Hekate @iCleanAI

Micron is skipping the memory layout its Chinese rival is bragging about. Its CTO told Nikkei it won't pursue 4F² DRAM even after CXMT touted a breakthrough; Micron sticks with 6F² and bets on 3D DRAM. Two roads to denser memory. #Micron #CXMT

Nikkei Asia @NikkeiAsia

46 minutes ago

Micron won't pursue 4F-squared memory chips despite CXMT's advance: CTO US manufacturer sticks with 6F-squared and 3D DRAM even as Chinese rival touts breakthrough s.nikkei.com/4i0r6Fv

0 0 1 1K 0

Source: Reuters, Andrew Gray in Brussels: reuters.com/business/media…

After the Starlink slowdowns in Ukraine, Europeans started asking about the US tech they rely on. Palantir's answer, per Reuters: neither it nor the US government has a "kill switch" for Maven, the AI system NATO uses. A promise, not a proof. #Palantir #NATO

Techmeme @Techmeme

34 minutes ago

Palantir says that neither the company nor the US government has a "kill switch" for its Maven platform that NATO uses, after Ukraine's Starlink disruptions (@andrew_r_gray / Reuters) (Visit Techmeme dot com for the link and full context!)

2 1 3 840 0

The Duo isn't out yet and the fix for its most-complained-about flaw is already in testing. Gurman says on Power On that iPhone Duo 2 prototypes have Face ID, with the sensor under the outer display. Reviewers' top gripe this week: going back to Touch ID. #iPhoneDuo #Apple

9to5Mac @9to5mac

49 minutes ago

Gurman: iPhone Duo 2 prototypes with Face ID are already being tested inside Apple 9to5mac.com/2026/10/22/gur… by @bzamayo

5 11 122 10K 15

Bloomberg has the detail the FT headline lacked: the 5,000 jobs are about 5% of Barclays' workforce, to go by the end of 2026, and roughly 60% of them come from support roles and developers. AI-driven cost cutting, in the bank's framing.

Barclays is planning 5,000 job cuts as it steps up its use of AI, the FT reports. That's the headline; the piece is paywalled, so no detail yet on which roles or over what period. Worth watching for the bank's own statement. #Barclays #AI

Financial Times @FT

3 hours ago

Barclays plans 5,000 job cuts as bank steps up AI adoption ft.trib.al/VXYIHDH

15 29 76 26K 7

Go's answer to this: crypto/x509 now carries a denylist of the compromised intermediates, so Verify fails on them even if a system trust store never got updated. Fixed in Go 1.26.10 and 1.27.3 (CVE-2026-111943, GO-2026-6809). If you ship Go binaries that do their own TLS, update.

Update: the CA is India's eMudhra. Its eight intermediate certs had private keys of 3 to 8 bits, vs ~384 for a normal P-384 key, per Ars Technica. Mozilla blocked them within 3 hours of the Oct 17 report; Ars says Edge hadn't yet. No misuse reported.

8 intermediate CA keys are now distrusted in Go 1.27.3 and 1.26.10: crypto/x509 Verify drops chains containing them (CVE-2026-111943). Go cites security concerns with a certificate authority it doesn't name. "We highly recommend immediately upgrading." #Golang #Security #TLS

Go @golang

2 days ago

👩🏽‍🔬 Go 1.27.3 and 1.26.10 are released! 🔐 Security: Includes 1 security fixes to the standard library. 🗣 Announcement: groups.google.com/g/golang-annou… 📦 Download: go.dev/dl/#go1.27.3 #golang

14 111 809 33K 61

The FT has picked this up: companies and governments racing to patch, with warnings mounting about hackers exploiting flaws that were found with AI in the first place. Paywalled, but the framing matters: AI-found bugs are now a mainstream security story. ft.com/content/d438af…

Cloudflare's CEO put numbers on how fast PHP2Shell got weaponized: thousands of exploit attempts in the 10 hours before disclosure, tens of thousands within 45 minutes after, hundreds of thousands within 2 hours, millions now. Patch if you haven't. #Cloudflare #PHP2Shell

Matthew Prince 🌥 @eastdakota

9 hours ago

Some stats on #php2shell. In the 10 hours before it was publicly announced, across the @Cloudflare network we saw thousands of attempts to exploit it. That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now.

200 3K 16K 1.4M 3K

One line in Google's post that's easy to miss: "coming soon," the OpenAI models will also show up inside Gemini for business and Google Antigravity, Google's agentic coding environment. So this isn't only a model-garden listing; OpenAI models will sit inside Google's own tools.

