Follow the developments that change your market view.

Alibaba Launches Qwen Intelligence for Agentic Smartphones

Alibaba has unveiled Qwen Intelligence, a new platform designed to power agentic AI on smartphones. The platform enables device manufacturers to develop intelligent assistants capable of planning and executing complex tasks across multiple apps. HONOR is the first partner to integrate Qwen Intelligence, with the HONOR Magic9 Series and the upcoming HONOR Robot Phone among the first devices to feature it. The platform includes three pre-configured agents: Mobile Planner, Mobile-Use, and Mobile Creative. It is built on three layers: optimized Qwen models, a modular platform for deployment and tool control, and a device-cloud architecture. Alibaba plans to add personalized memory, multimodal interactions, and specialized agents from partners. The company has also developed four benchmarks, including MobilePA-Bench with over 1,000 tasks, to evaluate mobile agent performance. HONOR and Alibaba report up to 91.8% task precision and orchestration of over 100 steps, with first devices launching end of September.

• •Alibaba launched Qwen Intelligence, a platform for agentic AI on smartphones.
• •HONOR is the first partner, integrating Qwen Intelligence into the Magic9 Series and Robot Phone.
• •The platform includes three agents: Mobile Planner, Mobile-Use, and Mobile Creative.

Amazon Launches US Beta of AI-Powered Seller Assistant Plugin

Amazon is expanding its Seller Assistant into a persistent AI agent for merchants, introducing new capabilities such as a permanent memory for business context, autonomous workflows for recurring tasks, and a new plugin that integrates with Amazon Quick and Anthropic's Claude. The Seller Assistant, built on Amazon Bedrock and using Claude models, now serves over 90% of global sellers in their respective languages and reaches hundreds of thousands of active users. Merchants can define workflows in natural language to automate tasks like inventory monitoring, price changes, and account checks, with options for recommendations or executed actions, all logged and controlled by the seller. The plugin, initially in beta for US sellers, allows merchants to access sales data, inventory, and performance metrics directly within Quick and Claude, and to trigger actions within their Amazon business. Amazon is also offering a free 12-month Quick Plus subscription for account holders. This move positions Amazon as a provider of agentic infrastructure for merchant operations, with plans for further integrations with other AI platforms.

• •Amazon's Seller Assistant will gain a persistent memory and autonomous workflows, transforming it into a continuously working agent for sellers.
• •Amazon introduces a new Selling-Partner Plugin for Seller Assistant, initially in beta in the US, integrating with Amazon Quick and Anthropic Claude.
• •The Seller Assistant is built on Amazon Bedrock and uses Claude models from Anthropic, and is now available to over 90% of global selling partners.

NZZ ordnet kommerzielles Geschäft neu – Corine Blesi leitet den neuen Bereich NZZ Experience und wird Mitglied der Geschäftsleitung

Künftig wird der Nutzermarkt mit dem Abonnementgeschäft eigenständig geführt und Werbemarkt, Veranstaltungen, Partnerschaften und weitere kommerzielle Angebote werden im Bereich NZZ Experience gebündelt. Beide Bereiche sind direkt in der Geschäftsleitung...

OpenAI Testing Third-Party Tracking in ChatGPT Ads

OpenAI is reportedly testing a third-party-style tracking mechanism in its ChatGPT advertising ecosystem. An independent researcher found that a cookie named '__obi' can be set by ChatGPT and persist for up to a year. When the user later visits websites that use OpenAI's advertising pixel, the same identifier is transmitted back to OpenAI along with page and conversion data. This enables cross-site attribution of ad interactions to subsequent user activity. The cookie is classified by OpenAI as an analytics cookie, which raises concerns about consent, as it may be set even when users decline marketing cookies. For marketers, this could improve conversion measurement in ChatGPT ads, but it also raises privacy and regulatory questions, especially given the sensitive personal data shared with ChatGPT.

• •OpenAI is testing a third-party-style cookie called '__obi' for ChatGPT ad tracking.
• •The cookie can persist for up to a year and is sent to OpenAI when visiting sites with its ad pixel.
• •The cookie may be set even when analytics consent is given but marketing cookies are declined.

