aidigester

341 Malicious Skills Discovered on OpenClaw ClawHub: The Naked Truth of AI Agent Security

by

341 Malicious Skills Found on OpenClaw ClawHub

  • Koi Security found malware in 341 out of 2,857 ClawHub skills
  • Atomic Stealer steals cryptocurrency wallets and API keys
  • Expert calls it a “security dumpster fire,” 3 additional RCE vulnerabilities disclosed

What Happened?

341 malicious skills were found on ClawHub, the marketplace for AI agent OpenClaw.[The Hacker News] Koi Security audited 2,857 skills and found that 12% contained malware.

Attackers distributed skills disguised as cryptocurrency bots and wallet trackers. When installed, Atomic Stealer or a trojan is executed.[VirusTotal]

Why is it Important?

OpenClaw is an AI agent that runs locally. It has system-wide access. The only requirement for ClawHub registration is a week-old GitHub account.[The Register]

Laurie Voss of Arize called it a “security dumpster fire.” Recently, 1 RCE and 2 command injection vulnerabilities have been additionally disclosed in the last 3 days.

What Happens Next?

OpenClaw has introduced a reporting feature. Skills reported more than 3 times are automatically hidden. VirusTotal has started analyzing AI skills with Code Insight.

Frequently Asked Questions (FAQ)

Q: What is OpenClaw and why is it dangerous?

A: OpenClaw is a locally executed AI agent. It was formerly known as Clawdbot. It has system-wide privileges. Malicious skills can access files, credentials, and browser data.

Q: What data is being stolen?

A: Cryptocurrency API keys, wallet private keys, SSH credentials, and browser passwords are targeted. Keylogging data and environment variables are also collected.

Q: What if I have already installed a skill?

A: Delete any suspicious skills. Immediately replace cryptocurrency wallets and API keys. Also, change browser passwords.

If this article was helpful, please subscribe to AI Digester.

References

a16z’s $15 Billion AI Infrastructure Investment: Where Are They Betting and What Are They Ignoring?

by

a16z’s $15 Billion AI Infrastructure Investment: Where They’re Betting and What They’re Ignoring

  • $1.7 Billion Allocated to Infrastructure Team, Largest Ever
  • Focus on Agent-Native Infrastructure and Multimodal Data Management
  • Existing Systems Mistake Agent Traffic for DDoS Attacks

What Happened?

a16z has raised its largest fund ever, totaling $15 billion. Of this, $1.7 billion has been allocated to the infrastructure team.[a16z] The infrastructure team, led by General Partner Jennifer Li, is investing in companies like OpenAI, ElevenLabs, Cursor, and Black Forest Labs.[TechCrunch]

The investment direction is noteworthy. Li identified unstructured multimodal data within enterprises as the biggest bottleneck and untapped treasure. Unstructured data such as PDFs, videos, and logs are crippling RAG and agent systems.[a16z Big Ideas 2026]

Why Does It Matter?

The AI infrastructure paradigm is changing. Existing systems tailored to human speed cannot handle the bursty and recursive traffic patterns generated by agents. When a single agent goal triggers thousands of subtasks in milliseconds, legacy databases mistake this for a DDoS attack.[a16z Big Ideas 2026]

The areas a16z is ignoring are also clear. Simple fine-tuning tools or general SaaS are not of interest. Instead, Li emphasizes that search infrastructure is far more important than people think. True enterprise AI is only possible when agents have continuous access to accurate data semantics, solving the context problem.

What Happens Next?

Building agent-native infrastructure will be a key task in 2026. Reducing cold starts, eliminating latency variations, and increasing concurrency by orders of magnitude are required. With this $15 billion, a16z aims to define the AI infrastructure standard, accounting for 18% of the US venture capital market.[Crunchbase]

Ultimately, the winners will be startups that redesign the control plane for agent workloads. Existing cloud infrastructure won’t cut it.

Frequently Asked Questions (FAQ)

Q: How much is a16z investing in AI infrastructure?

A: $1.7 billion has been allocated to the infrastructure team out of a total fund of $15 billion. This is managed alongside other teams such as Apps ($1.7 billion), Bio + Health ($700 million), American Dynamism ($1.176 billion), and Growth ($6.75 billion). Infrastructure targets AI, cloud platforms, and distributed systems.

Q: What is agent-native infrastructure?

A: Infrastructure designed to handle the massive concurrent execution, recursive, and bursty workloads generated by agents. Existing systems are optimized for predictable human-speed traffic, causing them to mistake agent patterns for attacks. Coordination of routing, locking, and state management is a key challenge.

