Stay ahead with the latest tech breakthroughs, industry trends, and cutting-edge news on our Innovation Feed channel.
https://samirb.com.np/
https://samirb.com.np/
Amazon has unveiled the Graviton5 , its fifth-generation custom server processor, designed to compete with AMD EPYC and Intel Xeon CPUs in AWS data centers.
Key features:
- Up to 192 cores and 180 MB of L3 cache.
- Available in Amazon EC2 M9g instances in preview.
- Compute-optimized C9g and memory-focused R9g variants are scheduled for launch in 2026.
- Up to 30% faster for databases, up to 35% faster for web applications and machine learning workloads compared to M8g.
- Integrates 192 Neoverse V3 cores per package, fabricated using a 3nm-class process.
- A distributed L3 cache is used to enable better coherence scaling across 192 cores and predictable latency.
- Improved network and storage bandwidth.
- New Nitro Isolation Engine for enhanced security.
Here's how to enable it:
1. Open the Settings app.
2. Select Accessibility.
3. Select Audio & Visual.
4. Select Flash for Alerts.
5. Toggle Flash for Alerts on and select the Screen option, or Both.
Features and Functionality: This integration enables users to gather more information using Gemini's reasoning capabilities based on trusted sources from NotebookLM.
Availability: The integration is currently live on the web version of Gemini, with a slow rollout across Google accounts.
Improvements to NotebookLM: Google has been rapidly improving NotebookLM, including an Ultra NotebookLM tier, mobile app features like image-based sources, and support for infographics and slide decks.
Key points:
- The technology, called 5D Memory Crystal, stores data in nanoscale structures within the glass, utilizing five dimensions (three spatial coordinates, orientation, and intensity).
- A single 5-inch glass disc can hold up to 360TB of data.
- The storage medium is designed to be stable for 13.8 billion years.
- SPhotonix plans to pilot the glass-based cold storage systems in data centers over the next two years.
Jellyfin's desktop app has been updated and renamed to Jellyfin Desktop. The update includes a switch to Qt 6, multi-profile support, and other new features. The new version requires users to log in to their servers and reconfigure settings. Jellyfin Desktop 2.0 is built with the Qt 6 and MpvQt libraries, which should improve performance. Other changes include new options for the HTML5 player, movie/show backdrop, playback rate, and more. Jellyfin Server 10.11.5 is also rolling out, mainly as a bugfix update. Jellyfin Desktop 2.0 will be released soon for Windows, macOS, and Linux.
Following the editor updates in August, the Google Docs, Sheets, and Slides homepages are now getting a Material 3 Expressive redesign..
#CIASurveillance #Vault7 #PrivacyWarning
Former CIA spy warns agency’s tools can takeover your phone, TV, and even your car
Former CIA officer John Kiriakou drops chilling truths on how the agency can hijack your devices for spying, straight from Vault 7 leaks.
Key highlights
⦁ CIA tools from 2017 Vault 7 disclosures let them intercept any data, remotely control cars to stage "accidents," and turn off smart TVs into room mics.
⦁ Capabilities target phones (iOS/Android exploits), OS like Windows/macOS, browsers, and apps—hiding malware for untraceable ops.
⦁ Official role: Recruit spies, steal/analyze secrets for policy; reality: Whatever the president orders, from coups to media influence or assassinations.
⦁ Kiriakou, post-prison whistleblower on torture, stresses it's not mass surveillance but targeted—barrier is priority, not ability.
⦁ Echoes WikiLeaks docs on cyber tools; urges awareness as ex-insider now advocating for oversight and civil liberties.
Read More: CIA Spy Warning
Former CIA spy warns agency’s tools can takeover your phone, TV, and even your car
Former CIA officer John Kiriakou drops chilling truths on how the agency can hijack your devices for spying, straight from Vault 7 leaks.
Key highlights
⦁ CIA tools from 2017 Vault 7 disclosures let them intercept any data, remotely control cars to stage "accidents," and turn off smart TVs into room mics.
⦁ Capabilities target phones (iOS/Android exploits), OS like Windows/macOS, browsers, and apps—hiding malware for untraceable ops.