Marketo Co-founder Launches AI Marketing Automation Platform Phave

Jon Miller, co-founder of Marketo, has launched Phave, a new AI-powered marketing automation platform for enterprise B2B organizations. After two years in stealth, Phave is now used by 10 companies. The platform replaces traditional rules-based triggers with AI reasoning models to handle complex B2B buying scenarios. Phave treats contacts, accounts, and buying groups as first-class objects, and introduces a 'playlist' concept that dynamically sequences marketing tactics based on real-time data and preferences. It also exposes 319 tools through the Model Context Protocol (MCP), allowing external AI agents to trigger actions. Miller aims to address the limitations of legacy MAPs, which he says have failed to adapt to modern B2B buying behavior.

• •Jon Miller launched Phave, an AI-powered B2B marketing automation platform.
• •Phave has been in stealth mode for two years and now powers marketing operations for 10 companies.
• •Phave replaces rules-based engines with AI reasoning models to handle ambiguity in B2B buying.

EU Commission to Scrutinize Uber-Delivery Hero Deal

The European Commission's DG COMP is preparing to scrutinize Uber's planned acquisition of Delivery Hero, a deal valued at around €13 billion. The review will consider whether to assess the main acquisition and the parallel sale of Delivery Hero's 14 country businesses to SSW Partners jointly or separately. Notably, Uber is lending SSW Partners most of the €1.4 billion needed for this purchase, raising concerns about market overlap and competitive impact. The deal has been structured to avoid antitrust issues, with Delivery Hero divesting in several European markets. However, some industry players are urging Brussels to examine both transactions together. The formal review is pending, with pre-notification talks underway. The deal is expected to close in the second half of 2027, subject to regulatory approvals in multiple jurisdictions, including South Korea and the Middle East.

• •Uber plans to acquire Delivery Hero for approximately €41.50 per share, valuing the company at around $14.8 billion.
• •Delivery Hero is selling its business in 14 countries to SSW Partners for around €1.4 billion.
• •Uber has agreed to lend SSW Partners funds to finance the majority of the SSW transaction.

Kroger AI Shopping Assistant Drives Larger Baskets

Kroger reported that customers using its AI-powered shopping assistant are adding more items to their carts, contrary to expectations. The tool, launched in July across Kroger's websites and apps, helps with meal planning, recipe discovery, and cart building based on budgets and dietary needs. Yael Cosset, Kroger's EVP and Chief Digital Officer, discussed the basket size growth at Groceryshop 2026. Kroger is integrating more personalization features, leveraging its loyalty data from over 63 million households. The retailer also uses AI for demand forecasting and operational functions. Kroger partners with Google Cloud for Gemini Enterprise to power the assistant. The article also mentions similar results at Albertsons and increased conversion from AI-referred traffic per Adobe Digital Insights. Kroger's retail media business, Kroger Precision Marketing, is integrating sponsored products within the assistant.

• •Kroger's AI shopping assistant is driving larger basket sizes, though the exact increase is undisclosed.
• •The assistant was rolled out across Kroger's websites and apps in July.
• •Yael Cosset discussed the basket increase at Groceryshop 2026.

Paramount Legal Hurdle Cleared; Amazon Bars Meta's Muse

The article is a digest covering two major ad tech news items. Paramount Global has cleared a legal hurdle, likely an antitrust or regulatory challenge, in its proposed merger with Warner Bros. Discovery, setting the stage for the combination of the two media giants. In a separate development, Amazon has barred Meta's AI-powered shopping agent, Muse, from its platform, preventing the AI agent from accessing Amazon's retail ecosystem for agentic shopping tasks. This move highlights the growing tensions and strategic decisions as AI agents begin to play a larger role in e-commerce and advertising.

• •Paramount Global cleared a legal hurdle ahead of its merger with Warner Bros. Discovery.
• •Amazon has barred Meta's AI agent, Muse, from its platform.
• •The action prevents Muse from performing agentic shopping on Amazon.

