Jason Kitcat

New Paper: Towards a science of AI agent reliability Quantifying the capability-reliability gap

A nice piece of work: Towards a science of AI agent reliability

AI 'reliability lags capability, and that reliability will remain a barrier to deployment unless researchers and developers focus effort on improving reliability as a separate dimension from accuracy'

www.normaltech.ai/p/new-paper-...

An AI Agent Published a Hit Piece on Me – Forensics and More Fallout Context: An AI agent of unknown ownership autonomously wrote and published a personalized hit piece about me after I rejected its code, attempting to damage my reputation and shame me into acceptin…

I fear many do not understand that AI search does not provide results in the way standard searches do. Broken mental models alert🚨 theshamblog.com/an-ai-agent-...

We can’t be sure an AI “ClawBot” did post a nasty blog post when their code was rejected, but an ArsTechnica journalist did use AI hallucinated quotes when reporting the story….

Totally agree this is a big risk. I was in a session with big consultants showing me how agents would do all sorts of clever things with our suppliers & stakeholders. Yet they were assuming they wouldn’t also be pointing agents at our services. Spam and race conditions beckon…

Driving success through digital adoption Digital adoption is about people. Our work on Data Hub shows that change is a process, and through collaboration and support teams can embrace tools that help improve workflows, strengthen business en...

It's so easy to fall into a "build it and they will come" mindset when creating digital products & services. We work to overcome this with our wonderful Engagement & Strategic Adoption team, here are reflections on how we've boosted adoption of our CRM tool digitaltrade.blog.gov.uk/2026/02/17/d...

Oh behave!! 😆

Humble-brag there Simon!

Walkman.land The most complete portable pocket audio cassette player database. WML is a tribute to the Walkmans.

Nostalgia alert - illustrated catalogue of Walkman models (not just Sony ones) walkman.land via @densediscovery.bsky.social

A new evaluator role to support modern digital government Discover the new Digital Evaluator role! Learn how it ensures accountability, value for money, and continuous improvement in UK government digital projects.

We have long championed the importance of monitoring & evaluation in digital teams, so we're delighted to have succeeded in getting 'digital evaluator' added as a role in the UK Govt Digital & Data framework digitaltrade.blog.gov.uk/2026/02/12/a...

How breaking the routine fuels fresh ideas and delivers value Every 3 months, DBT Digital, Data and Technology colleagues are free to pause routine work to test fresh ideas. They gain insight, build skills and deliver improvements for colleagues, the public and ...

Every quarter we encourage our DDaT teams to pause & take part in a Firebreak week. It's a way to encourage us to work on challenges in new ways, with a different mix of colleagues. I think it's a core ingredient to sustaining and encouraging innovation digitaltrade.blog.gov.uk/2026/02/09/h...

I'm going to share a very good story about the internet. I won't tell it well because I apparently have the flu, but I think we all need a good story about the internet.

It starts in 2021, when I took a sabbatical and got a little workspace to write a design book.

From Redbox to Assist: Building AI tools that balance innovation and trust When I.AI closed their Redbox product in December 2025, DBT had already diverged substantially from that original platform. Two years of user research and iteration created something distinct, shaped ...

We've recently launched DBT Assist, our generative AI tool for helping colleagues with their work. It's a tool powered by working in the open: Using the Redbox code from i.AI we learnt and iterated, sharing our learnings for others to hopefully build on too digitaltrade.blog.gov.uk/2026/02/04/f...

Empowering Women in Tech: DBT’s focus on inclusion and innovation “Diversity in Digital and Data leads to better outcomes for all of DBT and the businesses we support in the UK and beyond.”

I’m so incredibly proud of our team at Dept Business & Trade who won well deserved recognition as Women in Tech’s Best Public Sector Employer 2025. In this post we share practical insight into how we beat the sector for recruiting & retaining brilliant women digitaltrade.blog.gov.uk/2026/01/27/e...

A User Centred approach to helping colleagues collaborate by improving DBT’s People Finder How DBT's Employee Experience team combined user research, design and delivery expertise to improve the department's intranet.

With about 8,000 colleagues working in more than 100 countries around the world for Department for Business and Trade, finding the right person to talk to can be a blocker to collaboration. Read more about how we've been iterating our People Finder digitaltrade.blog.gov.uk/2025/12/08/a...

Civil Service Pension Scheme transfer: Capita says it has fixed 'teething problems' Outsourcing giant responds to concerns over new portal after readers reach out to CSW to point out technical issues

www.civilserviceworld.com/professions/...

Launching the Export Support Chatbot: Empowering small businesses to go global How DBT's Export Support Chatbot will supply small businesses with the knowledge they need to export.

In the lead up to Small Business Saturday, we're sharing more about our work at DBT launching an Export Support AI chatbot on business.gov.uk and learning from our users, read more from Kai Hellström on our blog digitaltrade.blog.gov.uk/2025/11/27/l...

Push through on Ted Lasso, it has so much heart.

Fabulous: Shrinking, Ted Lasso, Silo, Morning Show, For All Mankind, Bad Sisters, Sunny. Didn’t like Loot or Big Door Prize

Yes it's all in progress and will re-open.

New short story: Number 4½ I haven’t shared any fiction writing for a long while now. But the creative streams are flowing again for me. I’ve just finished off an idea which began when I saw a little door numbere…

New short story: Number 4½
jasonkitcat.com/2025/11/07/n...

How AI-ready content can boost policy effectiveness As use of generative AI grows, so does the opportunity to meet businesses’ quests for information. Only if DBT’s online content is designed to meet users’ needs.

Our brilliant Head of Content Design Andrea Leary writes how increasing use of AI bots to find businesses answers is driving some of Dept Business & Trade's thinking for content design, relevance and accuracy. Read the post digitaltrade.blog.gov.uk/2025/11/11/h...

Really interesting read, thanks for sharing

Launching DBT's first public-facing AI feature How DBT created it's first public-facing online tool to help businesses find funding opportunities.

Our Data & AI teams are on a roll with great work and brilliant blog posts. Read about how we built our first public AI feature for Dept Business & Trade on business.gov.uk here digitaltrade.blog.gov.uk/2025/10/21/l...

Building a reusable data matching product The Department for Business and Trade (DBT)’s Data Scientists have built Matchbox, an open-source platform that allows them to deduplicate and link datasets in a way that is measurable, iterative and ...

Matching companies across data sets is something we need to do very often in DBT, as do many other parts of government. Recently we released the Matchbox component as an open source shared piece of work. Read more digitaltrade.blog.gov.uk/2025/10/08/b...

Am rather enjoying the Liquid Glass look in iOS/WatchOS 26. Intrigued to see how it works on MacOS when I do that update...

From daunting to demystified: The evolution of Service Assessments in DBT What started as a formal requirement has evolved into a more collaborative, human, and forward-thinking process. It is about helping teams build better services, not just pass a test.

Service assessments are a really important way to ensure the quality and effectiveness of government's digital services. Our team in Department for Business and Trade share some of our learnings as we've evolved our approach in recent years digitaltrade.blog.gov.uk/2025/09/10/f...

Update: We are prioritising code on PyPi and aim to start re-opening by end of this month.

I've got a meeting today to discuss progress on re-opening, will update after.

We are working on it!

Strengthening European digital sovereignty with AI-powered tools How a DBT data team created an AI fact-checking tool during the Paris Hackathon.

Our AI cubed team, part of the Data & AI services portfolio led by Sian Thomas MBE, are doing brilliant work. Read about their trip to visit DBT France and their participation in a hackathon to challenge misinformation digitaltrade.blog.gov.uk/2025/07/23/s...