🛡 Anthropic says Claude Mythos found thousands of zero-day bugs across major systems
Anthropic reports its unreleased model Claude Mythos discovered thousands of zero-day vulnerabilities in operating systems and browsers within days, working largely автономously.
Examples include a 27-year-old flaw in OpenBSD enabling remote crashes, a 16-year-old bug in FFmpeg missed by millions of test runs, and a Linux kernel exploit chain giving full server control.
The company says the model can now find and exploit vulnerabilities at the level of top specialists. To prepare defenses, Anthropic launched Project Glasswing, giving 40+ companies access and $100M in credits to secure their systems first.
✔️@venture
Anthropic reports its unreleased model Claude Mythos discovered thousands of zero-day vulnerabilities in operating systems and browsers within days, working largely автономously.
Examples include a 27-year-old flaw in OpenBSD enabling remote crashes, a 16-year-old bug in FFmpeg missed by millions of test runs, and a Linux kernel exploit chain giving full server control.
The company says the model can now find and exploit vulnerabilities at the level of top specialists. To prepare defenses, Anthropic launched Project Glasswing, giving 40+ companies access and $100M in credits to secure their systems first.
✔️@venture
Ⓜ️ Meta unveils Muse Spark from new Superintelligence Lab
Meta introduced Muse Spark, the first model from its new Superintelligence Lab led by Alexandr Wang. It is not open source yet, though the company says future versions may be.
The model uses a redesigned architecture and data pipeline, reaching comparable performance to earlier models with much lower compute. It lags top models like Opus 4.6 and GPT-5.4 in coding, shows solid HLE results, and performs strongly in medical and multimodal tasks.
Meta also launched a Contemplating mode for running multiple agents, similar to Deep Think setups. Muse Spark is available on meta.ai and will expand to WhatsApp, Instagram, Facebook, and Ray-Ban devices.
📊 @tech
Meta introduced Muse Spark, the first model from its new Superintelligence Lab led by Alexandr Wang. It is not open source yet, though the company says future versions may be.
The model uses a redesigned architecture and data pipeline, reaching comparable performance to earlier models with much lower compute. It lags top models like Opus 4.6 and GPT-5.4 in coding, shows solid HLE results, and performs strongly in medical and multimodal tasks.
Meta also launched a Contemplating mode for running multiple agents, similar to Deep Think setups. Muse Spark is available on meta.ai and will expand to WhatsApp, Instagram, Facebook, and Ray-Ban devices.
📊 @tech
🤖 Anthropic explores building its own AI chips amid compute shortages
Anthropic is considering developing in-house AI chips as demand for Claude continues to grow. The company currently relies on a mix of Nvidia GPUs, Google TPUs, and Amazon hardware.
Ongoing shortages of accelerators are pushing Anthropic to diversify its infrastructure and reduce dependence on external vendors.
OpenAI is pursuing similar plans, turning custom AI chips into a clear industry direction.
✔️@venture
Anthropic is considering developing in-house AI chips as demand for Claude continues to grow. The company currently relies on a mix of Nvidia GPUs, Google TPUs, and Amazon hardware.
Ongoing shortages of accelerators are pushing Anthropic to diversify its infrastructure and reduce dependence on external vendors.
OpenAI is pursuing similar plans, turning custom AI chips into a clear industry direction.
✔️@venture
🧪 Researchers show 100% scores on AI coding benchmarks with simple exploits
Berkeley researchers demonstrated that SWE-bench Verified and Terminal-Bench can be fully gamed without solving tasks.
Their agent achieved 100% by exploiting benchmark logic.
In SWE-bench, the agent added a short script that forced all tests to return “passed,” scoring 100% across hundreds of tasks with zero real fixes.
In Terminal-Bench, it replaced dependencies during setup and injected a binary that wrote correct outputs, reaching 89/89.
The team found similar issues across multiple benchmarks, showing how easily agents can optimize for scores instead of real problem solving.
Benchmark results are becoming easier to manipulate than to trust.
📊 @tech
Berkeley researchers demonstrated that SWE-bench Verified and Terminal-Bench can be fully gamed without solving tasks.
Their agent achieved 100% by exploiting benchmark logic.
In SWE-bench, the agent added a short script that forced all tests to return “passed,” scoring 100% across hundreds of tasks with zero real fixes.