Trump-Xi Summit Discusses AI Safety, Chips Remain Sticking Point

Ahead of the Trump-Xi summit, U.S. and Chinese officials are exploring an AI safety dialogue, including a channel for national security-related incidents. Treasury Secretary Scott Bessent discussed the proposal with Chinese counterpart He Lifeng. Both countries acknowledge risks from autonomous AI models, but neither is willing to slow development. Export controls on advanced Nvidia chips and accusations of AI 'distillation' remain contentious. The U.S. restricts China's access to leading chips, while China seeks relaxation of controls. Analysts see a modest step in establishing incident communication, but agreements on slowing AI progress are unlikely. Verification mechanisms are seen as crucial for any cooperation.

• •U.S. Treasury Secretary Scott Bessent discussed a 'U.S.-China AI dialogue' with He Lifeng over the weekend.
• •The dialogue could include an emergency communication channel for national security-related AI incidents.
• •Washington continues to restrict China's access to Nvidia's most advanced AI chips.

Telefónica España convierte los dispositivos en una nueva unidad dentro de B2C e incorpora a Alberto Ruano para liderarla

Telefónica España reorganiza su área de dispositivos como una unidad independiente dentro del negocio B2C, nombrando a Alberto Ruano como líder. Además, la compañía anuncia la venta del edificio de Gran Vía 28 y nuevas alianzas como socio oficial del Espai Barça.

Uber's Delivery Hero Acquisition Faces EU Antitrust Scrutiny

The European Commission's Directorate-General for Competition (DG COMP) is examining Uber's planned acquisition of Delivery Hero, a deal valued at around €13 billion. The scrutiny focuses on whether to assess the main acquisition and the parallel sale of Delivery Hero's operations in 14 countries to SSW Partners jointly or separately. Uber is financing a large portion of SSW's €1.4 billion purchase of those markets. Industry players are lobbying regulators, and MEP Andreas Schwab has called for careful assessment. The deal involves Uber offering €41.50 per share for Delivery Hero, with an equity value of approximately $14.8 billion, and is expected to close in the second half of 2027. Regulatory approvals are required in several jurisdictions, including South Korea, the Middle East, and Latin America, with concerns remaining in Spain.

• •European Commission's DG COMP is scrutinizing Uber's ~€13 billion acquisition of Delivery Hero, potentially assessing the main deal and SSW Partners sale jointly.
• •Uber is offering €41.50 per share for Delivery Hero, equating to an equity value of ~$14.8 billion.
• •Delivery Hero is selling its business in 14 countries to SSW Partners for ~€1.4 billion, financed largely by Uber.

AI Overviews Impact on German E-Commerce SEO Revealed

A study by LEAP and SISTRIX analyzing 2,496 top German online shops reveals that organic visibility in e-commerce has declined for the first time in years, with a median drop of 16.7%. The share of searches with AI Overviews tripled from 4.3% to 14.2% in a year. AI Overviews appear in 69.7% of informational searches but only 0.2% of transactional ones, indicating the disruption is currently at the research stage, not purchase. However, this may change as Google tests AI Mode and the Universal Commerce Protocol. Top organic shops capture 72% of AI citations, and only 9.3% of informational AIO cases include the shop. The decline in organic traffic share (from 42.5% to 33.3%) translates to an estimated €7.5 billion in lost reach.

• •Organic visibility in German e-commerce declined by 16.7% year-over-year.
• •AI Overviews now appear in 14.2% of search queries, up from 4.3%.
• •AI Overviews are present in 69.7% of informational searches but only 0.2% of transactional searches.

