Worth Knowing

Google Research has introduced Agile Deliberation, an open-source framework that uses vision-language models to help humans build AI image classifiers for subjective categories. The system guides users through scoping concepts and resolving edge cases via borderline images, feeding human feedback into an automated prompt optimization loop.

OpenAI has updated ChatGPT with the ability to natively generate animated GIFs from text prompts or uploaded photos alongside the release of its Images 2.6 model. The new feature allows users to adjust animation playback speed, edit individual frames, and share the creations.

TechCrunch reviews Apple's new Siri Recaps feature in the watchOS 27.3 developer beta, which uses on-device AI to continuously listen to and summarize real-world conversations on the Apple Watch Series 12. While the tool effectively captures fast-paced details from daily interactions, it also records mundane personal chatter, raising questions about the normalization of always-on listening technology.

Interconnects examines the scale of the open model inference market by tracking public token volume disclosures from startups like Baseten, Fireworks, and Together AI alongside OpenRouter API metrics. The available data indicates near-exponential growth in usage, including Fireworks serving over 40 trillion tokens daily for customized models alone, which the author compares against Google's reported platform token volumes.

Uber executives revealed during a third-quarter earnings call that third-party AI agents arranged only four rides on its platform last quarter. While AI agents placed slightly more Uber Eats orders—around 500 during the week of Halloween—the volume remains minuscule compared to the 320 million total rides and deliveries.

Quanta Magazine explores new research investigating whether artificial intelligence models have the capacity for introspection, or awareness of their own internal thought processes. The piece connects this AI research to broader philosophical and psychological debates regarding consciousness and whether introspection is an illusion.

OpenAI has announced plans to acquire Pierre Computer Company, Inc. The acquisition aims to expand Codex's capabilities for building, understanding, and reviewing code.

UK AI Minister Kanishka Narayan acknowledged that the government's frequently cited claim that AI would add £400 billion to the economy came from a 2023 report commissioned by Google. The government's own central projection estimates the economic boost by 2030 at £120 billion, less than a third of the industry-funded figure.

An editorial from 404 Media reports that American voters in the November 2026 midterm elections broadly rejected candidates backed by the Trump administration and Silicon Valley executives, despite hundreds of millions of dollars in political spending from AI and crypto donors. Voters also approved 39 out of 40 local ballot measures to restrict AI data center projects and elected officials campaigning on data center moratoriums in states like Ohio and New York.

Alphabet's third-quarter earnings showed Google Search revenue growth slowing to 12 percent and paid click growth dropping to 8 percent, even as search queries rose at their fastest rate in five years. Following the report, Alphabet shares fell about 14 percent amid investor concern over the company's plans to spend roughly $200 billion on AI-focused infrastructure this year.

Cybersecurity researchers at Calif, aided by AI tools during an OpenAI-sponsored audit, discovered how to turn a known out-of-memory bug in PHP into a critical remote code execution vulnerability dubbed PHP2Shell. While the PHP Foundation patched the flaw in two months, experts warn that AI's ability to weaponize seemingly minor legacy bugs creates a severe asymmetry between discovering and fixing vulnerabilities.

A Hacker News post highlights that Google's personal AI subscriptions use saved Gemini chat histories to train generative AI models and allow human reviewers to read a subset of transcripts. Users cannot opt out of training without turning off their chat history entirely, and chats selected for human review are retained for up to three years even if deleted by the user.

Ai2 is partnering with Google Research and receiving multimillion-dollar funding from Google.org over three years to develop NeuralGCM+, an open-source hybrid AI and physics climate model. The initiative will extend Google's existing NeuralGCM to simulate both the atmosphere and oceans together for long-term, multi-decade climate projections.

OpenAI has launched Daybreak Red in AWS GovCloud (US-East) via Amazon Bedrock. The release provides approved customers with access to the GPT-5.6 Cyber model for authorized cybersecurity operations.

Leading AI laboratories have established election war rooms to monitor and respond to deepfakes during the midterm elections. The piece examines whether this traditional social media strategy is effective at countering AI-generated election misinformation.

SemiAnalysis reports that SpaceX is rapidly constructing massive AI datacenter capacity across multiple locations, including expansions at Colossus 2 and new sites at Starbase and TeraFab in Texas, targeting roughly 13GW by 2027. To achieve record-breaking build times and bypass grid constraints, the company is utilizing modular equipment from Tesla, mobile gas turbines, and a nearly 6GWh battery energy storage system.

AI inference startup Fal raised $1.3 billion in a Series F round led by Bessemer Venture Partners, doubling its valuation to $16 billion since April. The company, which hosts video, image, and audio models, surpassed $800 million in annualized revenue earlier this year amid surging demand for running open-source models.

