What AI benchmarks miss about real-world performance - 惯性聚合

推荐订阅源

aimingoo的专栏

LINUX DO - 最新话题

News and Events Feed by Topic

Forbes - Security

Security Affairs

Secure Thoughts

Threat Intelligence Blog | Flashpoint

CERT Recently Published Vulnerability Notes

The Last Watchdog

Hacker News: Front Page

Cyber Attacks, Cyber Crime and Cyber Security

Lohrmann on Cybersecurity

Attack and Defense Labs

News | PayPal Newsroom

Privacy International News Feed

cs.CV updates on arXiv.org

Troy Hunt's Blog

Simon Willison's Weblog

News and Events Feed by Topic

The Hacker News

www.infosecurity-magazine.com

Hacker News: Ask HN

Google DeepMind News

Threat Research - Cisco Blogs

PCI Perspectives

Kaspersky official blog

Hacker News - Newest: "LLM"

Vulnerabilities – Threatpost

Know Your Adversary

Proofpoint News Feed

Recent Commits to openclaw:main

TaoSecurity Blog

cs.AI updates on arXiv.org

cs.CL updates on arXiv.org

The Exploit Database - CXSecurity.com

Security @ Cisco Blogs

Full Disclosure

The Blog of Author Tim Ferriss

VentureBeat

Anthropic says it hit a $30 billion revenue run rate after 'crazy' 80x growth OpenAI voice models get GPT-5-class reasoning AI agent identity: how to govern agentic AI in 6 stages Anthropic wants to own your agent's memory, evals, and orchestration — and that should make enterprises nervous Enterprise GPU utilization: why 95% of AI infrastructure spend is wasted Governance, not gatekeeping: How SAP brings enterprise‑grade safety to AI connectivity Anthropic introduces "dreaming," a system that lets AI agents learn from their own mistakes RL orchestration: how a 7B model routes tasks across GPT-5, Claude, and Gemini Meet ZAYA1-8B, a super efficient open reasoning model trained on AMD Instinct MI300 GPUs Anthropic Skill scanners passed every check. The malicious code rode in on a test file. Why AI breaks without context — and how to fix it Market research is too slow for the AI era, so Brox built 60,000 identical 'digital twins' of real people you can survey instantly, repeatedly The app store for robots has arrived: Hugging Face launches open-source Reachy Mini App Store with 200+ apps Scaling AI into production is forcing a rethink of enterprise infrastructure Miami startup Subquadratic claims 1,000x AI efficiency gain with SubQ model; researchers demand independent proof. GPT-5.5 Instant shows you what it remembered — just not all of it One command turns any open-source repo into an AI agent backdoor. OpenClaw proved no supply-chain scanner has a detection category for it AI agents are missing all the discussions your team is having. SageOX has an answer: agentic context infrastructure OpenAI turns its sold-out GPT-5.5 party into a monthlong Codex giveaway for 8,000 developers Inside AMEX’s agentic commerce stack: How intent contracts and single-use tokens enforce AI transactions Microsoft takes Agent 365 out of preview as shadow AI becomes an enterprise threat The RAG era is ending for agentic AI — a new compilation-stage knowledge layer is what comes next Salesforce Agentforce Operations fixes workflows breaking enterprise AI MCP command execution flaw: what security teams need to know The scaffolding era is over. LlamaIndex says context is the new moat xAI launches Grok 4.3 at an aggressively low price and a new, fast, powerful voice cloning suite Hidden IT problems are quietly creating risk, shadow IT, and lost productivity Alibaba's HDPO cuts AI agent tool overuse from 98% to 2% One tool call to rule them all? New open source Python tool Runpod Flash eliminates containers for faster AI dev Why OpenAI's 'goblin' problem matters — and how you can release the goblins on your own AI coding agents breached: attackers targeted credentials, not models | VentureBeat Writer launches AI agents that can act without prompts, taking on Amazon, Microsoft and Salesforce Netomi raises $110 million as Accenture and Adobe bet on AI for customer service Cheaper tokens, bigger bills: The new math of AI infrastructure Amazon’s OpenAI gambit signals a new phase in the cloud wars — one where exclusivity no longer applies Enterprise RAG rebuild: hybrid retrieval adoption tripled in Q1 2026 IBM launches Bob with multi-model routing and human checkpoints to turn AI coding into a secure production system AWS Quick's knowledge graph creates an orchestration blind spot Why enterprise GPU utilization is stuck at 5% — and why the fix makes it worse Definity embeds agents inside Spark pipelines to catch failures before they reach agentic AI systems How to build custom reasoning agents with a fraction of the compute American AI startup Poolside launches free, high-performing open model Laguna XS.2 for local agentic coding Mistral AI launches Workflows, a Temporal-powered orchestration engine already running millions of daily executions Microsoft and OpenAI gut their exclusive deal, freeing OpenAI to sell on AWS and Google Cloud Open source Xiaomi MiMo-V2.5 and V2.5-Pro are among the most efficient (and affordable) at agentic 'claw' tasks AI framework autonomously outperforms human-designed R&D baselines Why supply chains are the proving ground for automation‑led iPaaS RAG precision tuning can quietly cut retrieval accuracy by 40%, putting agentic pipelines