SPRIND and NADI Launch €40M AI-Native Chip Design Challenge

Germany's SPRIND and the Netherlands' NADI have launched a joint €40 million program, the AI-Native Chip Design Challenge, to accelerate European chip design and strengthen AI sovereignty. The initiative aims to cut development timelines from years to weeks by leveraging AI-driven design methods and combining Germany's industrial strength with Dutch semiconductor expertise. Running in two stages over 20 months, the first stage will fund seven teams with €2.6 million each, and the second will shortlist three teams for €7 million each, with additional technical and commercial coaching. The program seeks to produce production-ready designs for AI training and inference, enhance energy efficiency, and reduce Europe's dependence on US and Chinese semiconductors. This is the first collaborative funding challenge between the agencies, with NADI formally announced on 15 September 2026.

• •SPRIND and NADI commit €40 million to the AI-Native Chip Design Challenge.
• •The challenge aims to cut chip development time from years to weeks.
• •Stage 1 selects seven teams (€2.6 million each); Stage 2 shortlists three teams (€7 million each).

Publishers Score Google's Ad Tech Remedies 5/10

Publishers and sell-side executives gave a mixed scorecard, averaging 5/10, to the remedies proposed in the DOJ's antitrust case against Google's ad tech dominance. They acknowledge progress, particularly in opening AdX to rival ad servers and banning first-look advantages, but express skepticism about practical impact and enforcement. The opening of AdX to non-Google ad servers received the highest average score (6/10), seen as a potential shift, though caveats remain about Google's own demand participation. Other remedies, like data portability and unbundling AdX/DFP, scored lower, seen as transparency or billing changes rather than true competitive resets. Concerns persist about the hollowed-out alternative ad server market and risks for publishers switching stacks.

• •Seven publisher revenue and sell-side executives rated Google's ad tech remedies, assigning an average total score of 5/10.
• •The remedy to open AdX bidding to rival ad servers received the highest average score of 6/10.
• •A ban on Google's 'first look' and 'last look' auction advantages scored 6/10, though Google claims it already stopped these practices in 2019.

Discord Rolls Out Age Verification Despite Privacy Backlash

Discord is rolling out age verification measures to all users starting Wednesday, following a delay from February due to privacy concerns. The company will use existing account data and machine learning to estimate users' age groups—adult, teen, or unconfirmed—to comply with expanding age assurance laws, including in Brazil and Texas. More than 90% of users will not need to provide biometrics or ID; instead, Discord analyzes signals like account tenure and usage patterns. Users who cannot be confidently classified can appeal using credit card or app store data. Teens will face restrictions on certain servers and communication features. The rollout follows a previous data breach affecting 70,000 users, raising concerns about new privacy risks.

• •Discord begins age verification rollout to all users on September 23, 2026.
• •Discord uses machine learning on account data (tenure, activity) to estimate age groups, without collecting new data.
• •Users not confidently classified as adults can verify via credit card or Apple/Google Play account data.

Anthropic launches Claude Opus 5.5 for coding and knowledge work

Anthropic has launched Claude Opus 5.5, its new flagship model designed for complex coding, knowledge work, and long-running agentic tasks. It is available on the Claude Platform, AWS, Google Cloud, and Microsoft Azure. Opus 5.5 offers improved performance, being up to 40% cheaper per task and over 30% faster than Opus 5. Token prices have been reduced to $4 per million input and $20 per million output tokens. The model can handle large codebases, completing a 200,000-line repair in three hours compared to over 20 hours for Opus 5. Enhanced security features include resistance to prompt injection, an action classifier, and Preserved Thinking to combat distillation attacks. It has achieved record alignment scores in Anthropic's Behavioral Audit. Sonnet 5.5 and Haiku 5.5 are expected in the coming weeks.

• •Claude Opus 5.5 is available on AWS, Google Cloud, and Microsoft Azure.
• •Opus 5.5 is up to 40% cheaper and over 30% faster than Opus 5.
• •Token prices: $4 per million input and $20 per million output.

Haatch exits Trigify to HubSpot at up to 3.53x

We are delighted to announce the successful exit of our full position in Trigify following its acquisition by HubSpot, the agentic customer platform for scaling businesses. The all-cash transaction delivered up to 3.53x for our investors. This exit represents Haatch’s 6th profitable exit in the last 18 months.

