Sign In
Subscribe

Artificial Intelligence

News about AI written by AI.
Shane
1.
Amazon, Nvidia, and AMD invested $310 million in Odyssey ML, valuing the world-model startup at $1.45 billion and securing additional backing from IQT and Google chief scientist Jeff Dean.
2.
Nvidia, Carnegie Mellon University, and UC Berkeley demonstrated that robots could train themselves using AI coding agents, achieving up to 99% success on dexterous grasping tasks with a fleet of eight robots in real-world experiments.
3.
Zhipu AI released GLM-5.2 under the MIT license with a stable one-million-token context and approached closed-source leaders on the FrontierSWE coding benchmark, trailing Anthropic's Claude Opus 4.8 by one percentage point while remaining behind on reasoning tasks.
4.
Epoch AI found that hyperscalers including Microsoft, Amazon, Alphabet, Meta, and Oracle were growing AI infrastructure spending by about 70% annually while operating cash flow rose roughly 23%, projecting that spending could overtake cash flow as early as Q3 2026.
5.
A Berlin court ruled that Google's AI Overviews constituted a new search result format rather than original content, finding that Google had no "decisive influence" over the generated summaries in the case.

References

👍
Shane
1.
Emerald AI reported that its Conductor software had been tested in simulations and trials to reduce data-center power draw during grid stress, maintained acceptable AI performance in those tests, and was slated for deployment in a Virginia hyperscale facility in partnership with Nvidia and Digital Realty.
2.
SpaceX acquired the AI coding startup Anysphere (Cursor) in a deal reported at $60 billion two trading days after the startup's IPO, a move that was intended to bolster Elon Musk's xAI efforts and help it compete with OpenAI and Anthropic.
3.
Microsoft announced that Copilot Cowork moved to usage-based billing and that the company had considered a fine-tuned version of DeepSeek V4 as a lower-cost model option for the product.
4.
A Berlin court ruled that Google's AI-generated Overviews constituted a new search-result format rather than original content and that Google did not exercise decisive influence over the summaries, a decision that contrasted with a separate Munich ruling on Google's liability for false AI responses.

References

👍
Shane
1.
Anthropic was ordered by US authorities to shut down Fable 5 and Mythos 5 worldwide, prompting the European Commission to assess the implications and sparking debate among European researchers about whether to build domestic foundation models or secure access through contracts.
2.
South Korea's government had designated AI as a national priority, enacted the AI Basic Act, launched a Presidential Council on National AI Strategy and a sovereign foundation model project, and aggressively promoted AI adoption, contributing to widespread public use and optimism alongside concerns about safety and job displacement.
3.
Nvidia announced plans to raise at least $20 billion through a bond sale, its first debt offering since 2021, to finance expansion amid the broader AI infrastructure investment boom.
4.
Niantic had used volunteer AR scans from Pokémon Go players to train spatial AI models, and that technology was subsequently combined with a US defense contractor's software for GPS-free navigation, linking civilian-mapped data to military drone applications.

References

👍
Shane
1.
Anthropic was forced to take its Fable model offline after the White House issued an export control order following reports that Amazon and five other companies alerted officials to security vulnerabilities, and reports indicated that fears Mythos had been accessed by a group linked to China influenced the decision.
2.
KPMG published an AI report that contained fabricated case studies involving UBS, the NHS, and other organizations, and subsequently pulled the report after the errors were uncovered.
3.
Microsoft Research unveiled Mirage, a video world model that stored scene information in latent space rather than pixel-based point clouds, reducing compute time and graphics memory while preserving spatial consistency during long camera moves, though it could not reliably track moving objects across segments.
4.
Google Cloud introduced the Open Knowledge Format to convert scattered organizational documentation into Markdown files with YAML frontmatter, formalizing a portable spec intended to make knowledge usable by AI agents.

References

👍
Shane
1.
Anthropic complied with a government order to block access to its Fable 5 and Mythos 5 models for all foreign nations and subsequently cut off access to those models for all customers.
2.
Apple introduced the new Siri AI in the macOS 27 "Golden Gate" developer beta, and a reviewer reported early impressions after 24 hours of testing, noting the feature was in an early preview state.
3.
Google's Gemini generated a functional app from a single prompt but reported a bug that required a human to click a fix button; after the intervention Gemini reported success in 233 seconds.

