AI News Archive: August 28, 2026 — Part 3
Sourced from 500+ daily AI sources, scored by relevance.
- Inside Meta’s Push to Put Robots to Work in Data Centers
The company is testing robots that can swap cables, reset servers, and take on other tasks performed by technicians, fueling concerns among some workers that their jobs could be at risk.
Score: 61🌐 MovesAug 28, 2026https://www.wired.com/story/inside-metas-experiments-with-data-center-robots/ - South Korea is funding free AI to help citizens book appointments, file taxes, and everyday tasks
South Korea is handing out free AI subscriptions to boost homegrown AI services and reduce dependency on US and China based services.
- Beyond answers: New Genie One features to turn insights into action
We all know the pattern: you ask an AI tool a question and get an answer in seconds,...
Score: 61🌐 MovesAug 28, 2026https://www.databricks.com/blog/beyond-answers-new-genie-one-features-turn-insights-action - Supporting Thailand’s next generation of AI startups
OpenAI and Thailand’s MHESI launch an eight-week accelerator helping 10 health, wellness, and education startups turn AI prototypes into trusted products.
Score: 60🌐 MovesAug 28, 2026https://openai.com/index/supporting-next-generation-ai-startups-thailand - Researcher shows how Claude Code can be tricked simply by asking it to summarize a website
More prompt-injection hijinks from wunderwuzzi
- Gemini Omni 1.1 🎬, Cohere Parse 📄, Codex persistent mode 👨💻
Gemini Omni 1.1 🎬, Cohere Parse 📄, Codex persistent mode 👨💻
- How AI agents "radicalized" a top Meta exec into quitting her job
Clara Shih on how AI will soon displace the first few rungs on the career ladder — and why she left Big Tech to do something about it
Score: 59🌐 MovesAug 28, 2026https://www.platformer.news/clara-shih-new-work-dear-cc-interview-meta-salesforce/ - DLSS 5 leaked and modders are putting Nvidia’s AI effects on everything
Modders are trying out an unofficial version of Nvidia's DLSS 5 on Skyrim, Cyberpunk 2077, GTA V, and a bunch of other games after code for the AI upscaling tech appeared in an early-access build of NBA 2K27. Members of the RenoDX modding channel on Discord reportedly found a way to extract the DLSS "Neural […]
- New risk categories for AI change risk assessment
New risk categories for AI change risk assessment Atlassian Community
- ICE Plans to Spend Millions on Boston Dynamics Dog Robots
ICE wants the robot dogs to improve “officer safety.”
Score: 59🌐 MovesAug 28, 2026https://www.404media.co/ice-plans-to-spends-millions-on-boston-dynamics-dog-robots/ - Z.ai’s Latest Model Intensifies Competition for Low-Cost Offerings
Z.ai’s Latest Model Intensifies Competition for Low-Cost Offerings The Information
Score: 59🤖 ModelsAug 28, 2026https://www.theinformation.com/briefings/z-ais-latest-model-intensifies-competition-low-cost-offerings - Google and Microsoft were in talks for Nscale compute deal that went to Anthropic
Nscale is building one of the world's largest data center campuses in West Virginia, as the battle for compute heats up.
- OpenAI rolls out ads on select ChatGPT plans in India to boost monetization, support wider access
Ads have been rolled out for free users and those on the 399 rupee ($4) per month 'Go' plan.
Score: 58🌐 MovesAug 28, 2026https://www.cnbc.com/2026/08/28/openai-strategy-india-anthropic-ipo.html - Cohere Parse 5 loses the benchmark on points. It wins on cost per page.