⦁ Official role: Recruit spies, steal/analyze secrets for policy; reality: Whatever the president orders, from coups to media influence or assassinations.
⦁ Kiriakou, post-prison whistleblower on torture, stresses it's not mass surveillance but targeted—barrier is priority, not ability.
⦁ Echoes WikiLeaks docs on cyber tools; urges awareness as ex-insider now advocating for oversight and civil liberties.
Read More: CIA Spy Warning
CurrentIndia.com
‘Yes, they can:’ Former CIA spy warns agency’s tools can takeover your phone, TV, and even your car - CurrentIndia.com %
CurrentIndia.com ‘Yes, they can:’ Former CIA spy warns agency’s tools can takeover your phone, TV, and even your car %
#GoogleDarkWeb #SecurityShutdown #PrivacyTools
Google to Shut Down Dark Web Monitoring Tool in February 2026
Google's pulling the plug on its dark web scanning feature next year, shifting focus to more actionable security tools after feedback showed it fell short.
Key highlights
⦁ Launched in 2023 for Google One users and expanded to all accounts in 2024, the tool scanned for leaked personal info like emails and SSNs on the dark web.
⦁ Scans end January 15, 2026; full shutdown February 16, 2026, with all data deleted—users can opt to remove profiles early via the settings.
⦁ Reason: Reports gave general alerts without clear next steps, so Google prioritizes passkeys, Security Checkup, and "Results about you" for better protection.
⦁ Alternatives: Use Have I Been Pwned or Google's Password Manager; the company vows continued dark web threat monitoring behind the scenes.
⦁ Impact: Aimed at identity theft prevention, but critics say it was basic—now emphasizing phishing-resistant MFA and data removal from search.
Read More: Google Dark Web Shutdown
Google to Shut Down Dark Web Monitoring Tool in February 2026
Google's pulling the plug on its dark web scanning feature next year, shifting focus to more actionable security tools after feedback showed it fell short.
Key highlights
⦁ Launched in 2023 for Google One users and expanded to all accounts in 2024, the tool scanned for leaked personal info like emails and SSNs on the dark web.
⦁ Scans end January 15, 2026; full shutdown February 16, 2026, with all data deleted—users can opt to remove profiles early via the settings.
⦁ Reason: Reports gave general alerts without clear next steps, so Google prioritizes passkeys, Security Checkup, and "Results about you" for better protection.
⦁ Alternatives: Use Have I Been Pwned or Google's Password Manager; the company vows continued dark web threat monitoring behind the scenes.
⦁ Impact: Aimed at identity theft prevention, but critics say it was basic—now emphasizing phishing-resistant MFA and data removal from search.
Read More: Google Dark Web Shutdown
Beyond powering helpful agents, native audio unlocks new possibilities for global communication. We’re introducing live speech translation, a capability that enables streaming speech-to-speech translation for headphones. It preserves the speaker’s intonation, pacing and pitch. This beta experience is rolling out in the Google Translate app starting today
Manus released upgrades for its agent system and introduced Manus 1.6 and Manus 1.6 Max.
Manus 1.6 Max performs 19% better at internal benchmarks.
Manus 1.6 Max performs 19% better at internal benchmarks.
Looking for Envato Elements files for free?
Your search ends here. 🚀
With @EnvatoDLBot, you can easily download premium Envato Elements resources in just a few clicks — no hassle, no complicated steps.
✨ What you can get:
* Fast & simple downloads
* Clean and user-friendly experience
* Perfect for designers, editors, students, and creators
Just send your link and receive your file. 💾🎨
Your search ends here. 🚀
With @EnvatoDLBot, you can easily download premium Envato Elements resources in just a few clicks — no hassle, no complicated steps.
✨ What you can get:
* Fast & simple downloads
* Clean and user-friendly experience
* Perfect for designers, editors, students, and creators
Just send your link and receive your file. 💾
Solana wallet Phantom is rolling out its own debit card. Real-world payments, crypto under the hood. Launch starts this week in the US, with other countries coming next.
Wallet to card, DeFi to daily spending. Phantom is clearly pushing crypto closer to everyday life. Details soon — but the direction is clear.