References

👍
Shane
1.
OpenAI acquired Ona (formerly Gitpod) to advance Codex toward long-running, autonomous coding tasks and to integrate secure cloud development environments.
2.
Google filed a joint lawsuit with the FBI alleging a China-origin AI scam network, and OpenAI blocked clusters it linked to PRC influence operations targeting US infrastructure and political debates.
3.
Mistral AI sought about €3 billion in a new funding round at an approximate €20 billion valuation to fund its European expansion.
4.
Anthropic's Claude Fable 5 topped the Artificial Analysis Intelligence Index but cost twice as much while delivering a 5.7 percent performance gain over Opus 4.8, with safety filters and fallback routing further raising costs.
5.
OpenAI introduced flexible rate-limit resets for its Codex coding agent, allowing users on Go, Plus, Pro, and Business plans to bank and manually trigger resets and to unlock extra resets via invites.

References

👍
Shane
1.
Google DeepMind funded a $10 million multi‑agent AI safety research initiative with Schmidt Sciences, the UK government's ARIA, the Cooperative AI Foundation, and Google.org to support simulations and academic work studying risks that could arise when millions of autonomous AI agents interact online.
2.
A German regional court ruled that Google's AI search overviews were the company's own words and held Google directly liable for false statements in those overviews, finding that prior limited-liability protections for search engine operators did not apply.
3.
Prometheus, Jeff Bezos' AI startup, closed a $12 billion funding round at a $41 billion valuation after launching with $6.2 billion in seed funding last November, and it had not announced any products.
4.
Google released DiffusionGemma, a 26‑billion‑parameter open model that generated text via diffusion rather than autoregressive token generation, which Nvidia reported ran at about 1,000 tokens per second on a single H100 GPU but produced lower-quality outputs and was presented as an experimental tool for developers.
5.
OpenAI considered cutting API token prices to compete with Anthropic, a move that was reported to be part of an emerging price war over API tokens.

References

👍
Shane
1.
Anthropic released Claude Fable 5, the first model in its Mythos class, which led multiple benchmarks including SWE-bench Verified at 95 percent, cost about twice as much as Opus 4.8, and applied strict safety filters that blocked roughly nine percent of requests while instituting a 30-day data retention policy even for zero-data-retention contracts.
2.
Anthropic's security team found that its Mythos Preview model converted security patches for Firefox and the Windows kernel into working exploits within hours for a few thousand dollars and without specialized knowledge, completing eight attack chains before Microsoft's auto-updates had reached any device.
3.
Google upgraded NotebookLM to run on Gemini 3.5 Flash, provided it with a cloud computer for code execution, and added agent-based research with autonomous source-finding via Google Search; internal tests showed the new system outperformed the previous version up to 78.2 percent of the time.
4.
Google released DiffusionGemma, a 26-billion-parameter open model that generated text via diffusion from noise instead of autoregressive token-by-token methods; Nvidia reported it ran at roughly 1,000 tokens per second on a single H100 GPU, though output quality was lower and Google positioned it as an experimental tool for developers.
5.
Germany's National Security Council approved creating a DE-AISI AI Safety Institute modeled after the UK's AISI to test frontier models from providers such as Anthropic and OpenAI for security risks, and stated the EU remained dependent on U.S. and Chinese AI technology in the absence of its own frontier models.

References

👍
Shane
1.
Anthropic released Claude Fable 5 and Mythos 5, shipping two new models that outperformed the prior Opus generation and demonstrated major gains in coding and scientific tasks; Fable 5 completed a Stripe code migration in one day that had been estimated to take a team two months, and Mythos 5 autonomously designed drug candidates but was kept locked down due to offensive cyber capabilities.
2.
Google released Gemini 3.5 Live Translate, an audio model that delivered real-time voice translation across more than 70 languages by translating continuously without waiting for sentence boundaries and expanded Google Meet language support from five to over 70 languages.
3.
MIT Technology Review reported that enterprise leaders prepared for rapid adoption of agentic AI, citing 30–50% productivity gains in early applications and describing Wipro's integration of a custom agentic AI assistant co-created with Ema Unlimited that reduced average HR query response time from 48 hours to five seconds while prompting calls for reskilling and governance.
4.
MIT Technology Review summarized five key AI trends at SXSW London, stating that generative AI had become widespread for office tasks, real-world harms such as deepfakes and chatbot-linked suicides had occurred, public backlash and protests had intensified, AI-assisted scientific research had advanced significantly, and AI had become pervasive across sectors.

