HackerNews Digest

English Heritage seeks a permanent Head of Stonehenge, working 36 hours weekly at the World Heritage Site. Responsibilities include leading a large, diverse team to deliver visitor experience, operational, retail and F&B performance, managing complex budgets, meeting revenue targets, overseeing solstice events, ensuring compliance and safety, and shaping long‑term strategy. Candidates must demonstrate proven strategic and operational leadership in visitor‑facing environments, strong commercial and financial acumen, high‑level communication and external representation, sound judgment in governance and health‑and‑safety, and inclusive leadership skills. Salary starts at £64,189 p.a., rising with experience, plus 25‑28 days holiday, pension matching up to 10 %, flexible hours, free site access for household, discounts, enhanced parental leave, and professional development support. Applications close 21 June 2026; first interviews via MS Teams the week of 6 July, final interview on 17 July at Stonehenge. Submit CV (max 2 pages) and cover letter (max 1 page) responding to “What attracts you to this role and why now?” Contact Danielle Reed for queries; no agencies.

Apple announced a major redesign of its Apple Intelligence platform, introducing a new architecture that incorporates foundation models co‑developed with Google using the Gemini technology. The models are engineered to operate both on‑device and via Apple’s Private Cloud Compute, enabling advanced multimodal functions such as realistic image creation, sophisticated photo editing, visual question answering, speech generation, and enhanced dictation accuracy. A central system orchestrator coordinates these capabilities across Apple’s ecosystem, tailoring responses to the active app and user task while maintaining on‑device processing and strict privacy safeguards—user data is used only for the immediate request and is not accessible to Apple or third parties. The announcement also referenced related developments: potential Gemini integration in Google Maps for CarPlay, Google’s upcoming Android 17 featuring “Gemini Intelligence,” and expanded cross‑platform file‑sharing (Quick Share) to additional Android devices.

Apple Intelligence embeds on‑device processing into iPhone, iPad, and Mac, enabling personalized features while keeping personal data local. When more complex tasks arise, Apple’s Private Cloud Compute runs larger models on Apple silicon servers; user data is used only for the specific request, never stored, and the privacy promise is verifiable. The platform powers Siri‑based interactions across the ecosystem, including: visual analysis of photos (object identification, nutrition facts, image‑search, and editing tools such as Clean Up, Extend, Reframe); contextual actions in Messages, Mail, Calendar, and Reminders (adding notes, creating events, proof‑reading, live translation); multimodal queries that combine text, voice, and images; and integration with Safari, Shortcuts, and password‑security features. Demonstrations span a range of devices—from iPhone 17 models to iPad Pro and MacBook Pro—showing Siri responding to queries about plants, sports equipment, food nutrition, and travel details, while maintaining end‑to‑end privacy.

None

- xAI, now a SpaceX subsidiary, has signed large‑scale compute‑leasing agreements with Anthropic and Google, using its older “Colossus 1” datacenter in Memphis. - Anthropic pays up to $1.25 billion / month for 300 MW (≈220 k GPUs) to lift peak‑hour usage caps; Google pays about $920 million / month for 110 k GPUs. Both contracts allow 90‑day cancellation after an initial lock‑in. - At the projected rates, xAI would recoup the estimated $40 billion construction cost of the facility in roughly 18 months, even before accounting for operating expenses and depreciation. - The deals address a persistent global GPU shortage, giving Anthropic immediate capacity relief and providing xAI with a high‑margin revenue stream. - Potential conflicts of interest are noted: Elon Musk’s legal dispute with OpenAI and Google’s stake in SpaceX may influence the arrangements. - Leasing capacity reduces the compute available for xAI’s own Grok model, suggesting a shift from frontier‑lab ambitions toward a datacenter‑REIT business model. - The rapid build time (Colossus 1 constructed in 122 days) highlights SpaceX/xAI’s execution advantage over traditional hyperscalers.

The page is titled “performative‑ui | AI‑native React Components,” indicating a project or library named performative‑ui that provides React UI components built with native AI integration. The focus appears to be on offering ready‑to‑use, AI‑enhanced components for developers working within the React ecosystem, aiming to simplify the incorporation of artificial‑intelligence functionality into user interfaces. No further description, documentation, or technical details are provided in the excerpt.

Xiaomi’s MiMo‑V2.5‑Pro‑UltraSpeed is a 1‑trillion‑parameter language model that achieves >1000 tokens / s decoding by combining FP4 quantization (restricted to MoE Expert layers) with the DFlash block‑level speculative decoding method and TileRT’s ultra‑low‑latency GPU inference system. The model runs on a single standard 8‑GPU commodity node, using TileRT’s persistent engine kernels and warp‑specialized pipelines to eliminate operator launch overhead. A limited‑time API (June 9‑23 2026, Beijing time) is offered at three times the regular MiMo‑V2.5‑Pro price, delivering roughly tenfold faster generation; access is application‑based, with per‑account caps of 10 queue entries per day and 30‑minute sessions. The speed enables real‑time use cases such as high‑frequency trading, anti‑fraud interception, interactive dialogue, and rapid code generation. Xiaomi has released the FP4‑quantized checkpoint and DFlash parameters on HuggingFace for community evaluation. UltraSpeed support for the broader MiMo‑V2.5 series is announced for future rollout.

GoGoGrandparent, a profitable, boot‑strapped company founded in 2016, provides phone‑based ride‑hailing and other on‑demand services for older adults with mobility, visual, cognitive or dexterity impairments. The fully remote engineering team seeks a Senior Backend Engineer with 6+ years of Node.js/TypeScript experience to design, implement, and maintain scalable backend systems (Node.js, TypeScript, MySQL, REST, GraphQL). Responsibilities include end‑to‑end feature delivery, performance monitoring, bug resolution, code reviews, architecture discussions, and mentoring. The role requires collaboration with product, QA, DevOps and occasional front‑end work (Vue.js). Preferred deployment experience includes AWS, Docker/Kubernetes. Compensation ranges $80k–$180k plus quarterly bonuses, 15 days PTO, health benefits, 401(k), and a remote work setup. Candidates must overlap ≥ 4 hours with US time zones and demonstrate ownership, clean code practices, and initiative.

A laboratory analysis of 64 food items—rice grain, various teas, paprika powder, chili, cumin seeds, and curry powder—revealed widespread contamination with pesticide residues. 49 samples (77 %) contained at least one pesticide, and 45 (70 %) contained substances not approved for use in the EU. Fourteen samples (22 %) exceeded the maximum residue limits set by EU law, rendering them non‑compliant for market sale. Every tested paprika, chili and cumin sample showed residues of non‑approved pesticides; one paprika sample contained 22 distinct pesticides, including six prohibited compounds. The most frequently detected unapproved pesticides were chlorfenapyr, bifenthrin, spirotetramat, clothianidin, thiamethoxam, imidacloprid and isoprothiolane. European Chemicals Agency data indicate that six of these pesticides were exported from EU member states to third‑country markets during 2024‑2025.

The scraped snippet consists solely of a page title “Core AI | Apple Developer Documentation” and a notice stating that JavaScript must be enabled to view the page content. No further sections, descriptions, API references, or technical details are present in the captured text. Consequently, the excerpt provides no substantive information about Core AI features, frameworks, or usage within Apple’s developer ecosystem. The only actionable detail is the requirement for JavaScript to render the underlying documentation. No additional content can be summarized. Accessing the full Core AI documentation would require a browser with JavaScript enabled to retrieve the complete resource.