VAUNET Criticizes German Media Policy Implementation

At its annual get-together in Berlin, German private media association VAUNET celebrated but also criticized the political landscape. Chairman Claus Grewenig highlighted a 'grave implementation problem' in digital regulation, noting that 72% of digital ad revenues go to global Big Tech platforms. He criticized new legislative burdens instead of relief for private media, advocated for eased media merger rules, and urged faster implementation of the Digitale-Medien-Staatsvertrag (DMStV). Rhineland-Palatinate State Secretary Torsten Welling, who coordinates the Rundfunkkommission, defended the DMStV draft as a 'media policy masterpiece' aiming to create fair rules and address media concentration in the digital age. The event also saw a debate between Grewenig and Welling on platform regulation versus deregulation of private media.

• •VAUNET forecasts 72% of German digital ad revenues will go to global Big Tech platforms in 2026.
• •Claus Grewenig, VAUNET chairman, criticized new legislative burdens on private media during the get-together.
• •The Digitale-Medien-Staatsvertrag (DMStV) draft is to be presented for public hearing in early October by the Rundfunkkommission.

Google Launches Googlebook Laptop with Adaptive App Development

Google announced Googlebook, a new laptop category built on Android, featuring partner hardware from HP, Dell, Lenovo, Acer, and Asus. The device integrates Google Play optimization, with dedicated badging for optimized apps. Developers are encouraged to use adaptive development to scale their apps from phones to desktop, leveraging window size classes, multi-pane layouts, and multi-instance support. The post details technical features like Navigation 3, Scene decorators, and Continue On for cross-device handoffs. Google Play will highlight optimized apps, and developers can join the Apps Experience Program for new rate cards. Android Studio Canary includes a desktop emulator for testing, and an adaptive skill assists AI agents in refactoring layouts. The announcement emphasizes a single codebase to optimize across form factors.

• •Google announced Googlebook, a new Android-based laptop category, with partners HP, Dell, Lenovo, Acer, and Asus.
• •Google Play will feature optimized apps with badging, enhanced search, and featured spots.
• •Developers are encouraged to use adaptive development to scale apps across phones, foldables, tablets, and laptops.

Anthropic Unveils Opus 5.5; OpenAI Launches GPT-6 Sol and Luna

In a major AI model release day, Anthropic introduced Opus 5.5, a refreshed flagship model with enhanced safety guardrails and improved conversational tone, while OpenAI launched two new models: GPT-6 Sol (a faster, cheaper daily driver) and GPT-6 Luna (a lighter, more efficient variant). The new models focus on cost reduction, speed, and token efficiency, with significant improvements in caching. A blind taste test conducted by a tech reviewer evaluated the models across multiple tasks including front-end coding, creative SVGs, and agentic workflows. The reviewer found that OpenAI models (Astra and Sol) excelled in user-friendly interactions and creative illustrations, while Opus 5.5 won the week for its broad consistency, particularly in agentic tasks and long-running research. The review also highlighted ongoing differences in model behavior and the importance of optimizing cache usage for cost savings.

• •Anthropic released Opus 5.5, a refreshed flagship model with new safety guardrails (cyber and bio) and improved user interaction.
• •OpenAI launched GPT-6 Sol and GPT-6 Luna, both positioned as faster and cheaper daily-driver models.
• •The reviewer's blind taste test ranked Opus 5.5 as the most consistently high-performing model across a wide range of tasks (won the week).

Announcing Claude Opus 5.5 on Snowflake Cortex AI

Snowflake announces the availability of Claude Opus 5.5 on Cortex AI, a major product launch enhancing AI capabilities.

Snorkel AI raises $350M Series E, triples valuation to $3.5B

Snorkel AI, a startup providing AI training data and simulated environments, has raised a $350 million Series E round at a $3.5 billion valuation, nearly tripling its previous $1.3 billion valuation from 17 months ago. The round was led by Insight Partners and S32, with participation from existing investors Addition, Lightspeed, Greylock, GV, and Wells Fargo. Snorkel has shifted from data-labeling software to offering complete datasets and reinforcement learning environments. The company reports an annualized revenue run rate of $375 million, an eighteenfold increase over the past year, driven by high demand from AI labs. The funding will support scaling its data-as-a-service offering.