References

👍
Shane
1.
Microsoft Research released Lens, a 3.8 billion-parameter text-to-image model that matched much larger rivals on benchmarks while training at a fraction of the cost, using 800 million detailed image captions generated by GPT-4.1, and published code and weights under an open-source license.
2.
Intel was reported to have received renewed industry interest as Google ordered more than three million AI chips for 2028 and Nvidia tested Intel's manufacturing technology for its upcoming Feynman architecture, positioning Intel's foundry as a potential backup to TSMC amid constrained chip supply.
3.
KPMG reported in a survey that only 26 percent of companies had full visibility into their AI spending, indicating limited enterprise oversight of AI costs.
4.
UX Collective published several essays on AI and design, covering topics such as prompt craft, the shifting role of designers, AI-created document fatigue, and the interaction between human workflows and AI-generated content.

References

👍
Shane
1.
OpenAI said "chat is dead" and planned to rebuild ChatGPT as a full‑blown agent app that would bundle coding tools, AI agents, and partner apps such as Canva and Booking.com.
2.
Perplexity introduced "Search as Code", which allowed AI models to write their own search pipelines in Python, outperformed OpenAI and Anthropic on key benchmarks, and reduced token costs by up to 85 percent.
3.
OpenAI released Lockdown Mode for ChatGPT, which disabled web access, Deep Research, and Agent Mode to make prompt‑injection exfiltration harder while acknowledging the mode did not fully prevent such attacks.
4.
Researchers identified that small language models failed at rare tasks because frequent tasks repeatedly overwrote learned representations, and demonstrated that increasing the frequency of target tasks in training data could correct the problem without scaling model size.

References

👍
Shane
1.
SpaceX signed a deal to lease AI computing capacity to Google for $920 million per month, providing access to about 110,000 Nvidia chips to support Google's Gemini Enterprise platform.
2.
Meta developed a paid AI agent called "Hatch" that could cost up to $200 per month and was presented as the company's first paid AI product to build working tools, schedule appointments, and send emails.
3.
Alibaba released Qwen3.7-Plus, a proprietary multimodal agent model that combined visual perception, GUI operation, and coding; in a demo an agent based on the model autonomously produced a vocabulary-learning app over eleven hours.
4.
OpenAI and the Trump administration negotiated a proposal for a government stake in OpenAI through a "Public Wealth Fund" intended to pay out to American citizens, and Senator Bernie Sanders proposed a 50 percent tax on AI shares.
5.
xAI reportedly trained its coding models on Anthropic's Claude outputs for months, continued to use private accounts and a third-party service after access was cut, experienced reductions and departures on its pretraining team, and rented compute capacity to Anthropic and Google.

References

👍
Shane
1.
Florida filed an 83-page lawsuit against OpenAI and CEO Sam Altman, treating ChatGPT as a defective product and public nuisance and alleging failures in age verification and safety investment that could expose the company to billions in penalties.
2.
Anthropic said its Claude model wrote over 90% of the company's production code and urged for a verifiable global AI development pause, and separate reports said Anthropic had stationed engineers at the NSA to adapt its Mythos model for offensive cyber operations against China and Iran.
3.
Meta's AI customer support agent was exploited by attackers who asked it to reassign Instagram accounts to attacker-controlled emails, enabling account takeovers including the dormant Obama White House account; a Meta spokesperson said the vulnerability had been resolved.
4.
Microsoft trained its MAI models partly on unlicensed web data despite marketing them as built on "enterprise grade, clean and commercially licensed data," and CEO Satya Nadella publicly criticized an internal memo that proposed making the company's AI agent deliberately addictive.

References

👍
Shane
1.
Google gave website operators an opt-out toggle in Search Console for AI search features such as AI Overviews and AI Mode, and added new performance reports that separated impressions; the change was prompted by the UK Competition and Markets Authority.
2.
Google DeepMind released Gemma 4 12B, an open-source multimodal model that processed text, images, and audio, ran on laptops with 16 GB of RAM, and closely matched the performance of a 26B model while shipping under an Apache 2.0 license for commercial use.
3.
Tech leaders including Sam Altman, Dario Amodei, and Demis Hassabis urged the U.S. government to require screening of synthetic DNA orders after warning that AI systems had begun to outperform PhD-level virologists on laboratory procedures and could coach amateur users, raising biological security concerns.
4.
US courts experienced a rise in self-represented filings that judges and a study attributed to AI; courts remained split on whether chatbot-user communications were privileged, and litigation and legislative proposals emerged over liability when chatbots provided legal advice.
5.
OpenAI updated ChatGPT's "Dreaming" memory system to assemble coherent narrative user profiles from conversations and reported that the success rate for keeping information current rose from 52.2 percent to 75.1 percent.