Q: What are the main portfolio companies of the a16z infrastructure team?

A: OpenAI, ElevenLabs (valued at $11 billion), Cursor, Black Forest Labs, Ideogram, Fal, and others. Jennifer Li, General Partner, leads the team, focusing on data systems, developer tools, and AI infrastructure. Dozens of other investments are also underway.

If you found this article useful, please subscribe to AI Digester.

References

OpenAI vs Anthropic: Super Bowl Ad Feud, Sam Altman Directly Refutes

by

OpenAI vs Anthropic: Super Bowl Ad Feud

  • Anthropic Criticizes ChatGPT’s Ad Policy
  • Sam Altman Rebuts, Calling it “Dishonest”
  • AI Ad Model Debate Intensifies

What Happened?

Anthropic satirized the introduction of ChatGPT ads in a Super Bowl ad.[9to5Mac] The key message is “Ads are coming to AI. But not to Claude.”[Ad Age]

Why Does it Matter?

Sam Altman retorted that “the premise is dishonest.” He attacked Anthropic as an “authoritarian company.”[The Verge]

What Happens Next?

OpenAI does not include ads for paid subscribers. Anthropic keeps Claude ad-free.

Frequently Asked Questions (FAQ)

Q: What is the Anthropic Super Bowl ad?

A: It satirizes ChatGPT with a scene where AI gives advice and suddenly shows an ad. The message is that Claude is ad-free.

Q: Why did Sam Altman refute it?

A: OpenAI claims they don’t insert ads in the way depicted in the ad. He criticized it as deceptive advertising.

Q: Do ChatGPT paid subscribers also see ads?

A: No. There are no ads for paid subscribers. It only applies to free users.

If you found this helpful, please subscribe to AI Digester.

References

Resolve AI, Unicorn Status Achieved with $125 Million Investment: A New Horizon for SRE Automation

by

Unicorn in 16 Months: The SRE Automation Market is Changing

  • Resolve AI Raises $125M in Series A, Achieves $1B Valuation
  • Autonomous SRE Platform Created by Splunk Alumni
  • Secured Over 20 Enterprise Customers Including Coinbase, DoorDash, Salesforce

What Happened?

AI-powered SRE automation startup Resolve AI raised $125 million in a Series A round, earning a $1 billion valuation.[TechCrunch] Lightspeed Venture Partners led the investment, with additional participation from Greylock Partners, Unusual Ventures, Artisanal Ventures, and A*.[Pulse 2.0]

The company is led by CEO Spiros Xanthos and Mayank Agarwal. Both were co-creators of OpenTelemetry and key figures in Splunk’s observability business unit. They have two previous exit experiences with Splunk and VMware.[Pulse 2.0]

Why Does It Matter?

SRE (Site Reliability Engineering) is the practice of diagnosing and resolving failures in production environments. Traditionally, this involved manually analyzing logs and identifying root causes. Resolve AI automates this process with AI agents.

Unlike GitHub Copilot or Cursor, which focus on the development phase, Resolve AI specializes in production operations. CEO Spiros Xanthos stated that “the next frontier of software engineering is applying AI to production operations.” A multi-agent system encompassing code, infrastructure, and telemetry is key.

It is also noteworthy that they became a unicorn just 16 months after coming out of stealth mode. More than 20 large enterprises, including Coinbase, DoorDash, MongoDB, Salesforce, and Zscaler, are already customers. This signals that large companies are actually starting to spend money on SRE automation.

What Happens Next?

The funding will be used to accelerate product development and expand the engineering/GTM teams. With intense competition already in the observability market, it remains to be seen whether Resolve AI can create a new category of SRE automation.

Existing observability vendors like Datadog and New Relic are also enhancing their AI capabilities. However, it is still uncertain how they will compete with platforms like Resolve AI, which are designed from the ground up with AI agents.

Frequently Asked Questions (FAQ)

Q: What is Resolve AI?

A: Resolve AI is an SRE automation platform that uses AI agents to automatically diagnose and resolve failures in production environments. It was founded by Spiros Xanthos and Mayank Agarwal, co-creators of OpenTelemetry.

Q: How is it different from existing observability tools?

A: Datadog and New Relic focus on monitoring and alerting. Resolve AI goes a step further by automatically analyzing the root cause of failures and suggesting solutions or even executing them directly.

Q: Which companies are using it?

A: More than 20 large enterprises, including Coinbase, DoorDash, MongoDB, MSCI, Salesforce, and Zscaler, are customers. Most of them have complex cloud infrastructures and require high production stability.

If you found this helpful, please subscribe to AI Digester.