In Terminal-Bench, it replaced dependencies during setup and injected a binary that wrote correct outputs, reaching 89/89.
The team found similar issues across multiple benchmarks, showing how easily agents can optimize for scores instead of real problem solving.
Benchmark results are becoming easier to manipulate than to trust.
📊 @tech
This media is not supported in your browser
VIEW IN TELEGRAM
Charlie Munger: “The whole trick of the game is to have a few times when you know that something is better than average and to invest only where you have that extra knowledge.”
“And if you get just a few opportunities, that's enough.”
✔️@venture
“And if you get just a few opportunities, that's enough.”
✔️@venture
🚨 US Treasury warns banks after Anthropic’s Claude Mythos tests
The US Treasury held an urgent meeting with executives from Bank of America, Citigroup, Goldman Sachs, Morgan Stanley, and the Fed over the capabilities of Claude Mythos.
Officials described the model as highly effective at finding and exploiting security flaws, with early tests showing thousands of potential attack paths across banking systems.
Banks were told that human security teams are falling behind and that rapid AI adoption in defense is now required to keep up.
✔️@venture
The US Treasury held an urgent meeting with executives from Bank of America, Citigroup, Goldman Sachs, Morgan Stanley, and the Fed over the capabilities of Claude Mythos.
Officials described the model as highly effective at finding and exploiting security flaws, with early tests showing thousands of potential attack paths across banking systems.
Banks were told that human security teams are falling behind and that rapid AI adoption in defense is now required to keep up.
✔️@venture
⚖️ Harvey hits $200M+ ARR in 3 years by targeting top law firms first
In 2022, the founders tested GPT-3 on 100 legal questions from Reddit. 86 answers passed review by real lawyers. They sent results to Sam Altman and secured a seed check with early GPT-4 access.
Instead of starting with small firms, they went straight to Allen & Overy while still a 4-person team. The first 50 enterprise clients came through referrals, driven by trust networks in the legal industry.
About 20% of Harvey’s 460 employees are lawyers, helping close deals through peer credibility. The company is now shifting from selling licenses to enabling firms to build AI workflows and sell them via revenue share.
✔️@venture
In 2022, the founders tested GPT-3 on 100 legal questions from Reddit. 86 answers passed review by real lawyers. They sent results to Sam Altman and secured a seed check with early GPT-4 access.
Instead of starting with small firms, they went straight to Allen & Overy while still a 4-person team. The first 50 enterprise clients came through referrals, driven by trust networks in the legal industry.
About 20% of Harvey’s 460 employees are lawyers, helping close deals through peer credibility. The company is now shifting from selling licenses to enabling firms to build AI workflows and sell them via revenue share.
✔️@venture
💸 Top seed rounds jump to $174M valuations in Q1 2026
The top 5% of seed rounds reached $174M valuations in Q1 2026, up from $62M a year earlier. The 90th percentile also climbed from $50M to $94M.
The surge is driven by AI infrastructure, new LLMs, chips, and data center startups. Median valuations are rising slower, but the top end is heating up fast.
Early-stage pricing is starting to reflect infrastructure-scale bets.
✔️@venture
The top 5% of seed rounds reached $174M valuations in Q1 2026, up from $62M a year earlier. The 90th percentile also climbed from $50M to $94M.
The surge is driven by AI infrastructure, new LLMs, chips, and data center startups. Median valuations are rising slower, but the top end is heating up fast.
Early-stage pricing is starting to reflect infrastructure-scale bets.
✔️@venture
🚨 Second incident in days near Sam Altman’s home as two suspects arrested after shooting
Police responded around 2:56 a.m. to reports of shots fired in San Francisco’s Russian Hill. Investigators traced a vehicle linked to 25-year-old Amanda Tom and detained her with 23-year-old Muhamad Tarik Hussein on Taylor Street.
A search warrant at a residence uncovered three firearms. Both were arrested and booked for negligent discharge. No injuries were reported, and the investigation remains active.
This is the second incident targeting the same area within days. Earlier, a 20-year-old suspect allegedly threw a Molotov cocktail at Altman’s home.
✔️@venture
Police responded around 2:56 a.m. to reports of shots fired in San Francisco’s Russian Hill. Investigators traced a vehicle linked to 25-year-old Amanda Tom and detained her with 23-year-old Muhamad Tarik Hussein on Taylor Street.
