Category
Intel has tapped former Marvell chief commercial officer Dean Jarnac to strengthen ASIC sales. Comparing the Google IPU and Fortinet SP6 collaborations, we look at what must happen for custom chip design wins to lead to Intel manufacturing, and at the volume and profitability challenges ahead.
Anthropic's Claude Dashboards builds updatable dashboards from company data, while Claude Motion turns reports and charts into animations. Here's what the official materials say about visible SQL, editable code, update timestamps, and the data access permissions of people you share with.
Meta's ban on ByteDance advertising covers Japan and extends to third-party ads that send users to TikTok. We look at how TikTok's link features differ, the teen-protection settlement terms, and Japan's platform transparency rules.
Anthropic has added dynamic workflows to Claude Managed Agents. We look at the gap between 1,000 total agents and 64 concurrent threads, the internal test's detection rate, and how budget overruns and completion status work, to see when delegating tasks and verification makes sense.
OneDrive storage in Microsoft 365 Family and Premium is shifting from 1TB per person to a shared 2TB. We look at how the change affects different households, conflicting start dates in Microsoft's documents, the overage grace period, and Japanese pricing.
Analyst Jeff Pu says NVIDIA may use Intel's Foveros for back-end packaging from the second half of 2028. The timing overlaps with Intel's 14A volume production and packaging revenue plans, but a January report named EMIB. The official materials reviewed contain no AI GPU packaging contract between the two companies.
Samsung is reportedly ending development of the next Exynos Auto automotive processor. We check the report against official sources, including HARMAN's Qualcomm partnership, Samsung Foundry's automotive chip manufacturing, and the focus on mobile APs and image sensors.
A harness connects an AI model's answers to file edits and verification. Using a bug-fix example, this article explains "Agent = Model + Harness," traces how the term spread from software testing to AI agents, and looks at how to confirm a task is truly done.
SpaceX has announced a deal to acquire a portfolio of U.S. 800MHz licenses. We look at the planned terrestrial network for indoor coverage, the FCC approvals still pending, and how this differs from partner services in Japan.
Radeon prices reportedly rise in October. In China, the RX 9060 XT 16GB and RX 9070 GRE are now just 100 yuan apart, though the GRE has 4GB less memory. Here is how official specs and tighter GDDR6 supply affect the choice.