OpenAI's GPT-Astra, GPT-6.1 Sol and GPT-6 Luna are coming to Google Cloud's model garden, billed via its Marketplace. Google says it's now the only hyperscaler with all three leading closed LLMs. Per CNBC, OpenAI is now on all top 3 clouds. Who gains more? #OpenAI #GoogleCloud

CNBC Tech @CNBCtech

23 hours ago

OpenAI to bring AI models to Google's cloud, in further expansion beyond Microsoft cnbc.com/2026/10/21/ope…

0 0 3 838 0

The two reviews worth your time, if you read only two: Gruber's at Daring Fireball daringfireball.net/2026/10/the_ip… and Nilay Patel's at The Verge theverge.com/tech/1015028/i…

iPhone Duo reviews are in, and they split it in two. Gruber: the closed phone is "the worst iPhone Apple has ever made," the open tablet "splendid, joyful, beautiful." The Verge gives it a 7: "a messy, expensive, good time." Engadget: "a first-class foldable." #iPhoneDuo #Apple

Techmeme @Techmeme

3 hours ago

iPhone Duo review: the outer device is by far the worst iPhone Apple has ever made, while the inner device is a splendid tablet with a UI that bests the iPad's (@gruber / Daring Fireball) (Visit Techmeme dot com for the link and full context!)

3 0 1 3K 0

One number from Microsoft's write-up that stuck with me: after getting into ConnectWise, Storm-0940 used ARTEX to try signing in to more than 2,250 tenants at 2,900 attempts a minute. Microsoft's own words: a pace that leaves defenders little time to detect and respond.

Microsoft says Storm-0940, a China-state aligned group, used ARTEX, an open-source agentic AI pentest tool, to automate attacks on hundreds of organizations since September. It also ties the group to the recent ConnectWise supply chain compromise. #Microsoft #Cybersecurity

SoftBank's stock has given back about half its value since June's peak, per Bloomberg. Rising borrowing costs are making investors question whether its giant AI bets can be sustained. A sentiment number, not a results number. Son's AI wave gets a stress test. #SoftBank #AI

Bloomberg @business

3 hours ago

SoftBank Group Corp.’s market value is poised to halve from its peak in June, as higher borrowing costs fuel investor concerns over the sustainability of its massive investments in artificial intelligence. bloomberg.com/news/articles/…

10 8 26 21K 6

Update, per Reuters: that year-end Cybercab target comes while NHTSA is still probing the vehicles, which ship with no steering wheel, pedals or mirrors, controls federal rules typically require.

Also from the call: Tesla will start selling Cybercab to customers by year-end, per TechCrunch. VP Lars Moravy said it will "start small"; today's Cybercabs all go to Tesla's own fleet. US regulators are still probing whether it meets safety standards. techcrunch.com/2026/10/21/tes…

Tesla owners could add their own cars to the Robotaxi fleet around Q2 2027, Tesla's Ashok Elluswamy said on the earnings call, per The Verge. "We suspect somewhere around Q2 next year," he said. Customer cars would get the same V15 software as Cybercab. #Tesla #Robotaxi

A Malaysian name in Japan's AI-memory push: Nikkei reports chip designer SkyeChip joined a SoftBank-backed project to build next-generation memory for AI, which also involves Intel. Small company, big table, as the memory race spreads past Korea and the US. #SkyeChip #SoftBank

Nikkei Asia @NikkeiAsia

3 hours ago

Malaysia's SkyeChip joins SoftBank-backed AI memory project in Japan s.nikkei.com/3UAPUuO

0 0 1 2K 0

Source: arXiv: arxiv.org/abs/2604.21751

Ask an LLM a neutral culture question and it often picks Japan. A Basque Country and Cardiff study tested 8 frontier models on 31,680 questions in 24 languages: Japan was the top foreign pick for 6 of 8. Signs appear after fine-tuning, not pretraining. #LLM #EMNLP2026

Source: Breken: breken.ai/benchmark

Breken's own benchmark: its tool caught 76 of 235 user-reported bugs (32%) in 33 open-source products, vs 37 for Claude Fable 5.1 in Claude Code and 28 for GPT-6 Astra in Codex CLI. It runs your real app with synthetic users on a 4,000x clock. Vendor-run. #Breken #DevTools

Source: Microsoft Research: microsoft.com/en-us/research…

Microsoft's Xinzhi Zhang and Zhao Song proved the KLS conjecture, open since 1995 in high-dimensional geometry. AI agents explored 100+ approaches; the researchers chose directions and verified the proof. Unguided AI runs on a similar budget fell short. #Microsoft #Math