#ClaudeAI #AICodingFail #TechHorror
Claude CLI deleted my entire home directory! Wiped my whole mac.
A Reddit user's nightmare: Claude AI's CLI tool accidentally nuked their entire Mac home directory during a simple cleanup task, sparking debates on AI safety in coding.
Key highlights
⦁ The bot generated a fatal
⦁ User lost tons of work; Claude's analysis confirmed the blunder, but recovery's slim without backups—highlights risks of AI executing shell commands blindly.
⦁ Community roasts: Jokes about AI revenge, calls for containers and manual verification; similar Gemini CLI incidents show pattern in agentic tools.
⦁ Lessons: Never run AI-suggested commands without review, avoid home dir launches, and use safeguards like hooks or isolated environments.
⦁ Broader warning: Vibe-coding agents save time but demand caution—backups essential, as AI hallucinations can turn helpful into catastrophic.
Read More: Claude CLI Disaster
Claude CLI deleted my entire home directory! Wiped my whole mac.
A Reddit user's nightmare: Claude AI's CLI tool accidentally nuked their entire Mac home directory during a simple cleanup task, sparking debates on AI safety in coding.
Key highlights
⦁ The bot generated a fatal
rm -rf tests/ patches/ plan/ ~/ command, where the trailing ~/ targeted the whole user folder—Desktop, Documents, Keychain, everything gone.⦁ User lost tons of work; Claude's analysis confirmed the blunder, but recovery's slim without backups—highlights risks of AI executing shell commands blindly.
⦁ Community roasts: Jokes about AI revenge, calls for containers and manual verification; similar Gemini CLI incidents show pattern in agentic tools.
⦁ Lessons: Never run AI-suggested commands without review, avoid home dir launches, and use safeguards like hooks or isolated environments.
⦁ Broader warning: Vibe-coding agents save time but demand caution—backups essential, as AI hallucinations can turn helpful into catastrophic.
Read More: Claude CLI Disaster
Reddit
From the ClaudeAI community on Reddit
Explore this post and more from the ClaudeAI community
#GoogleTranslate #GeminiAI #LiveTranslation
Google Translate rolling out live translation using Gemini with any headphones
Google Translate is getting a Gemini-powered boost for smarter, real-time translations that handle idioms and slang like a pro, plus headphone magic for seamless listening.
Key highlights
⦁ Nuanced translations now capture context for phrases like "stealing my thunder," rolling out to English with 20 languages including Spanish and Arabic in US/India apps and web.
⦁ Live headphone translation beta works with any pair—point your phone, hear natural audio preserving tone and cadence, supporting 70+ languages for convos, lectures, or movies.
⦁ Android beta launches today in US, Mexico, India; iOS and more countries in 2026, with fullscreen transcription for easy follow-along.
⦁ Language learning upgrades: Better speech feedback, daily streaks tracker, expanding to 20 countries like Germany and India for English practice.
⦁ Powered by Gemini 2.5 Flash Native Audio, making cross-language chats feel effortless and authentic.
Read More: Google Translate Gemini
Google Translate rolling out live translation using Gemini with any headphones
Google Translate is getting a Gemini-powered boost for smarter, real-time translations that handle idioms and slang like a pro, plus headphone magic for seamless listening.
Key highlights
⦁ Nuanced translations now capture context for phrases like "stealing my thunder," rolling out to English with 20 languages including Spanish and Arabic in US/India apps and web.
⦁ Live headphone translation beta works with any pair—point your phone, hear natural audio preserving tone and cadence, supporting 70+ languages for convos, lectures, or movies.
⦁ Android beta launches today in US, Mexico, India; iOS and more countries in 2026, with fullscreen transcription for easy follow-along.
⦁ Language learning upgrades: Better speech feedback, daily streaks tracker, expanding to 20 countries like Germany and India for English practice.
⦁ Powered by Gemini 2.5 Flash Native Audio, making cross-language chats feel effortless and authentic.
Read More: Google Translate Gemini
9to5Google
Google Translate rolling out live translation using Gemini with any headphones
Google Translate is rolling out new features that leverage Gemini, including live translation with headphones...