• •Snorkel AI raised $350 million in a Series E round.
• •The Series E was led by Insight Partners and S32.
• •Snorkel AI's valuation is now $3.5 billion, up from $1.3 billion.

Introducing GPT-6 Sol and Luna

OpenAI announces the launch of GPT-6 Sol and Luna, a major product release. Also introduces Astra for Law, a new product for the legal sector, and expands OpenAI Academy with new learning paths.

OpenAI improves prompt caching for GPT-6

OpenAI announced improved prompt caching for its GPT-6 model family, designed to enhance performance for persistent agents handling complex, multi-step tasks. The new system achieves higher cache hit rates by default, offers cache discounts of up to 90% on eligible shared prefixes within a 30-minute window, and introduces tools for monitoring and diagnosing cache performance. A new Prompt Caching Dashboard helps developers track hit rates and compare cached vs. uncached tokens. The diagnostics tool identifies causes of cache misses, such as changes to tools or settings. New features include explicit cache breakpoints, the ability to adjust reasoning effort without breaking cache, and guidance on preserving cache when tools or instructions change. Prewarming the cache is also supported to reduce latency. These updates aim to help developers optimize caching for cost efficiency and response times.

• •OpenAI launched improved prompt caching for GPT-6 models.
• •Cache discounts of up to 90% on eligible shared prefixes within a 30-minute window.
• •New Prompt Caching Dashboard allows monitoring of cache hit rates.

Claude Opus 5.5 Review: Cheaper, Faster, Less Annoying

Claire Vo, host of the How I AI podcast, reviews Anthropic's Claude Opus 5.5 after having abandoned Claude for months due to its previous 'annoying' behavior. She highlights that Opus 5.5 is 40% cheaper than Opus 5 and 30% faster, and it now feels less annoying to interact with. She tested the model across long-running agentic tasks, frontend prototyping, SVG illustration, and daily knowledge work tasks. The model performed well in agentic tasks (ignoring prompt injections, catching bugs), excelled in frontend design and SVG generation, and improved email writing in her voice. However, it still has a safety-first 'scold' personality, can be slow on long turns, and lags behind Codex in computer use and video editing. She concludes that Claude is back in her stack for specific use cases like PR review, architecture questions, and frontend work.

• •Anthropic released Claude Opus 5.5, priced at $4 per million input tokens and $20 per million output tokens, with a fast mode at $8 and $40.
• •Opus 5.5 is 40% cheaper than Opus 5 and 30% faster, aiming for performance close to Opus 5.1.
• •Opus 5.5 includes first-model alignment safeguards, with external evaluations and rejection of prompt injections, and cybersecurity tasks are diverted to Opus 4.8.

VideoAmp Receives JIC Re-Certification as National TV Measurement Currency Through 2028

VideoAmp announced it received JIC re-certification as a national TV measurement currency through 2028, as covered by Adweek. Additionally, VideoAmp is powering DeepIntent's new Cora platform for linear TV planning and forecasting, and GumGum's sports media solution for addressable audiences.

Anthropic, OpenAI Release Cheaper AI Models

Anthropic and OpenAI both announced new, more cost-effective AI models on Tuesday, marking their first releases since industry leaders called for a slowdown in advanced AI development. OpenAI introduced GPT-6 Sol and GPT-6 Luna, with API prices cut by 50% per token compared to GPT-5.6. Sol costs $2 per million input tokens and $10 per million output tokens, targeting complex workloads like coding, while Luna costs $0.10 and $0.50 respectively, designed for high-volume tasks such as information extraction. These models complement the flagship GPT-6 Astra. OpenAI attributes the price reduction to improved caching and inference efficiencies. Independent benchmarks show mixed results: Sol improves coding and reduces hallucinations but shows declines in knowledge work. Anthropic launched Claude Opus 5.5, a more token-efficient version costing about 40% less to run than Opus 5, intensifying price competition. The releases respond to customer demand for cheaper models and growing competition from open-weight rivals like Alibaba, Moonshot AI, and DeepSeek, amid heightened safety debates following a former Anthropic researcher's resignation.

