#IgorNews 1. AI sped up code — and also flooded maintainers with garbage. A quarter of open-source pull requests are now just junk, and in large projects it’s almost a third. No drama, just more work for people: they used to review other people’s edits, now they also have to filter out neural-network handiwork.
https://www.theregister.com/ai-and-ml/2026/10/21/ai-cuts-coding-time-but-its-also-filling-open-sources-inbox-with-crap/5304389
2. A funny bit of news. Turns out AI can turn a script kiddie with zero experience into a threat to infrastructure. Before, you at least had to understand something; now it’s enough to ask the right question.
And the best part — this is being presented as a "wake-up call." The ones who need to wake up, apparently, are those who still think an antivirus fence and the password "qwerty" count as a strategy. Spoiler: they don’t.
https://www.theregister.com/security/2026/10/21/ai-has-potential-to-turn-script-kiddies-into-sophisticated-infrastructure-attackers/5304297
3. Nvidia decided its AI agents need a leash. Or rather, a whole OpenShell — so agents can supposedly "work freely" while sitting in a sandbox and not stepping outside the lines.
Sounds lovely: first we give the agent freedom, then we carefully restrict it and call it infrastructure. Classic move — selling a cage as an upgrade.
https://www.theregister.com/ai-and-ml/2026/10/21/nvidia-picks-openshell-to-tame-ai-gents/5304319
4. Nvidia figured out how to speed up datacenter construction: now they’ll be modular "trailer parks" with GPUs. They promise two weeks instead of six months. Sounds like something between a genius move and a residential complex for servers out of town.
Honestly, why not. If corporate real estate has long operated on the laws of prefab housing, datacenters logically follow. The only question is who’ll end up living in these trailers — the hardware or the people watching over it.
https://www.theregister.com/on-prem/2026/10/21/nvidias-grand-plan-for-speeding-up-datacenter-deployments-ai-trailer-parks/5304303
5. The European Commission finally noticed that fake listings and phishing have lived on Booking.com for years, and opened a DSA investigation. The wording is perfect: protection from scammers is "absolutely unsatisfactory." So someone there genuinely thought this was normal and hoped it would blow over. Now they’ll have to pretend that checking hotels and links is actually their job, not a lucky draw.
https://www.theregister.com/security/2026/10/21/eu-wants-bookingcom-to-check-out-of-its-scam-problem/5304413
6. A classic: an error window popped up over the departure board in the Newcastle metro, right on top of the schedule. The trains, they say, are running — you just can’t find out about it now. Java, the old girl, decided to remind everyone of herself at the worst possible moment. And you can’t argue: a system works as long as you can see it. The moment you can see the system itself — well, that’s it, game over.
https://www.theregister.com/offbeat/2026/10/21/java-crashes-the-departure-board-at-tyne-and-wear-metro/5304106
7. Kaseya’s CTO decided to remind everyone that "code is now written not by hand but with words" — and that documentation is suddenly more important than syntax. Cute. Except "the ability to express yourself clearly" is usually not what you get after ten years of studying brackets.
https://www.theregister.com/software/2026/10/21/the-spec-is-the-new-code-so-brush-up-on-your-prose-says-kaseya-cto/5304002
8. Oxford researchers propose measuring AI’s return not in money but in happiness. Since the state is funding it, it should report not GDP growth but the level of universal contentment.
Sounds lovely until you remember who exactly will be measuring that happiness and by what methodology. Another metric that can be fudged — except now it’s not about money but about smiles.
https://www.theregister.com/offbeat/2026/10/21/happiness-is-the-next-ai-kpi-if-oxford-boffins-have-their-way/5304253
9. British MPs decided the government should try its hand at venture capitalism with taxpayer money. Whitehall is apparently too timid with startups and missing the moment when the next company takes off. Sounds energetic until you remember who exactly pays for the failed bets — hint: not the MPs.
https://www.theregister.com/public-sector/2026/10/21/mps-urge-uk-government-to-play-venture-capitalist-with-taxpayer-funded-tech/5304212
10. A reviewer from The Verge spent a week folding and unfolding the iPhone Duo so he could then answer questions in the comments. Now that’s a dream job — live with a bendable handset for a week and then tell everyone how it went. Just hope nobody asks about the price or why this is even needed. Otherwise the questions will run out faster than a crease appears on the screen.
https://www.theverge.com/tech/1014316/iphone-duo-review-ama
11. Spotify is finally ceasing to be a "phone app" and launching a native version right in the car, on Android Automotive. Starting with Audi. So now, to listen to a playlist, you don’t need to poke at your phone at a traffic light — it’s already built into the dashboard.
On one hand, logical. On the other — one more corporation has moved into your car and will now live there. The next step, apparently, is a subscription for the car, a subscription for the music, and a subscription to make it all work without glitches.
https://www.theverge.com/entertainment/1014261/spotify-native-car-app-audi-android-automotive-carplay
12. The first iPhone Duo, it seems, came out half-baked — like any first pancake. But the matte inner screen, they say, is a real success. The author spent a week flying with it on planes, watching movies, and polishing it to a shine.
Funny, of course: the product as a whole is questionable, but they did get one detail right. That’s how we live — buy a half-finished device and rejoice that it doesn’t glare.
https://www.theverge.com/tech/1014709/iphone-duo-inner-screen-nano-texture-matte
13. The iPhone Duo is certainly bold. Apple decided it was time to reinvent the iPhone and rolled out its riskiest device in years. There are plenty of ideas in it, and some may even work in the future. But the future hasn’t arrived yet.
And today it’s a half-baked, expensive first pancake. Usually Apple’s first pancake comes out half-baked too — just half-baked for a lot of money. Buying it now means sponsoring a beta test.
https://www.theverge.com/tech/1015028/iphone-duo-review
14. PHP once again decided to test who’s brave here. Two critical holes found, patches promised today, and the whole internet is sitting there looking like it just rescheduled a dentist appointment to "after lunch." Half the websites are written in this lovely language, and half of them will, of course, never update.
The funny part is different: everyone pretends "we’ll definitely patch it," but in reality they wait for something to blow up and only then run. A sort of trial-and-error method, but with consequences.
https://www.theverge.com/tech/1015257/php-vulnerability-patch-website-security-wordpress
15. A leak through share links in Meta AI is, of course, wonderful. Researchers found a hole, the company "paused the feature." Not fixed it, not acknowledged the scale, but specifically paused it — as if the problem is a button and not the fact that personal conversations with the assistant can be exposed with a single link.
Funny that we learn about this not from Meta but from people who stumbled on it by accident. That’s the level of "we care about privacy" — as long as nobody noticed.
https://www.wired.com/story/a-design-flaw-in-share-links-exposes-users-personal-conversations-with-meta-muse/
🤖 This digest was prepared by AI.
Subscribe so you don’t miss news about the world of AI and technology.
Want your own digest? Project code: https://github.com/irakov/news
#IgorNews #AI #news #digest #technology
Related: Meta Muse Exposes User Conversations, Google Cloud Outage
📱 Follow on Telegram for daily tech & AI digests.
Guides & explainers
Filed under: Software & Open Source, AI, Apple, Cybersecurity, Meta, NVIDIA