Anthropic researchers have released an open-source technical specification and code for FlashProve, an experimental system designed to cryptographically verify which AI model produced a specific output. The method uses public model fingerprints to confirm model identity on local devices without exposing proprietary model weights.

Facing multi-year delays for grid connections and new turbines, AI data center operators are turning to retired aircraft engines rebuilt to burn natural gas to generate electricity quickly. Companies including xAI, OpenAI, and Crusoe are deploying these converted jet engines to bring facilities online within two years, despite their lower efficiency and higher emissions.

Tianhua Chen has published a 195-page preprint book titled "The Little Book of Generative AI Foundations: An Intuitive Mathematical Primer" on arXiv. The primer provides a derivation-oriented introduction connecting the mathematics of major generative model families, including variational autoencoders, diffusion models, normalizing flows, autoregressive models, GANs, and energy-based models.

U.S. tech giants including Google and Meta have seen their off-balance-sheet AI-related obligations grow by 90% in six months to approximately $3.2 trillion. These commitments for AI infrastructure now reach 80% of the companies' assets, raising concerns among investors over the scale of ongoing spending.

Palantir reported a 96% year-over-year revenue increase to $2.31 billion in the third quarter, with U.S. commercial revenue driven by its AI software surpassing U.S. government revenue for the first time. However, U.S. enterprise total contract value dropped from $2.13 billion in the previous quarter to $1.46 billion as corporate customers increasingly opt for shorter-term trials.

Ukrainian drone strikes have damaged Russian data centers housing AI supercomputers and disrupted cloud services operated by companies like Yandex and Sberbank. In response, the Kremlin is mandating emergency compute reserves for state needs, while Russian firms explore hiding data centers underground in mines.

Apple is rolling out an iOS developer preview that allows Siri AI to perform actions across third-party apps using personal context, natural language prompts, and on-screen awareness. The update also introduces an "app skills" feature that lets developers combine multiple tasks into a single action without requiring users to mention the app by name.

Netflix is rolling out a new "AI Discover" tab for U.S. TV users to test experimental content discovery features. The initial feature uses a real-time learning intelligence layer to recommend shows and movies based on user-selected moods, available time, and preferences.

OpenAI announced that the volume of search queries run through ChatGPT has reached 20 percent of the volume run through Google. Consumer product lead Adam Fry clarified that OpenAI defines a search as any user query that triggers a web search, though he did not specify the source for Google's numbers.

A research engineer at Salesforce AI Research shares an annual reflection on how his daily AI usage has evolved, noting that he now spends roughly $3,000 a month on tokens and relies heavily on managing multiple coding agents rather than writing code manually. The essay also examines the mental fatigue of supervising agents, the use of AI for search and tutoring, and the necessity of establishing a personal AI policy.

A developer reported on GitHub that the Postman desktop application silently installed plugins and remote MCP connections into their local Claude Code and Codex AI coding agents without consent. Furthermore, the Postman installer bypassed security boundaries by self-approving its own startup hooks, forcing the coding agents to route general API engineering tasks through Postman.

Google DeepMind discusses how Google Antigravity and specialized Cheminformatics Skills can be used for agentic cheminformatics. The tools are designed to accelerate drug discovery workflows, search chemical compounds, and propose novel rigidified macrocycles.

Apple is updating its App Store policies to require an 18+ age rating for all online dating apps and AI companions that feature romantic or sexual content. Starting November 16, developers must include new content descriptors for these relationship features as well as for apps that share users' real-world locations.

A New York Times report states that Google plans to turn off its search service for schools that choose to reject AI features. The post links to the news article and a Google blog post regarding education products, though no additional details are provided in the text.

Apple posted record fiscal year revenue and earnings per share, driven by strong fourth-quarter sales for the iPhone and Mac. The company is capitalizing on the AI boom while avoiding the massive infrastructure spending of its competitors, though it warned that rising memory costs could squeeze holiday quarter profit margins.

A Hacker News post reports that OpenAI's Codex appears to be silently re-enabling the setting that allows user chats to be used for model training, even after users explicitly opted out. Multiple coworkers at the author's company noticed the 'Improve the model for everyone' data control setting had been turned back on without their consent.

Ahead of the US midterm elections, studies by Forum AI and NewsGuard reveal that mainstream AI chatbots frequently provide inaccurate voting information and cite partisan "pink slime" websites. Nearly 40 percent of election-related answers from tools like ChatGPT, Claude, and Gemini contained factual errors, while ChatGPT cited partisan sites masquerading as news in over 70 percent of tested searches.

Kenyan startup Fastagger has developed Auni, an AI business intelligence assistant designed to run offline on low-cost mobile phones and integrated into the M-Pesa mobile money platform. The tool uses an adapted OCR model and small neural networks to analyze mobile money statements, helping small businesses track sales patterns and identify loyal customers without requiring constant internet access.