A search warrant at a residence uncovered three firearms. Both were arrested and booked for negligent discharge. No injuries were reported, and the investigation remains active.
This is the second incident targeting the same area within days. Earlier, a 20-year-old suspect allegedly threw a Molotov cocktail at Altman’s home.
✔️@venture
🎮 GTA VI PC release may land Feb 2027, 4 months after consoles
Insider DetectiveSeeds says three former Rockstar developers point to a February 2027 PC launch. That would follow the console release by about 4 months.
If confirmed, this would be Rockstar’s shortest gap between console and PC. GTA V took around 1.5 years, while RDR 2 took about 1 year.
GTA VI is set to launch on PS5 and Xbox Series on November 19.
📊 @tech
Insider DetectiveSeeds says three former Rockstar developers point to a February 2027 PC launch. That would follow the console release by about 4 months.
If confirmed, this would be Rockstar’s shortest gap between console and PC. GTA V took around 1.5 years, while RDR 2 took about 1 year.
GTA VI is set to launch on PS5 and Xbox Series on November 19.
📊 @tech
❤1
Forwarded from Business avec Moteki.co
En 2026, les produits digitaux sont l’un des moyens les plus rapides de générer des revenus en ligne 💻
Formations, templates, eBooks, prompts IA ou abonnements… certains produits se vendent beaucoup mieux que d’autres 📈
Nous avons sélectionné 5 produits digitaux particulièrement rentables à lancer dès maintenant, même sans expérience 🚀
Le plus important n’est pas seulement l’idée, mais ta capacité à la mettre en ligne rapidement et à vendre efficacement ⚡️
Avec Moteki, tu peux créer ta boutique et commencer à vendre tes produits digitaux en quelques minutes 🛒
Découvre les 5 produits à fort potentiel et comment les lancer dans l’article complet 👇
Formations, templates, eBooks, prompts IA ou abonnements… certains produits se vendent beaucoup mieux que d’autres 📈
Nous avons sélectionné 5 produits digitaux particulièrement rentables à lancer dès maintenant, même sans expérience 🚀
Le plus important n’est pas seulement l’idée, mais ta capacité à la mettre en ligne rapidement et à vendre efficacement ⚡️
Avec Moteki, tu peux créer ta boutique et commencer à vendre tes produits digitaux en quelques minutes 🛒
Découvre les 5 produits à fort potentiel et comment les lancer dans l’article complet 👇
🤖 Gemini Launches Native macOS App
Google has released a native Gemini application for macOS, allowing users to summon the assistant with the Option + Space shortcut over any task.
The app can analyze screen content, handle files, and quickly answer questions without switching tabs. It also supports image and video generation.
Gemini is now globally available for users running macOS 15 and later versions.
📊 news.nzoko.com
Google has released a native Gemini application for macOS, allowing users to summon the assistant with the Option + Space shortcut over any task.
The app can analyze screen content, handle files, and quickly answer questions without switching tabs. It also supports image and video generation.
Gemini is now globally available for users running macOS 15 and later versions.
📊 news.nzoko.com
This media is not supported in your browser
VIEW IN TELEGRAM
🤖 Claude turns into a full design tool
Anthropic launched Claude Design, letting users create clean websites, apps, and presentations directly inside Claude.
It works with existing design systems, just upload your assets and the bot builds around them. You can export everything to PDF, PPTX, HTML, or send it straight to Canva for edits.
👉 Link 👈
💻 @developpeur
Anthropic launched Claude Design, letting users create clean websites, apps, and presentations directly inside Claude.
It works with existing design systems, just upload your assets and the bot builds around them. You can export everything to PDF, PPTX, HTML, or send it straight to Canva for edits.
👉 Link 👈
💻 @developpeur
❤1
Media is too big
VIEW IN TELEGRAM
😺 Codex expands from coding tool into full AI workspace
OpenAI is pushing Codex far beyond code. The new pitch is simple: one tool for much more than engineering.
🔸 Computer use on your own machine, at least on Mac for now
🔸 Built-in browser
🔸 Image generation
🔸 Deep integrations
🔸 Artifacts inside the app, including tables
🔸 Native automations
This looks like OpenAI’s answer to Claude Code Epitaxy.
📊 @tech
OpenAI is pushing Codex far beyond code. The new pitch is simple: one tool for much more than engineering.
