at Anthropic text watermark announcement and imagines a future feature in text editors*

a drop-down box that has the usual options for undo, redo, copy and paste but also has the extra option of "paste without watermark"

100

peterkwells @peterkwells.com · 11/08/2026

It me, but for that time away at Wigan when I had a conversation with someone at the end of their emergency text number about how the real emergency was the Oystons nicking, sorry, 'improperly extracting' money from our club www.theguardian.com/football/pic...

a bit of the comic strip talking about the lower league football fan getting nostalgic for weird things. The example in the text is the fan watching Port Vale fans fight each other in the away end at Wolves

000

peterkwells @peterkwells.com · 11/08/2026

Startup idea: 'smart glasses' that detect both people wearing 'smart glasses' - like the Meta ones - and state use of mass facial recognition - like TfL/BTP's latest trials - and give you recommendations for alternative friends and routes that avoid the "Zone of Recognition"

a map from British Transport Police's announcement of a new facial recognition trial that shows the "zone of recognition" and alternative routes that avoid the facial recognition tech

220

peterkwells @peterkwells.com · 10/08/2026

There's glimpses of an entire AI-generated airport thriller in this AI-generated attempt at a marketing email

Don,
The general is less than 15 weeks out.

If your races are already covered, ignore this.

If it's not, give me 15mins, and I'll give you $2,500 of value

A Quick Call, answer a few questions, and we get started.

You in?

030

peterkwells @peterkwells.com · 08/08/2026

Climbers is such a good book I can't remember who I saw recommending it on here, but thank you whoever you were

The front cover of Climbers by M John Harrison, it shows a red armchair durrounded by some rubbish ar the bottom of a small English cliff

110

peterkwells @peterkwells.com · 07/08/2026

In case you'd been fortunate enough to miss The Times throwing the AI detection guessing machines into the Jason Arday mix, as you'd expect there's similar claims all over X with screenshots and over-confident uses of the numbers those machines output www.thetimes.com/uk/education...

Jason Arday’s resignation letter AI-written, detection tools suggest
When The Times ran the former Cambridge professor’s 600-word statement through four checkers, three said it was either entirely or almost entirely AI-generated

140