References

Gizmo: TikTok of Mini-Apps Made with Vibe Coding

by

New York startup Atma Sciences has launched Gizmo, a vibecoding-based mini-app platform. You can create interactive content with just prompts, without coding.

What is Gizmo?

Gizmo is a platform for sharing mini-apps in a TikTok-style vertical feed. Users can create interactive experiences using text, photos, sound, and touch. AI analyzes natural language prompts to automatically generate code.

How does it work?

No coding knowledge is required. If you describe your idea in natural language, AI generates code and transforms it into an interactive experience. You can create various forms of digital toys such as puzzles, memes, art, and animations.

Interact in various ways, such as tapping or swiping the screen, drawing, or dragging. You can like and comment on your favorite Gizmos, and remix existing Gizmos to create your own version.

What is the growth status?

After launching on iOS in July 2025, it expanded to Android in early 2026. It currently has approximately 600,000 installations, half of which are from US users. It achieved a growth rate of 312% from October to December 2025.

Atma Sciences was co-founded by Rudd Fawcett and Brandon Francis. It is operated by CEO Josh Siegel and CTO Daniel Amitay. It has raised $5.49 million in seed funding from First Round Capital, among others.

FAQ

Do I need coding skills to use Gizmo?

Not at all. AI automatically generates code when you describe your idea in natural language.

Which platforms are supported?

Both iOS and Android are supported. You can download it for free from the App Store and Google Play.

Can I modify someone else’s Gizmo?

You can create your own version based on an existing Gizmo through the remix feature.

Google Surpasses $400 Billion in Annual Revenue: AI Drives Record-Breaking Performance

by

Google’s Annual Revenue Exceeds $400 Billion: AI Drives Record-Breaking Performance

  • Alphabet, Annual Revenue of $400 Billion for the First Time
  • Google Cloud Grows by 48%
  • $185 Billion Investment in AI Expected in 2026

What Happened?

Alphabet announced its Q4 2025 results. Annual revenue exceeded $400 billion for the first time.[CNBC] Cloud led the growth with a 48% surge.[Benzinga]

Why Does It Matter?

Cloud’s 48% growth surpasses AWS and Azure. Surpassing 750 million Gemini users and reducing serving costs by 78% are also noteworthy.[9to5Google]

What Happens Next?

Capital expenditure of $185 billion is expected in 2026. The Big Tech AI arms race is in full swing.

Frequently Asked Questions (FAQ)

Q: What is the reason for the rapid cloud growth?

A: Companies are adopting cloud for AI training and inference. TPU and Gemini are key.

Q: What is the impact of large-scale investment?

A: Short-term pressure on profit margins, but the market sees AI investment as a prerequisite.

Q: What is the significance of 750 million Gemini users?

A: Performing well in the competition with ChatGPT. Platform integration is advantageous.

If you found this helpful, please subscribe to AI Digester.

References

Google Surpasses $400 Billion in Annual Revenue for the First Time: AI Investment in Numbers

by

Google Surpasses $400 Billion in Annual Revenue for the First Time: AI Investment in Numbers

  • First time surpassing $400 billion in annual revenue in 2025, 18% growth compared to the previous year
  • Google Cloud grows rapidly by 48%, annual run rate of $70 billion
  • Capital expenditure to double in 2026 to $175 billion ~ $185 billion

What Happened?

Alphabet announced its Q4 2025 results. Annual revenue surpassed $400 billion for the first time.[The Verge] Q4 revenue was $113.8 billion, an 18% increase year-over-year.[9to5Google]

Google Cloud led the growth. Q4 revenue was $17.66 billion, exceeding an annual run rate of $70 billion. YouTube also achieved $60 billion annually. Gemini has 750 million monthly users.[Finviz]

Why Does It Matter?

It shows Google’s unique position among Big Tech companies. It’s the result of successfully layering cloud and AI on top of search advertising.

The 48% growth of the cloud is key. It ranks third after AWS and Azure, but its growth rate is the fastest. The surge in AI workloads is leveraging the power of TPU infrastructure.

What Happens Next?

Alphabet has announced capital expenditures of $175 billion to $185 billion for 2026. This is almost double that of 2025. It is expected to focus on expanding data centers and securing AI chips.

Frequently Asked Questions (FAQ)

Q: Can Google Cloud catch up with AWS?

A: AWS has a larger market share, but Google is growing faster. 48% exceeds AWS and Azure. If it leverages its strengths in AI workloads, it can narrow the gap. A complete reversal will take time.

Q: Where will the $175 billion in capital expenditure be used?

