TORUS
Hello Mate,
For UK founders, operators and professionals who want a fast, useful view of what changed in AI in the last 24 hours.
Let’s dive in.
Today’s AI Signal
Scottish rescue AI wins KPMG's UK technology title
Edinburgh startup Zelim has won the 2026 UK edition of KPMG's Tech Innovator competition. Its software uses AI-enabled cameras to detect a person falling overboard and guide rescue teams during the search.
This is applied AI with a clear test: detect faster, find the person and reduce the time spent in dangerous water. Zelim will represent the UK in KPMG's global final in Lisbon in November.
NHS discharge software shows no noticeable improvement
A Health Foundation study found no noticeable improvement in hospital discharge performance at NHS trusts using Palantir's Federated Data Platform tool. NHS England had previously cited a 15% reduction in long-stay discharge delays.
The two sides measured the problem differently, but the broader lesson is straightforward. Public-sector AI needs agreed success measures before rollout, not after claims have entered the press release. An independent evaluation is under way, though the final report is not expected until 2029.
Nvidia may guarantee $250bn for OpenAI infrastructure
Nvidia is in talks to provide roughly $250 billion in financing guarantees for an OpenAI data-centre project in Ohio, according to a Wall Street Journal report cited by Reuters. The planned site would deliver 10 gigawatts of capacity and cost more than $500 billion.
The figure is not a completed deal, but it shows where the AI contest is heading. Access to chips is only part of the problem. Power, land, debt and the ability to reassure lenders are becoming part of the product stack. UK firms do not need to copy this scale, but they should price infrastructure dependence into every serious AI plan.
Global AI headlines
TODAY’S LESSON
Audit one workflow before buying another AI tool
The fastest way to waste money on AI is to start with a product rather than a process.
Use this prompt when a recurring workflow feels slow, manual or error-prone. It forces the model to separate work that AI can assist from work that should remain under human control. It also gives you a small test instead of a company-wide project.
Act as an operations analyst helping a UK small business.
I will describe one recurring workflow. Do not recommend tools until you understand the work.
First, rewrite the workflow as numbered steps and flag any missing information.
For each step, classify it as:
- Human only
- AI assisted
- Suitable for automation
- Remove or simplify
Then create a table with:
1. The current step
2. Time spent
3. Main source of delay or error
4. What AI could do
5. What a person must check
6. Data or system access required
7. The worst plausible failure
Finish with a 14-day test plan. Include the baseline to record, the smallest version to test, who approves outputs and the measures that would justify continuing.
Workflow:
[PASTE YOUR WORKFLOW HERE]Why it works: the prompt starts with process mapping, makes human checks explicit and asks for evidence before a wider rollout. That is a better sequence than buying software and looking for somewhere to put it.
UK AI tools and offers
Tool: Teable 3.0
Teable combines a spreadsheet-style interface with databases, automation and AI. The latest release can process groups of Excel or CSV files, work with PDFs of up to 100 pages, build internal apps and run recurring operational workflows.
A sensible first use is a small sales or service tracker where AI summarises records, drafts follow-ups and flags overdue actions. Keep the source data visible and require approval before messages are sent.
Pricing: free plan available. Pro costs $10 per seat per month when billed yearly, or $12 monthly.
Try it → https://teable.ai/
Kimi K3
What it does: Reads large collections of documents, works with images and office files, produces research and slide drafts, and supports coding tasks.
Best for: Comparing lengthy reports, reviewing document collections and creating an initial research brief.
Pricing: Kimi offers a free membership level. API access is usage-based, with K3 priced at roughly $3 per million input tokens and $15 per million output tokens.
Torus verdict: Worth testing on non-confidential work. Do not move sensitive company data into it until your security and privacy checks are complete.
Offer: up to £800,0v00 for advanced connectivity research
UK academic institutions can apply for a share of up to £800,000 through the ACTASAP pilot. The programme, funded by DSIT and delivered by Innovate UK, covers advanced connectivity ideas and may suit research involving AI-native or autonomous network systems.
The deadline is 11:00 am on Tuesday, 28 July 2026, so this is only useful for teams already prepared to apply. Businesses should check the collaboration and eligibility rules before spending time on it.
Free Offer: UK AI Skill Boost
The UK government's AI Skills Hub offers free workplace AI courses for adults across the country. Learners can create a training profile, follow a recommended path and earn a government-backed AI foundations badge through selected courses.
This is a sensible starting point for small businesses that want to train staff without buying a company-wide course.
Tool: Google AI Studio
Google AI Studio lets users test Gemini models and build simple prototypes in the browser. For a UK founder, the practical use is testing prompts, checking model behaviour and shaping a simple AI workflow before paying a developer. Google says AI Studio usage is free in available regions, with model and API limits. Read more.
UK AI Job Opportunities
AI Engineer, Sequence - London. Salary listed at £115,000 to £130,000.
Senior AI Engineer, AI Transformation, Multiverse - London, full-time, hybrid.
AI Engineer, Funding Circle - London, hybrid, two office days each week.
AI Engineer, Writer - UK role with a London office hub.
Staff AI Engineer, AI Transformation, Multiverse - London-focused role working across human and AI systems.
Stay in the loop - TORUS tracks UK and European AI so you do not have to.
You are reading TORUS because you want to stay ahead of where AI is going, not just where it is today. If you find this useful, forward it to one UK founder, operator or policy‑maker who should be thinking about AI infrastructure, tools and skills right now.