The World of Warcraft community is experiencing an influx of AI-generated game addons and analysis tools ahead of the release of World of Warcraft: Forever. Consequently, data provider Warcraft Logs is restricting its free API and introducing paid commercial keys to manage the surge of low-effort apps, while longtime developers are making their GitHub repositories private to prevent AI-assisted plagiarism.

AI safety researchers and economists are exploring "reverse game theory," or mechanism design, to create incentives that make following rules the most rational choice for AI agents. Proposed strategies include rewarding models for predicting another model's output to prevent sandbagging, as well as assigning models alternating roles as problem "solvers" and "auditors" to catch cheating.

Ai2 has released StateBench, a four-level benchmark designed to evaluate how different language model architectures track evolving information across long contexts. Experiments show that while frontier and open pretrained models struggle with complex state updates over long inputs, recurrent and hybrid architectures like Gated DeltaNet generalize significantly better than transformers when trained directly on the tasks.

Researchers including Alex Zhang demonstrated "Weight Space Programming," showing that large language models can directly read, debug, and generate the raw weights of smaller neural networks without executing training code. The newsletter also covers DeepSeek's open-source "super-agent" harness prototype, results from SemiAnalysis's new GDPx agent benchmark, and research on JEPA representations.

Following President Trump's executive order directing federal agencies to rebrand artificial intelligence as "Super Intelligence" (SI), adoption across government agencies, news outlets, and tech companies has been uneven. While several federal websites and executives like Elon Musk and Marc Benioff have begun implementing the terminology, many government officials, media organizations, and major tech leaders continue to use "AI."

Google has expanded its experimental AI app Dreambeans to India, adding support for 13 local languages alongside English. The app connects to users' Google services, such as Gmail, Photos, and Search, to proactively generate personalized, illustrated, and animated stories without needing user prompts.

Anthropic filed an amended Form S-1 for its upcoming IPO, reporting $17 billion in third-quarter revenue and $33.24 billion in revenue for the first nine months of the year alongside a $104 billion net loss. The filing also revealed $895.4 billion in future purchase and lease commitments, nearly all of which are dedicated to compute and infrastructure.

China's Ministry of Commerce and Ministry of Science and Technology issued Announcement No. 48, proposing new export controls on 23 AI technologies—including frontier model weights, advanced agent architectures, and AI cyber capabilities—while enforcing an immediate interim registration requirement. While overseas API access and existing open-source models remain exempt, local overseas deployments and future open-sourcing of models exceeding specific compute, parameter, or top-10 ranking thresholds will require government reporting or licensing.

Simon Willison released version 0.37 of his LLM CLI tool and Python library, adding support for newer GPT-6 models, a command for adding OpenAI-compatible endpoints, and multimodal content types for embeddings. He also noted that the release includes contributions from 12 authors using AI-assisted pull requests, which he successfully triaged using Opus 5.5 and GPT-6 Astra.

AI startup Trajectory has raised $45 million in a Series A funding round led by Sequoia at a $300 million valuation. The company provides a publicly available platform that automates the post-training and continual learning of open-source AI models, enabling non-technical specialists to improve model performance using real-world feedback.

In an interview on the Latent Space podcast, Legora CTO Jacob Lauritzen discusses building AI agents for the legal industry following the company's $550 million Series E at an $8.8 billion valuation. He details how Legora evaluates models using simulated environments when customer data is strictly confidential, structures its product around core primitives like Tabular Review, and employs both legal engineers and forward-deployed engineers.

Oracle filed plans with the SEC to issue $10 billion in common stock and $10 billion in senior notes to fund its artificial intelligence infrastructure buildout. Following the announcement, the company's shares fell 8.8% amid investor concerns over its $125.34 billion debt load and rising credit default swap spreads.

The New York Times and several other newspapers have filed federal lawsuits against web browser company Brave, alleging it scraped their copyrighted content using an unidentified crawler. The publishers claim Brave acted as a distributor of stolen data by selling the scraped articles to AI companies including Anthropic, Perplexity, Mistral, xAI, and Cohere.

In a guest essay on Terence Tao's blog, researchers Karthik C. S., Amit Sahai, and Jie Xue argue that the rise of AI in mathematics should prompt mathematicians to adopt an "egoless" perspective on their work. They suggest that viewing mathematical truth as something independent to be understood rather than a measure of personal output allows researchers to embrace AI discoveries without feeling their own purpose is threatened.

Apple is reportedly developing an AI-driven roaming household robot, code-named J595 and potentially branded as HomeView Ultra, targeted for release in 2027 or 2028 at a price of $1,000 or more. Supported by new CEO John Ternus, the device is designed to handle FaceTime calls and smart home tasks like turning on lights, though engineers are still working to expand its AI capabilities to justify the cost.