A: Most of it will be invested in AI infrastructure. Major items include data center construction, securing GPUs and TPUs, and expanding power infrastructure. It is a foundational investment to strengthen the Gemini platform.

Q: How does Gemini’s 750 million users compare to ChatGPT?

A: ChatGPT has over 300 million weekly active users. Gemini’s 750 million is a monthly figure, so direct comparison is difficult. Thanks to the Android ecosystem, the spread is rapid.

If you found this article helpful, please subscribe to AI Digester.

Reference Materials

Google Surpasses $400 Billion in Annual Revenue: AI and Cloud Drive Record-Breaking Performance

by

$400 Billion, A Historic First

  • Alphabet First to Surpass $400 Billion in Annual Revenue in 2025
  • Q4 Revenue of $113.8 Billion, Up 18% Year-Over-Year
  • Driven by Google Cloud’s Explosive 48% Growth

What Happened?

Alphabet announced its Q4 2025 earnings. Q4 revenue was $113.83 billion, exceeding Wall Street’s estimate of $111.43 billion.[CNBC] Net income surged 30% year-over-year to $34.46 billion. CEO Sundar Pichai described it as a “tremendous quarter.”[9to5Google]

Annual revenue surpassed $400 billion for the first time in the company’s history. This was primarily driven by the rapid growth of AI and cloud businesses.[TheWrap]

Why Does It Matter?

The most notable achievement is Google Cloud’s performance. Revenue was $17.66 billion, significantly exceeding the estimate of $16.18 billion. This represents a 48% year-over-year growth. The cloud contract backlog jumped to $240 billion, more than double what it was a year ago.[FinViz]

Search advertising remained strong. Google Search revenue was $63.07 billion, up 17% from $54.03 billion the previous year. This is evidence that AI features are successfully monetizing without harming the search experience.

YouTube surpassed $60 billion in annual revenue. Q4 advertising revenue grew 9% to $11.38 billion, but fell short of the estimate of $11.84 billion. Nevertheless, the platform as a whole has become a business on par with Netflix.[TheDesk]

What Happens Next?

Alphabet forecasts capital expenditures (CapEx) of $175 billion to $185 billion in 2026. This is more than double that of 2025 at the upper end. These funds will be used to expand AI computing capacity and meet cloud demand. This also includes expanding the research infrastructure of Google DeepMind.

In an environment where AI investment competition is overheating, this level of betting is an expression of the will to seize the market. The question is the speed of return on investment. How this investment is reflected in performance from the second half of 2026 will be key.

Frequently Asked Questions (FAQ)

Q: What is Google’s exact annual revenue for 2025?

A: Alphabet announced that its annual revenue for 2025 exceeded $400 billion. The exact figure can be found in the official annual report. This is a milestone achieved for the first time since the company’s founding.

Q: Why has Google Cloud grown so rapidly?

A: The main reason is the surge in demand for AI workloads. Companies are investing in cloud infrastructure to train and infer their own AI models. Alphabet’s AI chips and infrastructure are proving to be competitive.

Q: Why is the 2026 investment scale double that of 2025?

A: Building AI computing infrastructure costs astronomical amounts. GPUs, data centers, and power infrastructure all require large-scale upfront investment. As demand from cloud customers has surged, securing capacity has become urgent.

If you found this article useful, please subscribe to AI Digester.

References

Gemini App Surpasses 750 Million MAU: How Google is Chasing ChatGPT

by

AI Chatbot Used by 750 Million People

  • Google Gemini app achieves 750 million monthly active users (MAU)
  • Announced in the Q4 2025 earnings call
  • Full-fledged competition with ChatGPT and Meta AI

What Happened?

Google announced in its Q4 2025 earnings call that the Gemini app has surpassed 750 million monthly active users.[TechCrunch] Google is directly competing with OpenAI’s ChatGPT and Meta AI in the generative AI chatbot market.

Gemini is Google’s flagship AI chatbot. It is integrated throughout the Google ecosystem, including Search, Gmail, and Workspace. It has also rapidly gained users as a standalone app.

Why is it Important?

750 million MAU is a significant number. Compared to ChatGPT, which recorded 100 million weekly active users at the end of 2024, Google’s scale is evident. Of course, MAU and weekly active users are different metrics. However, it is clear that Google is catching up quickly despite being a latecomer.

Google’s strength is its distribution power. Gemini can be pre-installed on Android phones, Chrome browsers, and Google Search. Users can experience AI features without having to install a separate app. This strategy is working.

What Will Happen Next?

The AI chatbot market is now a three-way race. OpenAI, Google, and Meta are competing with their respective strengths. OpenAI has a technology leader image and a developer ecosystem. Google has distribution power and integration with existing services as its weapon. Meta has accessibility based on its social platform.

