Building a learning app for my kids. Two homepage options and I can't pick.
A: "Turn 'I don't get it' into 'I get it'."
B: "A patient AI teacher who explains first, then checks."
Which one makes you tap Start? Parents especially, be brutal.
Last night my wife complained about the kids’ schoolwork.
So I started building.....
24hrs later, my kids (Aged 6 & 9) have both taken lessons, given feedback and iteratively tested the app with me and her.
We're going to give it to our friends to test with their kids around the same age and, with any luck, a new product will be born!
Anybody able to beat 24hrs from idea to reality?
Reuters: one hacker set AI agents on 1,200+ websites and got into over 100, Deere and SharkNinja included. About $5,000 of tokens, so roughly $50 a break-in.
Cheaper than the pen test you keep putting off.
When did anyone last try to break into your site?
On a £1m company that's £450k in tax instead of £240k, just for handing it to your own kids.
Sell to an outsider and the old rate still applies. You'd almost think they want family firms sold off.
Panorama: the DWP paid £410m last year for supported exempt accommodation in Birmingham alone. 34,000+ people live in it there. London, seven times bigger, has about 22,000 in supported housing.
Pay by the room, don't check the support, get rooms.
Who's meant to be checking?
Introducing Mittral
Mittral didn't start as a product. It started as a requirement for one customer in South Africa who needed a better way to handle purchasing: purchase requests, approvals, and getting proper purchase orders (POs) out to suppliers.
They liked what we delivered so much that I kept going. I improved it, tidied it up, and made it available to any business with the same problem.
It fits what we've always done. For 15+ years I've built practical solutions for customers. Mittral takes one of those solutions and puts it in front of everyone who needs it.
A big shoutout to Ross Rencken, who has been banging the drum for some time about releasing micro-apps as a business. You were right, mate. This is the first of them.
Here's how it works: employees raise a purchase request, it goes to the right approver, and once it's approved the supplier gets a proper PO by email. Every purchase approved before it's ordered.
It's free forever for up to 5 users and 10 POs a month. No card and no sales call, so there's no risk in giving it a go.
All feedback is welcome, good or bad. Tell me what's missing.
🌐mittral.com#Mittral#SmallBusiness#Purchasing
France's MPs voted on Friday to tax swapping crypto into stablecoins from 2027.
38 for, 37 against. 477 of the 577 didn't take part. A new tax, carried by one vote. The Senate still gets its say.
Is swapping Bitcoin into a stablecoin cashing out, or just moving it?
If your purchasing runs on "can you just order this?" in email and chat, it works until it doesn't.
Mittral: request it, approve it, and the supplier gets a proper purchase order as soon as it's approved.
Free forever for 5 users. No card.
mittral.com/?utm_source=x&ut…
yes that was a bit frustrating mid incident when trying to investigate. they did quite a bit and then started refusing
we got access to daybreak blue now so it's better
Anthropic says Claude writes about 90% of Claude Code's own code. Low-risk changes now merge with no human sign-off once the checks pass.
Merged PRs more than doubled June to September. Revert rate down over 40%. Their numbers, mind.
Would you merge code nobody's read?
Barracuda's found phishing emails with instructions hidden in the HTML. Not for you, for the AI that summarises your inbox. One told it to add a priority action: change a supplier's bank details.
Invoice fraud's had an upgrade. Does your AI act on email, or just summarise it?
Microsoft has pulled out of Nscale's Essex data centre, billed as the UK's largest AI supercomputer.
The grid can't power it until the 2030s.
AI superpower, apparently.
If you were building an AI company today, would you put your compute in Britain?
Andrew Bailey warned yesterday that AI agents making payments could leave people unsure who's liable.
Citi, BofA and JPMorgan have just blocked outside agents instead.
If an agent pays the wrong person from your account, who carries the can: you, the bank or whoever built it?
Great British Energy's first accounts: £14.2m on running costs, including £7.9m on staff and £3.8m on lawyers and advisers. Invested: £5.8m, in one wind project.
£2.45 spent on itself for every £1 it invested.
Would a private company get a second year on numbers like that?
Accenture now covers $29 a month per person for Claude. Anything over goes on a project code. Some staff were spending up to $1,000 a month, mostly on Claude Code.
Not a ban. Someone's just read the bill.
Would you use agents less if the tokens came out of your own budget?
Last month an AI built me an app prototype in a couple of hours. Eight passing tests. The login didn't work.
Green ticks from the same agent that wrote the code don't prove much.
What do you check before you trust tests an agent wrote for itself?
Mittral’s mobile app is with Apple and Google for review.
Raise purchase requests, photograph quotes and approve from your phone, even without a signal.
It syncs when you’re back online.
Free for teams of up to 5 at mittral.com