🔸 Computer use on your own machine, at least on Mac for now
🔸 Built-in browser
🔸 Image generation
🔸 Deep integrations
🔸 Artifacts inside the app, including tables
🔸 Native automations
This looks like OpenAI’s answer to Claude Code Epitaxy.
📊 @tech
🤖 Google releases Gemini 3.1 Flash TTS with fine voice control
Google introduced Gemini 3.1 Flash TTS, a new speech model in the Gemini stack focused on controllable voice generation.
The model allows precise control over tone, pace, stress, and overall style using text tags. It also supports multi-speaker output while preserving each voice’s characteristics, making it usable for longer-form content.
Latency improved by tens of percent, including faster first-token response, bringing it closer to real-time use cases.
📊 @tech
Google introduced Gemini 3.1 Flash TTS, a new speech model in the Gemini stack focused on controllable voice generation.
The model allows precise control over tone, pace, stress, and overall style using text tags. It also supports multi-speaker output while preserving each voice’s characteristics, making it usable for longer-form content.
Latency improved by tens of percent, including faster first-token response, bringing it closer to real-time use cases.
📊 @tech
❤1
🍏 Apple rumored to build Mac Neo
Apple is reportedly working on a budget desktop called Mac Neo, priced around $300. The move follows Mac mini shortages and strong demand for lower-cost devices.
The device is expected to run on the A19 Pro chip from iPhone 17 Pro. Form factor is similar to Apple TV 4K, with 12GB RAM, USB-C ports, and about 35W power usage.
It is aimed at basic tasks like browsing, documents, and streaming, not heavy workloads where M-series chips dominate.
📊 @tech
Apple is reportedly working on a budget desktop called Mac Neo, priced around $300. The move follows Mac mini shortages and strong demand for lower-cost devices.
The device is expected to run on the A19 Pro chip from iPhone 17 Pro. Form factor is similar to Apple TV 4K, with 12GB RAM, USB-C ports, and about 35W power usage.
It is aimed at basic tasks like browsing, documents, and streaming, not heavy workloads where M-series chips dominate.
📊 @tech
This media is not supported in your browser
VIEW IN TELEGRAM
🤖 Anthropic launches Claude Design for UI, slides, and prototypes
Anthropic introduced Claude Design, a built-in studio inside Claude for creating presentations, interfaces, and layouts. The tool runs on Claude Opus 4.7 and generates a first version from a text prompt.
Users can edit through chat, inline comments, direct text changes, or auto-generated sliders for layout, colors, and spacing. The system also adapts to team style by analyzing codebases and design files, then applying the same palette, typography, and components to new projects.
Input supports text, images, and documents like DOCX or PPTX, as well as elements captured from websites. Outputs can be exported to PDF, PPTX, HTML, or Canva, or passed into Claude Code.
The product includes collaboration with shared projects and team editing. It is available in research preview for paid users.
📊 news.nzoko.com
Anthropic introduced Claude Design, a built-in studio inside Claude for creating presentations, interfaces, and layouts. The tool runs on Claude Opus 4.7 and generates a first version from a text prompt.
Users can edit through chat, inline comments, direct text changes, or auto-generated sliders for layout, colors, and spacing. The system also adapts to team style by analyzing codebases and design files, then applying the same palette, typography, and components to new projects.
Input supports text, images, and documents like DOCX or PPTX, as well as elements captured from websites. Outputs can be exported to PDF, PPTX, HTML, or Canva, or passed into Claude Code.
The product includes collaboration with shared projects and team editing. It is available in research preview for paid users.
📊 news.nzoko.com
🏎Rolls-Royce unveils Project Nightingale
Rolls-Royce introduced Project Nightingale, a limited coachbuilt model capped at 100 units with pricing starting around $1M.
The car features a convertible design, 24-inch wheels, and a starlight interior across the cabin.
It runs on a dual electric motor setup with at least 577 hp, though full specs are not yet disclosed.
Production is limited and positioned as a bespoke offering.
📊 news.nzoko.com
Rolls-Royce introduced Project Nightingale, a limited coachbuilt model capped at 100 units with pricing starting around $1M.
The car features a convertible design, 24-inch wheels, and a starlight interior across the cabin.
It runs on a dual electric motor setup with at least 577 hp, though full specs are not yet disclosed.
Production is limited and positioned as a bespoke offering.
📊 news.nzoko.com