Enterprises trying to feed PDFs, slides and scanned documents into AI pipelines keep running into the same wall: the tools either miss the structure — tables, charts, layout — or cost too much to run at scale. Cohere released Parse 5 on Thursday , positioning it on price-to-performance, not raw accuracy — the right cost-capability mix for enterprise scale. Parse 5 is a 2.3-billion-parameter vision language model built to convert PDFs, slides and images into structured Markdown at enterprise scale. Cohere's own published benchmark comparison puts Parse 5 behind three larger, general-purpose frontier models on accuracy. GPT-5.5, Opus 4.8 and Gemini 3.5 Flash all score higher than Parse on the three ParseBench dimensions Cohere reports. Cohere is not claiming the top score. It is claiming the best price for a score close to the top.The company priced the model at $1.50 per 1,000 pages through its API, with Model Vault, Cohere's secure, single-tenant platform for managed inference, available for higher-volume deployment. "Document parsing isn't solved because the hard part isn't reading text, it's preserving structure and meaning," Nils Reimers, VP of AI Search at Cohere, told VentureBeat. "Enterprise documents mix tables, diagrams, charts, and formatting that change the interpretation of the data. Most tools still drop structure or hallucinate content, and even frontier models break on layout‑heavy pages." Inside the single-pass architecture Parse 5 takes a page as an image, runs it through a single vision-language model pass and returns structured Markdown, collapsing the OCR-plus-model pipeline most tools run as separate steps. Architecture. It is a 2.3-billion-parameter vision language model built on Cohere Labs' North-Micro-Vision-Instruct architecture, with an 8,192-token context window and roughly a 4.6-gigabyte footprint. It accepts a PDF, PowerPoint or JPEG page as a base64-encoded image and returns Markdown in reading order, with tables rendered as HTML, image descriptions and bounding box coordinates for tables and images. Language coverage. Arabic, English, French, German, Italian, Japanese, Korean, Portuguese and Spanish get stable accuracy, with lower-accuracy zero-shot support elsewhere. Output modes. The default output returns a Markdown string per page. A blocks mode returns typed elements, where each table carries its own HTML, bounding box and description, the format Cohere positions as what makes citation-level traceability possible for agents. Availability. Parse 5 is generally available now through the Cohere API , Model Vault, Microsoft Foundry and AWS SageMaker. The benchmark shows a trade-off, not a win ParseBench is a benchmark that scores document-parsing tools against human-verified enterprise pages. Cohere reports Parse 5 scoring 79.2 across three dimensions: tables, content faithfulness and semantic formatting. That puts Parse 5 behind GPT-5.5 (84.4), Opus 4.8 (84.3) and Gemini 3.5 Flash (81.8), and ahead of LlamaParse's Cost Effective tier (78.3), Mistral OCR 4 (74.5), Databricks AI Parse (72.4) and Azure Document Intelligence (69.3). Cohere's table notes two excluded dimensions, Layout and Chart, and attributes both to product scope rather than a performance gap. Parse 5 returns reading-order Markdown instead of per-element bounding boxes for text, and describes charts rather than extracting their underlying data, with chart-data extraction planned for a future version. Reimers said that design choice reflects where agentic workflows actually break. "For charts, for example, we provide a general description of the chart together with an indicator, how Agentic AI can visually inspect the chart," Reimers explained. "Other solutions try to extract the data from the chart, but then miss out critical information (for example, the color or the pattern of a line) that leads to hallucinations in Chat and Agentic AI applications." Cost is where Cohere makes its real case. Reimers pointed to a workflow the company modeled for a large financial services firm. "We ran the numbers for a large financial services workflow that processes 750 million documents a year and showed that choosing Parse 5 over a large general‑purpose model like GPT‑5.5 would reduce costs by more than 98 percent." That figure is Cohere's own estimate for a single modeled workflow, not an audited deployment. Where Parse 5 sits against the field There is no shortage of options for enterprises looking at parsing solutions. General-purpose frontier models, GPT-5.5, Opus 4.8 and Gemini 3.5 Flash, top the accuracy comparison but carry the cost and latency of a large model on every page. Then there are specialized parsers, including Mistral OCR 4 , LlamaParse and open-weight options like Chandra OCR 2 and RedNote's dots.mocr. Hyperscaler document intelligence services, AWS Textract, Google Document AI, Azure Document Intelligence and Databricks AI Parse , compete more on ecosystem convenience than on raw parsing quality, and score lowest in Cohere's own comparison. Kevin Petrie, VP of Research at BARC US, said document parsing sits at the center of enterprise AI adoption right now. "We're completing a survey now that shows document analysis is by far the #1 use case for AI, with 62% adoption rates among organizations we polled," Petrie told VentureBeat . "Documents and other unstructured objects, including images and so on, hold the proprietary context that organizations need to differentiate their agentic AI initiatives." Petrie added that only time will tell how Cohere's cost-performance stacks up against frontier models, but strategically his view is that Cohere has the right focus. Stephanie Walter, Practice Leader for AI Stack at HyperFRAME Research, sees Cohere Parse 5 as sitting in a good spot between legacy OCR and using an expensive frontier model on every page. "Its potential advantage is delivering structure, spatial provenance and private deployment at a price suitable for high-volume ingestion," Walter told VentureBeat . "It does not need to win every benchmark. It needs to make reliable enterprise-scale parsing economical." The real test is downstream, not on the benchmark "Parsing is the first quality gate in the enterprise AI stack," Walter said. "If tables, headings, images, or reading order are lost during ingestion, better embeddings and larger models cannot recover that missing structure." A benchmark score isn't the only input that matters here. "Enterprises should test parsers against their own most difficult documents and measure downstream retrieval and task accuracy, not how clean the extracted text looks," Walter said. "The right question is not 'Did it read the PDF?' but 'Can the agent now use the information correctly?'"