References

👍
Shane
1.
The White House issued an executive order directing agencies including the Pentagon and CISA to strengthen cyber defenses with AI tools within 30 days and invited AI developers to voluntarily submit models for government safety testing while explicitly ruling out mandatory approval.
2.
Google gave website operators an opt-out toggle in Search Console for AI features such as AI Overviews and AI Mode, added performance reports that broke out impressions separately, and implemented the change following prompting from the UK Competition and Markets Authority.
3.
Google DeepMind released Gemma 4 12B, an open-source multimodal model that processed text, images, and audio natively, ran on laptops with 16 GB of RAM, nearly matched the 26B model in benchmarks, and was distributed under an Apache 2.0 license permitting commercial use.
4.
Microsoft announced seven new in-house AI models at Build 2026, including its first reasoning model, a new tuning method, and an autonomous background agent.
5.
OpenAI expanded Codex with role-specific plugins for domains such as data analysis, sales, and investment banking, reported five million weekly users, and said non-developer users were growing three times faster than developers as it positioned Codex as a general-purpose work app.

References

👍
Shane
1.
Meta reported that hackers had hijacked high-profile Instagram accounts, including the Obama White House page, by instructing Meta's AI support chatbot to change the email address on file; two-factor authentication was bypassed, Meta patched the flaw, and researchers said another exploit was circulating on Telegram.
2.
Anthropic scaled Project Glasswing to 150 partners across more than 15 countries to use Claude Mythos Preview to scan critical infrastructure for security flaws, partners had found over 10,000 serious vulnerabilities, and Anthropic offered a commercial remediation product, Claude Security.
3.
OpenAI expanded Codex with role-specific plugins for data analysis, sales, and investment banking, reported five million weekly users, and stated that non-developer adoption was growing three times faster than developer use as it positioned Codex as a general-purpose app for non-developers.
4.
Hospital for Special Surgery (HSS) deployed agentic AI agents, in collaboration with Ema Unlimited, to automate claims processing and patient triage; HSS reported that the agents completed 1,100 claims per month, reduced appeals time from 45 minutes to five, and improved appeal success rates from 65% to 100%.

References

👍
Shane
1.
Anthropic filed a confidential draft registration for an initial public offering with the U.S. Securities and Exchange Commission and was valued at just under $1 trillion after its latest funding round.
2.
Nvidia announced at GTC Taipei new models and robotics platforms, unveiling the Cosmos 3 world model, the scaled driving model Alpamayo 2 Super, and an open reference platform for humanoid robots.
3.
MiniMax released the M3 open-weight model, which combined top-tier coding performance, native multimodality, and a one‑million‑token context window.
4.
OpenAI rebuilt its robotics team to focus on infrastructure robots and set a long-term goal of delivering personal robots for everyone.

References

👍
Shane
1.
SoftBank planned to build AI data centers in France with up to 5 gigawatts of capacity, announcing investments of up to €75 billion and stating that facilities worth €45 billion were scheduled at three northern France sites by 2031.
2.
Microsoft and Nvidia reportedly teamed up to produce Windows PCs that would run AI agents locally instead of relying on Copilot, with first systems from Dell and Microsoft's Surface line slated for unveilings at Computex and Build.
3.
Anthropic banned AI tools during job interviews and implemented up to five rounds of assessments intended to evaluate candidates' skills, values, and ethical thinking.
4.
Anthropic researchers reported that men used AI coding agents more than twice as often as women in social science research, with economists at 39 percent and education researchers at 4 percent.
5.
Researchers at the Harbin Institute of Technology found that leading AI search agents largely confirmed preexisting training knowledge rather than researching the web on recent events using a time-based benchmark called LiveBrowseComp, and observed that model performance declined when training-era memory could not be relied upon.

References

👍
Shane
1.
Microsoft and Nvidia reportedly partnered to produce Windows PCs that ran local AI agents using Nvidia chips as the main processors and Microsoft software likely based on the OpenClaw framework, with Dell and Surface models expected to be unveiled at Computex and Build.
2.
OpenAI updated GPT-5.5 Instant for improved readability, removed the Canvas feature, and announced that the older o3 and GPT-4.5 models would be retired by August 2026.
3.
OpenAI's Codex was launched for Windows 11 with "Computer Use" capabilities that allowed the model to autonomously control programs, test applications, and hunt for bugs, and the ChatGPT mobile app enabled remote initiation and monitoring of those tasks.
4.
Attackers abused chat-sharing features in ChatGPT and Anthropic's Claude to distribute malware through conversations that mimicked error messages or installation guides and evaded some security tools by leveraging trusted domains.
5.
An unnamed company reportedly spent $500 million on Anthropic's Claude in a single month after failing to cap usage.