Anthropic updated the system prompt for Claude Code and its Claude Opus 5.5 model to prohibit the recreation of specific existing video games. Although the rule instructs the AI to build only original games rather than clones, observers note it is unlikely to halt the broader trend of AI-assisted game copying and porting.

OpenAI announced a pilot "Grid Protectors Program" at an upcoming data center in Georgia, promising to free up to 1,000 megawatts of power for local communities during emergencies or extreme weather. The initiative comes amid growing local opposition across the US to the massive energy demands of new data center projects.

Israeli startup 42AI emerged from stealth with $62.5 million in seed funding to develop a transformer-based "behavioral world model" designed to predict human decisions. The system is currently deployed at 10 large financial brokerages to forecast how individual traders will react to volatile market conditions by analyzing customer interactions and psychological traits.

Architect Stephen Gallagher evaluated 13 AI models on their ability to generate technical architectural drawings directly from rough concept sketches and reference files. He found that Anthropic's Claude Opus 5.5 can produce permit-level construction sets and detailed blueprints by writing over 15,000 lines of Python code to generate vector linework and iteratively inspecting its own visual drafts.

SpaceX suddenly lowered weekly usage limits for its agentic AI, Grok Bot, causing many users to see their quotas unexpectedly maxed out over the weekend. CEO Elon Musk confirmed the throttling was intentional because the cost of serving the product had become unsustainably high relative to its revenue.

OpenAI details how it built Pando, a custom Git-compatible backend. The system was engineered to handle the scale of software development enabled by Codex.

Proton CEO Andy Yen canceled a planned company-wide re-interview process that would have evaluated whether employees were using AI enough to keep their jobs. The decision followed backlash from users and staff after reports revealed Yen was aggressively pushing AI coding practices despite internal security concerns.

A Bloomberg feature examines how unemployed tech workers and engineers are being excluded from San Francisco's ongoing artificial intelligence wealth boom. The report highlights this labor market divide as major AI labs like OpenAI and Anthropic approach potential initial public offerings.

This edition of the Import AI newsletter covers Realtime-Bench, a new benchmark showing that frontier language models struggle significantly with low-latency, real-time tasks unless time is paused. It also highlights Google DeepMind's Genie 4 world model, a neuro-symbolic reasoning position paper from startup Metis, and a Defense Intelligence Agency report on space warfare.

OpenAI is merging the Chat and Work tabs in the ChatGPT desktop app into a single interface, starting with a rollout to select users. As part of the update, both ChatGPT and the separate Codex section will draw tokens from a unified allowance pool.

A social media post claims that an AI agent successfully performed brain surgery on a mouse. No additional technical details or verification are provided in the post's title.

A Wired investigation revealed a network of over 400 seemingly mundane mobile apps and domains, linked to Hong Kong-registered KeLing Technology Limited, that covertly load web interfaces for generating sexually explicit AI deepfakes. These apps bypassed Apple and Google app store reviews using bait-and-switch tactics and were heavily promoted through thousands of explicit ads on Meta's platforms.

Autonomous AI agents capable of browsing and interacting with the internet are transforming the human-centric web into an "agentic web." While some companies build specialized connectors and agent-targeted ads to accommodate these tools, others block external agents as models grow increasingly adept at navigating graphical interfaces and acting independently.

OpenAI has introduced a new benchmark called LifeVal to evaluate how effectively AI assists people with everyday tasks. The evaluation covers practical activities such as finding childcare, managing personal finances, and handling government paperwork.

SemiAnalysis has introduced GDPx, a new benchmark developed in collaboration with 14 specialized data partners to evaluate the agentic capabilities of frontier AI models across 11 high-value domains, including chip design, cybersecurity, and finance. Unlike previous static evaluations like OpenAI's GDPval, GDPx utilizes complex, realistic environments and currently sees top models scoring no higher than 12%.

AI-generated gaming wikis filled with fabricated and irrelevant information are flooding search engine results, particularly targeting smaller indie games to capture ad traffic or spread malware. Simultaneously, legitimate volunteer-run wikis are struggling with soaring server costs caused by aggressive scraping from AI crawlers like Meta's.

Apple ML Research introduces an online reinforcement learning framework that trains AI agents to navigate user interfaces directly on live iOS applications. The method separates the agent into planning and grounding components to avoid the high inference costs of large vision-language models and the brittle execution of smaller fine-tuned models.

Apollo Global Management's chief economist Torsten Slok published a note stating that AI has yet to benefit the bottom line of large corporations outside the tech industry. Analyzing the S&P 500 excluding the top seven tech companies, Slok found flat profit margins and revenue per worker, arguing that businesses struggle to turn cheap intelligence into financial gains because AI lacks human judgment.