From the user’s perspective, the choices are increasing. As competition intensifies, the speed of feature improvement will also accelerate. However, you should carefully consider which platform to entrust your data to.

Frequently Asked Questions (FAQ)

Q: Is the Gemini app free?

A: Basic features are provided free of charge. You can use it right away with just a Google account. Gemini Advanced, which includes advanced features, requires a subscription fee of $19.99 per month. It is included in the Google One AI Premium plan.

Q: Which is better, ChatGPT or Gemini?

A: It depends on the use case. ChatGPT excels in creative writing and coding. Gemini is excellent at integrating with Google services. If you need search result integration, Gmail summaries, or calendar integration, Gemini is more convenient.

Q: How big is 750 million MAU in reality?

A: It’s very big. There are approximately 5 billion internet users worldwide. That means 15% of them use Gemini at least once a month. There is a lot of overlap with Google Search users, but it is an overwhelming scale as a standalone AI chatbot service.

If you found this article helpful, please subscribe to AI Digester.

References

H Company Holo2: Achieved 1st Place in UI Localization Benchmark

by

235B Parametric model, UI automation을 완전히 뒤집다

  • ScreenSpot-Pro 벤치마크에서 78.5%로 SOTA 달성
  • Agent localization으로 성능 10-20% 향상
  • 4K 고해상도 인터페이스에서도 작은 UI 요소 정확하게 찾아냄

무슨 일이 있었나?

H Company에서 UI Localization (유저 인터페이스 요소 위치 식별)을 위한 전문가 모델 Holo2-235B-A22B를 출시했다. [Hugging Face] 이 235B 파라미터 규모의 모델은 스크린샷에서 버튼, 텍스트 필드, 링크와 같은 UI 요소의 정확한 위치를 찾는다.

핵심은 Agentic Localization 기술이다. 한 번에 정답을 제공하는 것이 아니라, 여러 단계에 걸쳐 예측을 개선한다. 덕분에 4K 고해상도 화면의 작은 UI 요소도 정확하게 잡아낸다. [Hugging Face]

왜 중요한가?

GUI agent 분야가 뜨겁다. Claude Computer Use, OpenAI Operator와 같은 빅테크 기업들이 UI automation 기능을 내놓기 위해 경쟁하고 있다. 하지만 작은 스타트업인 H Company가 이 분야 벤치마크에서 1위를 차지했다.

개인적으로 주목하는 점은 agentic 방식이다. 기존 모델은 한 번에 위치를 조정하려고 시도할 때 실패하는 경우가 많았지만, 여러 번 시도하여 모델을 개선하는 접근 방식이 효과적이었다. 10-20% 성능 향상 수치가 이를 증명한다.

솔직히 235B 파라미터는 꽤 무겁다. 실제 프로덕션 환경에서 얼마나 빠르게 실행될지는 지켜봐야 한다.

앞으로 어떻게 될까?

GUI agent 경쟁이 심화되면서 UI Localization 정확도가 핵심 차별화 요소가 될 것으로 예상된다. H Company 모델이 오픈 소스로 공개되었으므로 다른 agent framework에 통합될 가능성이 높다.

RPA (robotic process automation) 시장에도 영향을 미칠 수 있다. 기존 RPA 도구는 규칙 기반이었지만, 이제 비전 기반 UI 이해가 표준이 될 수 있다.

자주 묻는 질문 (FAQ)

Q: UI Localization이 정확히 무엇인가?

A: 스크린샷을 보고 특정 UI 요소 (버튼, 입력 창 등)의 정확한 좌표를 찾는 기술이다. 간단히 말해서, AI가 화면을 보고 어디를 클릭해야 하는지 아는 것이다. GUI automation agent의 핵심 기술이다.

Q: 기존 모델과 무엇이 다른가?

A: Agentic localization이 핵심이다. 한 번에 맞추려고 하는 것이 아니라, 여러 단계로 정교하게 다듬는다. 사람이 목표를 찾기 위해 화면을 스캔하는 방식과 유사하다. 이 방법으로 10-20%의 성능 향상을 달성했다.

Q: 모델을 직접 사용할 수 있나?

A: Hugging Face에서 연구용으로 공개되었다. 하지만 235B 파라미터 모델이므로 상당한 GPU 리소스가 필요하다. 실제 프로덕션 애플리케이션보다는 연구 또는 벤치마킹 용도에 적합하다.

이 기사가 유용했다면 AI Digester를 구독해 주세요.

참고 자료