Score: 58🌐 MovesAug 28, 2026https://venturebeat.com/data/cohere-parse-5-loses-the-benchmark-on-points-it-wins-on-cost-per-page - AI Agent Payment Risks Catch PBOC’s Eye
AI Agent Payment Risks Catch PBOC’s Eye Caixin Global
Score: 58🌐 MovesAug 28, 2026https://www.caixinglobal.com/2026-08-28/ai-agent-payment-risks-catch-pbocs-eye-102478958.html - The new management playbook for AI: How to move faster and create more value
Scale is where the value is. Enabling scale depends on how strong your foundations are.
- China’s daily AI token usage tops 500 trillion as compute demand grows
China’s daily AI token call volume exceeded 500 trillion as of June 2026, as large-model applications expanded across industries. The figure measures aggregate model processing activity rather than the number of users or models. Industry representatives said model update cycles have shortened from roughly three months to four to six weeks. Agent workflows also repeatedly […]
Score: 58🌐 MovesAug 28, 2026https://technode.com/2026/08/28/chinas-daily-ai-token-usage-tops-500-trillion-as-compute-demand-grows/ - MiniMax expands Alibaba cloud pact as compute needs surge for training and inference
Chinese artificial intelligence firm MiniMax has raised the three-year purchase ceiling on its cloud computing deals with Alibaba Group Holding by 220 per cent to US$1.2 billion, underscoring surging demand for computing power among the country’s top AI developers. The Shanghai-based company planned to spend up to US$300 million on Alibaba Cloud services this year, nearly triple its original US$115 million cap, after burning through two-thirds of its budget by the end of June, according to its...
- Ant Group launches finance-tuned Ling model, plans to open-source it next week
Ant Group has launched the finance-tuned Ling-3.0-flash-Fin model, retaining the Ling-3.0-flash architecture and its 124 billion total parameters and 5.1 billion active parameters. The model weights are scheduled to be released next week. Ling-3.0-flash-Fin is optimized for annual reports, financial workbooks and multi-document research, as well as information retrieval, investment analysis, valuation modeling and banking […]
Score: 57🤖 ModelsAug 28, 2026https://technode.com/2026/08/28/ant-group-launches-finance-tuned-ling-model-plans-to-open-source-it-next-week/ - Gnani AI launches Artha sovereign AI stack with 30-billion-parameter Evon 3.3
Gnani AI launched Artha, a sovereign AI stack for Indian companies and public institutions. This stack features the Evon 3.3 language model and the Plexus agentic platform. Evon 3.3 is an open-weights model trained on Indic languages and domain-specific data. The Artha stack allows organisations to run AI models within their own infrastructure.
- Georgia PSC Approves Secret OpenAI Contract
ATLANTA — After a tumultuous and secretive process, the Georgia Public Service Commission staff approved the contract between Georgia Power and OpenAI to power a 3.2 gigawatt hyperscale data center planned for Effingham County — without a direct vote from the commission itself. In response, Sierra Club and the Southern Alliance ... [continued] The post Georgia PSC Approves Secret OpenAI Contract appeared first on CleanTechnica .
Score: 57🌐 MovesAug 28, 2026https://cleantechnica.com/2026/08/27/georgia-psc-approves-secret-openai-contract/ - Google Flow just got much better at turning ideas into finished videos
New controls make AI-generated videos easier to shape and polish.