at risk Enterprises are obsessing over model accuracy while ignoring the infrastructure layer where AI systems actually break. Monitoring LLM behavior: Drift, retries, and refusal patterns CVSS vulnerability triage: 5 failures, 5 fixes DeepSeek-V4 arrives with near state-of-the-art intelligence at fraction of the cost of Opus 4.7, GPT-5.5 85% of enterprises are running AI agents. Only 5% trust them enough to ship. AI synthetic audiences are already here and poised to upend the consulting industry Mystery solved: Anthropic reveals changes to Claude's harnesses and operating instructions likely caused degradation OpenAI's GPT-5.5 is here, and it's no potato: narrowly beats Anthropic's Claude Mythos Preview on Terminal-Bench 2.0 New startup BAND debuts agentic mesh with deterministic routing to govern multiple enterprise AI agents across model providers, channels OpenAI unveils Workspace Agents, a successor to custom GPTs for enterprises that can plug directly into Slack, Salesforce and more Google and AWS split the AI agent stack between control and execution Are you paying an AI ‘swarm tax’? Why single agents often beat complex systems OpenAI launches Privacy Filter, an open source, on-device data sanitization model that removes personal information from enterprise datasets Google doesn't pay the Nvidia tax. Its new TPUs explain why. Salesforce’s Agentforce Vibes 2.0 targets a hidden failure: context overload in AI agents Google’s Gemini can now run on a single air-gapped server — and vanish when you pull the plug The modern data stack was built for humans asking questions. Google just rebuilt its for agents taking action. Google’s new Deep Research and Deep Research Max agents can search the web and your private data Vercel breach exposes the OAuth gap most security teams cannot detect, scope or contain The AI governance mirage: Why 72% of enterprises don’t have the control and security they think they do OpenAI's ChatGPT Images 2.0 is here and it does multilingual text, full infographics, slides, maps, even manga — seemingly flawlessly Kimi K2.6 runs agents for days — and exposes the limits of enterprise orchestration What AI model should you use for revenue intelligence? Von says all the big ones, and it will automate mixing and matching for you Three AI coding agents leaked secrets through a single prompt injection. One vendor's system card predicted it Train-to-Test scaling explained: How to optimize your end-to-end AI compute budget for inference AI agent security maturity audit: enterprises funded stage one, stage-three threats arrived anyway Anthropic just launched Claude Design, an AI tool that turns prompts into prototypes and challenges Figma Should my enterprise AI agent do that? NanoClaw and Vercel launch easier agentic policy setting, approval dialogs for messaging apps Salesforce launches Headless 360 to turn its entire platform into infrastructure for AI agents Are we getting what we paid for? How to turn AI momentum into measurable value OpenAI debuts GPT-Rosalind, a new limited access model for life sciences, and broader Codex plugin on Github OpenAI drastically updates Codex desktop app to use all other apps on your computer, generate images, preview webpages Anthropic releases Claude Opus 4.7, narrowly retaking lead for most powerful generally available LLM AI lowered the cost of building software. Enterprise governance hasn’t caught up Microsoft patched a Copilot Studio prompt injection. The data exfiltrated anyway Frontier models are failing one in three production attempts — and getting harder to audit Meta researchers introduce 'hyperagents' to unlock self-improving AI for non-coding tasks We tested Anthropic’s redesigned Claude Code desktop app and 'Routines' -- here's what enterprises should know AI's next bottleneck isn't the models — it's whether agents can think together Adobe’s new Firefly AI Assistant wants to run Photoshop, Premiere, Illustrator and more from one prompt Traza raises $2.1 million led by Base10 to automate procurement workflows with AI Agentic coding at enterprise scale demands spec-driven development Designing the agentic AI enterprise for measurable performance Five signs data drift is already undermining your security models Your developers are already running AI locally: Why on-device inference is the CISO’s new blind spot AI agent credentials live in the same box as untrusted code. Two new architectures show where the blast radius actually stops. Intuit compressed months of tax code implementation into hours — and built a workflow any regulated-industry team can adapt OpenAI introduces ChatGPT Pro $100 tier with 5X usage limits for Codex compared to Plus Mythos autonomously exploited vulnerabilities that survived 27 years of human review. Security teams need a new detection playbook Claude, OpenClaw and the new reality: AI agents are here — and so is the chaos Goodbye, Llama? Meta launches new proprietary AI model Muse Spark — first since Superintelligence Labs' formation LLM-referred traffic converts at 30-40% — and most enterprises aren't optimizing for it

What AI benchmarks miss about real-world performance

VB Staff · 2026-06-11 · via VentureBeat

此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。原文来自 — 版权归原作者所有。