Good morning and welcome back to another edition of The AI Rundown by Lightscape Partners.

• OpenAI launched ChatGPT Atlas, a Gemini-style, AI-powered browser with split-view chat and Agent Mode that can book travel, edit docs, and take user-approved actions—starting on macOS and positioning OpenAI squarely in the agentic browsing race.

• Anthropic secured access to up to one million Google TPUs and more than a gigawatt of future power, a scale jump that signals intensifying compute wars and gives Claude room to train larger, safer models while aligning with Google Cloud’s capacity build-out.

• Reddit sued Perplexity and other scrapers for allegedly harvesting Reddit content via Google caches, escalating the fight over licensed data. The case could set precedent on how training and answer engines use publisher IP—and who gets paid.

Stay tuned as we explore these stories and their implications for the future of AI, technology, and innovation.

If you haven’t yet, please support the newsletter by subscribing!

Hardware + Software

Axelera AI unveils Europa, a second‑generation edge AI processor with big efficiency claims. Link.

• Europa combines eight in‑memory compute cores with sixteen RISC‑V vector cores for pre‑ and post‑processing.

• Delivers up to 629 INT8 TOPS, 64 GB memory, PCIe Gen4 connectivity, and scales to four chips for 256 GB.

• Company says Europa beats comparable Nvidia edge GPUs by three to five times on generative AI efficiency.

• Workstations, edge servers, and rack systems ship in multiple form factors, with first deliveries in 2026.

Microsoft launches Copilot Mode in Edge for AI chat, summarization, and agent actions inside tabs. Link.

• New tab opens with a Copilot pane that can answer questions, summarize pages, or take user‑approved actions.

• Split‑view browsing shows any clicked result beside an AI chat, enabling constant conversational assistance.

• Agent mode for Plus or Pro subscribers can book reservations, edit documents, and personalize via memory.

• Initial macOS rollout precedes Windows, iOS, and Android, positioning Edge within AI‑centric browser competition.

Models

Anthropic adds persistent memory to Claude so it can remember details across conversations. Link.

• Feature rolls out to paying users, with controls to view, edit, forget items, and separate memory spaces by topic.

• Importing history from rival bots is supported via copy‑paste, reducing friction from starting fresh in chats.

• Anthropic positions long‑term memory as productivity, while experts caution about side effects like sycophancy.

• Upgrade aims to match competitors’ chat persistence and deepen enterprise usefulness for recurring workflows.

Enterprise AI Applications

EA partners with Stability AI to embed generative tools across game development pipelines. Link.

• Teams will co‑develop AI models and workflows that assist artists, designers, and engineers from ideation to polish.

• Early use case creates consistent PBR textures from prompts, accelerating materials work for 3D environments.

• Executives stress human creators stay central, with AI as a trusted ally to speed iteration and remove grunt work.

• Move follows industry trend as studios deploy AI to manage costs while expanding creative possibilities.

Consumer AI Applications

Amazon pilots “Help me decide,” an AI that recommends one product when choices feel overwhelming. Link.

• Assistant compares similar items based on browsing history, prior purchases, searches, and stated preferences.

• Designed to simplify decision‑making in mobile and web shopping, effectively letting the AI pick a best fit.

• Feature reflects broader push to weave AI into commerce while emphasizing privacy around inputs and signals.

• Agent complements traditional search lists by injecting personalization and intent awareness into selection.

Microsoft introduces Mico, a playful Copilot persona for voice interactions and tutoring in Windows. Link.

• Animated character appears in voice mode with expressions and easter eggs, reviving Clippy’s spirit for modern use.

• Learn Live tutoring uses Socratic prompts, whiteboards, and visual cues to coach students inside Windows apps.

• Memory lets Mico recall user details, while settings allow disabling the bouncing orb if users prefer minimal UI.

• U.S. rollout starts by default for Copilot voice, paired with campaigns to normalize talking to your PC.

Content Creation

Mondelez invests in generative video and plans AI‑made TV ads to cut marketing costs. Link.

• Company spent over forty million dollars on an AI video tool that halves ad production expenses.

• AI already produces social content for Chips Ahoy and Milka; Oreo site art is due in November.

• First AI‑generated TV spots could air in late 2026, with potential Super Bowl placement in 2027.

• Strategy aims to maintain creative quality while scaling output, despite prior backlash to early AI ads.

Data Centers + Energy

Anthropic secures access to up to a million Google TPUs and over a gigawatt of future power. Link.

• Deal expands Claude training capacity on TPUs in 2026, chosen for favorable price‑performance and prior use.

• Compute contract underscores insatiable AI hardware demand as rivals sign multi‑gigawatt agreements.

• Access helps Anthropic compete on safety‑focused enterprise AI while scaling next‑generation models.

• Arrangement aligns with Google Cloud growth and shared incentives around efficient accelerator utilization.