Score: 57🌐 MovesAug 28, 2026https://www.androidauthority.com/google-flow-gemini-omni-1-1-flash-video-update-3704288/ - Claude Memory Now Follows You Into Cowork — Here’s How to Turn It Off
Claude now shares memory across Chat and Cowork. Here’s how to pause, reset, or limit what Claude remembers and what controls users still lack. The post Claude Memory Now Follows You Into Cowork — Here’s How to Turn It Off appeared first on TechRepublic .
- Meta makes AI glasses slightly less creepy with limit on nonconsensual recording
Meta fixes AI glasses to stop recording any time users cover up the safety light.
- X busts 200,000-strong Chinese bot farm, including accounts making claims about AI data centers and electricity — suspect accounts posted claims about pricing and grid strain to 'manipulate' debate
The X Safety Team said that at least 200 bot accounts have been making posts to influence public opinion data centers and energy policy. The accounts share links to legitimate news stories and then add their own spin designed to inflame emotions.
- Think You’ve Eliminated Chinese AI? Check the Model’s Lineage, Cisco Says
New research shows that country-of-origin labels can obscure an AI model’s upstream dependencies, inherited behaviors and potential security risks. The post Think You’ve Eliminated Chinese AI? Check the Model’s Lineage, Cisco Says appeared first on SecurityWeek .
Score: 56🌐 MovesAug 28, 2026https://www.securityweek.com/think-youve-eliminated-chinese-ai-check-the-models-lineage-cisco-says/ - How to accelerate AI adoption without creating unnecessary security risk
Why sustainable AI adoption depends on stronger governance, verification, and data discipline.
Score: 56🌐 MovesAug 28, 2026https://www.techradar.com/pro/how-to-accelerate-ai-adoption-without-creating-unnecessary-security-risk - Large language models as uncertainty-calibrated optimizers for experimental discovery
Nature Machine Intelligence, Published online: 28 August 2026; doi:10.1038/s42256-026-01283-z Although language models can be helpful in molecular design, they are not typically calibrated for uncertainty. Rankovic and colleagues present a method to train language models while taking into account the uncertainty of the data.
- Weekly Must-Read: China Tech Firms Drive AI Data Center Boom in Southeast Asia
Weekly Must-Read: China Tech Firms Drive AI Data Center Boom in Southeast Asia Caixin Global
- AI tests 40 of 100 million ways to 3D-print a NASA rocket alloy & found 6 that worked
Washington State University researchers used artificial intelligence to find better printing methods. This advance significantly reduces the cost and time for printing a critical metal alloy. The AI strategy successfully identified printable configurations for the GRCop-42 alloy. This innovation makes the alloy accessible for use on common commercial 3D printers. The AI approach could also accelerate discoveries in other scientific fields.
- Anthropic's new AI tool can conduct scientific experiments: How it works
Anthropic has opened a research preview of its Model Hardware Standard, allowing AI agents such as Claude to control scientific equipment, automate experiments and manage workflows across labs
- ‘We go big’: Jon Gray on how Blackstone places its AI bets
The $1.3 trillion firm is staying bullish but investing “through the lens that this may change.”
Score: 55🌐 MovesAug 28, 2026https://www.semafor.com/article/08/27/2026/we-go-big-jon-gray-on-how-blackstone-places-its-ai-bets - Notion Plans Hiring Binge as CEO Makes All-In Bet on AI
Notion Plans Hiring Binge as CEO Makes All-In Bet on AI The Information
Score: 55🌐 MovesAug 28, 2026https://www.theinformation.com/articles/notion-plans-hiring-binge-ceo-makes-bet-ai - Always-on and self-starting AI agents might be OpenAI's next big play
OpenAI is building a "Persistent Mode" for its AI agent Codex that stays active indefinitely and generates its own follow-up tasks. WIRED found the relevant code, and OpenAI confirmed the tests. The feature comes with risks, though. With GPT-5.6 Sol, persistent behavior already led to unwanted actions, like deleting user data. The article Always-on and self-starting AI agents might be OpenAI's next big play appeared first on The Decoder .
Score: 55🌐 MovesAug 28, 2026https://the-decoder.com/always-on-and-self-starting-ai-agents-might-be-openais-next-big-play/ - How startups in regulated industries are using AI
The post How startups in regulated industries are using AI appeared first on Source .
- Temasek-backed investor behind Unitree pitches Singapore as Chinese robots' path to U.S.