• •OpenAI introduced GPT-6 Sol and GPT-6 Luna with API prices cut by 50% per token compared to GPT-5.6.
• •GPT-6 Sol costs $2 and $10 per million input/output tokens, while Luna costs $0.10 and $0.50.
• •Anthropic launched Claude Opus 5.5, costing about 40% less to run than Opus 5.

OpenAI launches GPT-6 Sol and Luna with lower cost, fewer errors

OpenAI launched two new GPT-6 models, Sol and Luna, on September 23, 2026. Sol targets complex tasks like coding and computer control, while Luna handles high-volume clerical tasks. Both are available in ChatGPT Work, Codex, and API at significantly reduced prices—about 50% cheaper than the previous 5.6 series—thanks to improved caching and inference, including up to 90% discounts on cached tokens. Sol achieves strong benchmark scores (e.g., 33.2% on AutomationBench), outperforming Claude Opus 5 in several tests, while Luna offers up to 93% cost savings on repetitive tasks. The release follows GPT-6 Astra and occurs in a competitive landscape, with OpenAI emphasizing safety and alignment in the deployment of more affordable AI agents.

• •OpenAI launched GPT-6 Sol and Luna on September 23, 2026, targeting complex and clerical tasks respectively.
• •GPT-6 Sol costs $2 per million input and $10 per million output tokens; Luna costs $0.10 and $0.50, both about 50% cheaper than previous series.
• •Prompt caching discounts up to 90% on reused tokens contribute to cost reductions.

Anthropic Launches Claude Opus 5.5 Despite Slowdown Call

Anthropic has released Claude Opus 5.5, its flagship AI model and the first in the 5.5 family, achieving the highest score ever recorded on the Artificial Analysis Intelligence Index (58) and leading six out of ten benchmarks, including outperforming OpenAI's GPT-6 Astra on Terminal-Bench 4.0. Priced 20% lower for input/output tokens and 60% lower for cache reads, it claims a 40% cost reduction for typical workloads, though higher output token usage may offset savings; it's also 30% faster and now the default in Claude Code and the Claude app. OpenAI responded by releasing GPT-6 Sol and Luna, priced ~50% lower than predecessors and offering up to 90% discounts on cached input. Both cite efficiency gains. Claude Opus 5.5 is available on Claude apps, Claude Platform, AWS, Google Cloud, and Azure, with smaller models to follow.

• •Claude Opus 5.5 scores 58 on the Artificial Analysis Intelligence Index, the highest recorded, and leads six out of ten benchmarks.
• •Pricing is reduced: $4 per million input tokens, $20 per million output tokens, and cache reads at $0.20 per million tokens, promising a 40% cost reduction per task (though higher output token usage may offset savings).
• •Opus 5.5 is 30% faster, supports multi-agent scaling up to 100 parallel agents, and is now default in Claude Code and the Claude app.

Microsoft Reveals AI-Centric Windows Plans for Agents

The Pragmatic Engineer interviews Microsoft Windows team executives about their strategy to integrate AI and agentic capabilities into the operating system. Key features include agent identity via Entra ID, a registry for MCP tools (ODR), an isolation mechanism for agent tools (MXC), a new hardware abstraction layer for local AI models (Windows ML), and embedded small language models (SLMs). Microsoft aims to win back developers by addressing past criticisms and supporting Linux via WSL. The article includes market share data showing Windows declining among developers, and highlights partnerships with NVIDIA for on-device AI performance.

• •Microsoft plans native agent identification using Entra ID and an On-Device Agent Registry (ODR) for MCP tools.
• •Microsoft Execution Containers (MXC) provide cross-platform isolation for agent tools with configurable containment policies.
• •Windows ML is a new hardware-agnostic layer for running local AI models across GPU, NPU, and CPU, based on ONNX runtime.