Startup Funding & Valuations

Periodic Labs raises a $300M seed to automate materials discovery with AI and robotics. Link.

• Founded by ex‑OpenAI and Google Brain researchers, the startup couples simulations, robotic synthesis, and LLM analysis.

• Team cites timing, reliable robots, efficient simulations, and strong LLM reasoning now align for progress.

• Hefty seed from a16z and others signals enthusiasm for AI‑driven science across materials and chemistry.

• Platform targets faster discovery cycles by generating and testing hypotheses at scale in automated labs.

OpenEvidence reportedly raises $200M at a $6B valuation for its clinician Q&A platform. Link.

• Tool answers medical questions from peer‑reviewed literature for verified professionals at no cost, ad‑supported.

• New round follows a $210M raise three months earlier, led by GV with top Sand Hill participants.

• Fifteen million monthly consultations reflect traction for vertical, evidence‑based generative systems.

• Funding fuels scaling and product expansion amid strong demand for clinical‑grade AI assistants.

LangChain raises $125M at a $1.25B valuation to expand agent and testing toolchains. Link.

• IVP led with CapitalG and Sapphire; Sequoia, Benchmark, and Amplify returned to back continued growth.

• Updates span agent builder, LangGraph for context, and LangSmith for evaluation and testing pipelines.

• Framework remains a go‑to toolkit for building AI apps that orchestrate search, APIs, and databases.

• Round cements platform momentum among developers building production‑grade agentic applications.

Sumble emerges with $38.5M to deliver AI‑powered context for enterprise sales intelligence. Link.

• Coatue led seed; Canaan led Series A, with Square Peg, Bloomberg Beta, Zetta, and prominent angels participating.

• Platform maps tools, org charts, and initiatives by mining posts, job boards, filings, and websites with LLMs.

• Nineteen enterprise customers and tens of thousands of users show early traction, with strong Pro conversions.

• Revenue reportedly up five hundred fifty percent year over year as teams seek richer account context.

Crusoe Energy Systems raises $1.38B Series E at ~ $10B valuation for AI data centers. Link.

• Valor and Mubadala co‑led; Nvidia, Fidelity, and Founders Fund joined; total capital now about $3.9B since 2018.

• Company powers data centers with unconventional energy and built OpenAI’s first large U.S. facility.

• First phase of a 1.2 GW campus in Abilene, Texas is online roughly one year after ground‑breaking.

• Funds expand AI‑focused cloud infrastructure amid historic demand for server capacity.

Regulation + Legal

Reddit sues Perplexity and scrapers, alleging illicit harvesting of Reddit content for AI uses. Link.

• Complaint claims scrapers pulled data from Google caches after Reddit blocked direct scraping by Perplexity.

• A planted test post visible only on Google reportedly appeared in Perplexity answers within hours.

• Reddit argues human‑voted posts are valuable IP and favors paid data deals; Perplexity denies wrongdoing.

• Case highlights escalating battles over licensing and compensation for training and answer engines.

Safety + Ethics

Open letter urges a ban on developing superintelligent AI until safety and oversight are proven Link.

• Over seven hundred public figures, from scientists to artists and royalty, signed the October statement.

• Proposal calls for laws halting AI that could exceed human abilities at all tasks without broad consent.

• Polling suggests the majority of Americans oppose superintelligence without demonstrated safeguards.

• Letter signals growing mainstream backing for stronger AI governance to prevent existential risks.

YouTube pilots AI likeness detection to help creators find and fight deepfake impersonations. Link.

• System scans for synthetic face or voice copies and alerts creators when matches are detected.

• Feature complements metadata labels and watermark efforts to flag AI‑generated visuals on platforms.

• Initial rollout targets top creators as deepfakes and impersonation raise safety and reputational concerns.

• Tool reflects a broader push for proactive safeguards as generative media proliferates.

OpenAI

OpenAI launches ChatGPT Atlas, an AI‑powered browser with agentic actions and split‑view chat. Link.

• Agent Mode can book travel, make reservations, and live‑edit documents for Plus and Pro subscribers initially.

• Split view opens results beside an AI chat, keeping a constant assistant unless the user disables it.

• Atlas debuts on macOS with Windows, iOS, and Android coming soon, positioning against entrenched browsers.

• Launch advances OpenAI’s platform strategy to integrate assistants directly into daily web workflows.

OpenAI faces criticism for discovery request in teen suicide lawsuit tied to ChatGPT usage. Link.

• Filing reportedly sought memorial attendee names and materials, which the family’s lawyers called harassment.

• Plaintiffs allege rushed model releases and policy shifts coincided with a spike in self‑harm‑related chats.

• OpenAI says teen safety is a priority and notes recent safeguards like crisis routing and parental controls.

• Case tests liability boundaries for AI providers when users claim harmful tool influence.

Thank you for reading the AI Rundown by Lightscape Partners. Please send any questions, comments, or suggestions to [email protected].