As Washington shuts out Chinese robotics, a Temasek-backed VC bets that startups with genuine development in the city-state can still reach American buyers.
Score: 55🌐 MovesAug 28, 2026https://www.cnbc.com/2026/08/26/temasek-unitree-humanoids-robotics-trump-ban-.html - IndiGo steps up AI focus, announces investment in Sarvam
The funding, made through IndiGo Ventures during Sarvam’s Series B round, aims to leverage advanced AI capabilities tailored for the aviation industry
- How finance leaders can close the AI trust gap
Most finance leaders at large organizations have made the right investments. A modern ERP, cloud data platforms, planning tools, and more. And now, increasingly, AI — for forecasting support, anomaly detection, close acceleration, and reporting at scale. The technology stack looks right. But when the board starts asking about results, the returns are harder to point to than the investments were. What your ERP was built to do — and what it wasn’t Your ERP is excellent at what it was designed for: capturing transactions, enforcing accounting standards, managing the chart of accounts. It is the system of record, and it performs that job well. But it doesn’t encode how your organization has decided to handle intercompany eliminations across a complex entity structure. It doesn’t carry your FP&A team’s cost allocation methodology, refined over three budget cycles. It doesn’t know what variance threshold triggers a controller review versus a VP escalation, or how your tax team has mapped jurisdictions for Pillar Two. That logic — specific, documented, organization-defined — isn’t in your ERP. It’s not in your data warehouse either. For most finance organizations, it lives in spreadsheets. Sometimes in the heads of the people who built them. Where AI runs into trouble in finance There’s a finding that gets cited a lot in finance AI conversations: research from MIT found that 95% of organizations are seeing no measurable return on their gen AI investments. Bain & Company looked at the same picture and reached a different conclusion for finance specifically. The fastest payback from AI in finance comes from embedding it in workflows — not from running pilots. The distinction matters because it explains why so many finance AI efforts stall after the proof of concept. AI can process data at speed and surface patterns across large datasets. What it cannot do is infer your business logic from raw inputs. Without that context, AI outputs in finance look confident but aren’t defensible — and in a function where auditability is a baseline requirement, that gap is not a minor limitation. It validates that trustworthy AI is critical for scaling workflows and AI pilots. Our own survey of 1,400 IT and business leaders asked what their biggest barriers to success with AI workflows were. One in two (49%) said inaccurate or biased outputs. Further, 38% said it was a reluctance to allow AI to make decisions without human oversight. While you don’t need perfect data to start using LLMs, you absolutely need trustworthy data. The layer that’s actually missing The gap between your ERP and your AI ambitions isn’t a data gap. It’s a business logic gap — the layer where your organization’s specific rules, methodologies, and decision criteria live, and where AI needs to operate to produce outputs you can stand behind. When that layer is built correctly — logic documented, workflows repeatable, outputs traceable — AI has validated, structured inputs rather than raw data it has to interpret. Outputs can be explained to auditors and to the board. And the sequencing question resolves itself: getting the process right is how you adopt AI. What it takes to build that layer Closing the gap takes more than a mandate to “use AI responsibly.” It takes three specific things, built and owned inside finance rather than handed off to IT. A purpose-built data asset for each process. Not another warehouse but a narrow, well-defined data set scoped to one process that reflects how your team measures it, not just what your ERP happens to store. Encoded logic, not tribal knowledge. The allocation methodology or the variance threshold that triggers escalation — built into a repeatable workflow instead of a senior analyst’s spreadsheet. The shift is building it once; in a form AI can use. A way to update it when the business changes. Comp plans get revised, tax jurisdictions shift, and the chart of accounts gets restructured after an acquisition. Logic that can only be changed by submitting a ticket to IT will be stale before it’s deployed — the people who own the process need to be the ones who can adjust the rule. None of this requires waiting for a perfect architecture. The highest-value starting point is whatever process has your analysts fielding the same question, the same way, every single cycle. Encode that one workflow first, connect it to the AI tools your team is already using, and the logic compounds from there: the same governed calculation that answers one controller’s question can feed the scenario model that runs your next planning cycle. To learn more, visit us here .
Score: 54🌐 MovesAug 28, 2026https://www.cio.com/article/4213444/how-finance-leaders-can-close-the-ai-trust-gap.html - Chinese automakers are following Tesla’s bet that robots are the next big profit machine
Technical progress has encouraged a new batch of companies to jump in on the promise of profits from humanoid robots. And they're all Chinese automakers.