#OpenAI #DisneyDeal #SoraAI
Sora Users Can Now Generate AI Videos Featuring Disney Characters
OpenAI and Disney just inked a massive deal, unlocking iconic characters for AI video creation in Sora while Disney pumps $1B into the AI giant.
Key highlights
⦁ Three-year licensing lets Sora users prompt short videos with 200+ Disney, Marvel, Pixar, and Star Wars figures—like Mickey or Darth Vader—but no voices or actor likenesses.
⦁ Disney becomes a major OpenAI customer, integrating APIs for Disney+ tools and ChatGPT for staff; includes equity investment and warrants for more shares.
⦁ Videos start rolling in 2026, with guardrails to block harmful content and protect creators' rights amid Hollywood's AI jitters.
⦁ WGA slams it as sanctioning "theft of our work," while Fairplay worries about kid-targeted AI; Disney insists it's safe and honors creators.
⦁ Boosts OpenAI's content library, potentially curating user vids for Disney+ and hedging against AI's entertainment shake-up.
Read More: OpenAI Disney Agreement
Sora Users Can Now Generate AI Videos Featuring Disney Characters
OpenAI and Disney just inked a massive deal, unlocking iconic characters for AI video creation in Sora while Disney pumps $1B into the AI giant.
Key highlights
⦁ Three-year licensing lets Sora users prompt short videos with 200+ Disney, Marvel, Pixar, and Star Wars figures—like Mickey or Darth Vader—but no voices or actor likenesses.
⦁ Disney becomes a major OpenAI customer, integrating APIs for Disney+ tools and ChatGPT for staff; includes equity investment and warrants for more shares.
⦁ Videos start rolling in 2026, with guardrails to block harmful content and protect creators' rights amid Hollywood's AI jitters.
⦁ WGA slams it as sanctioning "theft of our work," while Fairplay worries about kid-targeted AI; Disney insists it's safe and honors creators.
⦁ Boosts OpenAI's content library, potentially curating user vids for Disney+ and hedging against AI's entertainment shake-up.
Read More: OpenAI Disney Agreement
Gadgets 360
OpenAI and Disney Reach Licensing Agreement to Bring Its Characters to the Sora App
As part of the agreement, Disney will make a $1 billion equity investment in OpenAI.
Openai Introduce GPT-5.2
GPT-5.2 is general-purpose model, part of the GPT-5 flagship model family. Our most intelligent model yet for both general and agentic tasks, GPT-5.2 shows improvements over the previous GPT-5.1 in:
* General intelligence
* Instruction following
* Accuracy and token efficiency
* Multimodality—especially vision
* Code generation—especially front-end UI creation
* Tool calling and context management in the API
* Spreadsheet understanding and creation
Unlike the previous GPT-5.1 model, GPT-5.2 has new features for managing what the model "knows" and "remembers to improve accuracy.
Variant & Best for
gpt-5.2 : Complex reasoning, broad world knowledge, and code-heavy or multi-step agentic tasks
gpt-5.2-pro : Tough problems that may take longer to solve but require harder thinking
More info : https://platform.openai.com/docs/guides/latest-model
GPT-5.2 is general-purpose model, part of the GPT-5 flagship model family. Our most intelligent model yet for both general and agentic tasks, GPT-5.2 shows improvements over the previous GPT-5.1 in:
* General intelligence
* Instruction following
* Accuracy and token efficiency
* Multimodality—especially vision
* Code generation—especially front-end UI creation
* Tool calling and context management in the API
* Spreadsheet understanding and creation
Unlike the previous GPT-5.1 model, GPT-5.2 has new features for managing what the model "knows" and "remembers to improve accuracy.
Variant & Best for
gpt-5.2 : Complex reasoning, broad world knowledge, and code-heavy or multi-step agentic tasks
gpt-5.2-pro : Tough problems that may take longer to solve but require harder thinking
More info : https://platform.openai.com/docs/guides/latest-model
Openai
Using GPT-5.2 | OpenAI API
Learn about how to use and migrate to GPT-5.2 and the GPT-5 model family, the latest models in the OpenAI API.