- Northeast Ohio Hospitals Test AI to Improve Care, Cut Costs
The technology is swiftly taking hold in hospital exam rooms, doctors’ offices and health insurance systems in the northeast — shaping patient treatment, scheduling and how medical records are written.
Score: 53🌐 MovesAug 28, 2026https://www.govtech.com/artificial-intelligence/northeast-ohio-hospitals-test-ai-to-improve-care-cut-cost - Claude's new memory feature solves my biggest AI frustration — here are 5 prompts to try first
Claude's new memory feature solves my biggest AI frustration — here are 5 prompts to try first Tom's Guide
- IndiGo Ventures invests in Sarvam to enhance Artificial Intelligence in aviation
IndiGo Ventures is investing in Sarvam, an Indian AI company. This investment supports Sarvam's Series B funding round. The partnership aims to enhance customer experience and operational workflows. Both companies will develop AI applications for a modern airline infrastructure. This collaboration seeks to accelerate innovation and maximize stakeholder value.
- Meet Microduck, Hugging Face’s new open-source AI robot that can learn new tricks
Meet Microduck, Hugging Face’s new open-source AI robot that can learn new tricks
- OPM memo attempts to clear path for more AI use in federal hiring
The human capital agency said AI uses that support applicant evaluation and pre-offer quality control likely don’t fall into a category that requires more risk management scrutiny. The post OPM memo attempts to clear path for more AI use in federal hiring appeared first on FedScoop .
- US surgical robotics firm eyes Hong Kong listing to power push into mainland China
SoftBank-backed surgical robotics company Noah Medical has begun preparations for a Hong Kong initial public offering, joining a wave of cutting-edge technology firms tapping the city’s international fundraising pool. The US-headquartered company aimed to raise more than US$100 million and planned to file its listing application as early as next year, according to founder Zhang Jian. “We want to expand our business in [mainland] China, and Hong Kong – as an international financial centre with a...
- AI Has Human Doctors Asking: What’s Left for Us?
A recent paper argues that AI is often better at doctoring than doctors. Guess who isn't thrilled.
Score: 52🌐 MovesAug 28, 2026https://www.wired.com/story/ai-has-human-doctors-asking-whats-left-for-us/ - Everything we know about Z.ai, the Chinese company behind the mysterious Ox Alpha model
Everything we know about Z.ai, the Chinese company behind the mysterious Ox Alpha model Business Insider
Score: 52🌐 MovesAug 28, 2026https://www.businessinsider.com/what-is-ox-alpha-ai-model-openroute-opencode-z-ai-2026-8 - Hugging Face says sales for its robot ducks topped $2.6 million in 24 hours. Now there's a backlog.
Hugging Face says sales for its robot ducks topped $2.6 million in 24 hours. Now there's a backlog. Business Insider
Score: 52🌐 MovesAug 28, 2026https://www.businessinsider.com/hugging-faces-duck-robot-hits-sales-roller-skate-2026-8 - Labour rejects Zack Polanski’s call to ‘slam brakes’ on building AI datacentres
Government says Green leader’s proposed moratorium on ‘energy-guzzling’ AI projects would be disaster for economy Labour has rejected calls to pause construction of major AI infrastructure in Britain after the Green party leader, Zack Polanski, said it was time to “slam the brakes on these energy-guzzling, water-guzzling datacentres”. The government hit back at the opposition party’s proposal of a “moratorium other than [for] local-scale datacentres for the local community”, saying it “would be a disaster for jobs and national security”. Continue reading...
Score: 52🌐 MovesAug 28, 2026https://www.theguardian.com/uk-news/2026/aug/28/labour-green-party-zack-polanski-ai-datacentres - Baidu launches DuMateBench benchmark for real-world AI agent delivery
Baidu has launched DuMateBench, an evaluation leaderboard designed to measure whether AI agents can complete real-world tasks and deliver usable outputs, rather than only generate answers. The benchmark includes more than 200 office tasks across six categories and tests agent performance in complex operating environments. DuMateBench evaluates task understanding, tool use, continuous execution and final-result […]
Score: 52🌐 MovesAug 28, 2026https://technode.com/2026/08/28/baidu-launches-dumatebench-benchmark-for-real-world-ai-agent-delivery/