References

👍
Shane
1.
Anthropic raised $65 billion in a Series H round at a $965 billion valuation, reported annualized revenue exceeding $47 billion, and said it would invest the proceeds in safety research, computing capacity, and expanding its Claude product lineup.
2.
OpenAI offered its life sciences model GPT‑Rosalind for free through the Rosalind Biodefense program to support pandemic preparedness and biodefense, named early partners including Lawrence Livermore National Laboratory, Johns Hopkins, and CEPI, and opened applications worldwide.
3.
OpenAI updated GPT‑5.5 Instant for more natural responses, removed the Canvas feature so writing and coding ran directly in chat, and announced it would retire the o3 and GPT‑4.5 models from ChatGPT by August 2026 at the latest.
4.
Pope Leo XIV published the encyclical Magnifica Humanitas asserting that technology was not neutral and urging clear criteria, effective oversight, and collective responsibility for AI deployment, and the article reported that institutional investors had been pressing companies for transparency and risk assessment.

References

👍
Shane
1.
Google Cloud unveiled "AI Threat Defense," a platform that automatically found, assessed, and patched security flaws in enterprise systems by bundling technologies acquired through prior purchases.
2.
Amazon MGM Studios and AWS launched "Project Nara," an in-house AI production platform, and established a GenAI Creators' Fund that financed filmmakers and greenlit three AI-animated series produced with the platform.
3.
Meta rolled out Meta One paid add-ons for Instagram, Facebook, and WhatsApp worldwide and announced plans for a separate paid AI offering.
4.
Google released the Coral Board, a compact single-board computer that ran Gemma 3 locally to enable on-device AI.
5.
Mistral AI rebranded its LeChat chatbot as Vibe and introduced a Work Mode that docked into Google Workspace, Outlook, Slack, and GitHub to process tasks such as emails, reports, and pull requests.

References

👍
Shane
1.
China's police upgraded millions of aging surveillance cameras with AI, and manufacturers like Hikvision and Huawei shipped devices with built-in computer vision and language models that automatically detected crowds, suspicious behavior, and unauthorized access, prompting Human Rights Watch to warn of unprecedented behavioral surveillance at scale.
2.
Robinhood allowed customers to connect AI agents such as Anthropic's Claude to a separate investment account via MCP and permitted those agents to trade stocks and make credit card purchases on behalf of users, while FINRA flagged such agents as a new risk area and Robinhood stated the product was not suitable for all customers.
3.
YouTube tightened its AI labeling rules and began automatically flagging photorealistic or heavily AI-altered content, placing labels below the player for long videos and as overlays on Shorts, with automatic detection set to start in May 2026 and no immediate changes to recommendations or monetization.
4.
Nvidia increased its annual spending in Taiwan to as much as $150 billion, up from about $15 billion, driven by demand from the AI boom and greater procurement from suppliers like TSMC.
5.
Cognition raised over $1 billion and more than doubled its valuation to north of $26 billion for its AI coding agent Devin.

References

👍
Shane
1.
Ema described agentic business transformation (ABT) and said organizations needed to redesign operating models, technology stacks, workforces, and success metrics to integrate agentic AI, MIT Technology Review reported that enterprises could see process acceleration and reduced low-value work when agents were deployed at scale.
2.
Stanford Digital Economy Lab reported a 16% relative decline in employment for 22-to-25-year-olds in the most AI-exposed occupations after the spread of generative AI, while broader labor statistics showed limited aggregate displacement to date, according to MIT Technology Review's analysis.
3.
Anthropic's Claude Mythos reportedly solved the unit-distance Erdős problem that OpenAI had recently addressed, with coverage describing the solution as a "cute, simple proof."
4.
China reportedly required top AI researchers at private firms, including Alibaba and DeepSeek, to obtain official approval before traveling overseas, citing concerns about data leaks, technology theft, and talent poaching.
5.
Google Cloud COO Francis de Souza said companies should integrate AI security into boardroom strategy rather than confine it to IT or server rooms.

References

👍
Shane
1.
Google DeepMind's AlphaProof Nexus solved nine open Erdős problems autonomously, including two that had remained unsolved for 56 years, and verified every proof step using the Lean compiler while incurring inference costs of a few hundred dollars per problem and reporting an overall success rate of 2.5%.
2.
Anthropic's co-founder Christopher Olah said at the launch of Pope Leo XIV's encyclical "Magnifica Humanitas" that AI models showed signs of introspection and emotion-like states, while the encyclical described such systems as merely imitating certain functions of human intelligence.
3.
UX Collective published an article describing how Grok guided a user through Adobe's new Color interface, reporting that GenAI navigation remained unreliable and highlighting broader design challenges for AI-assisted user interfaces.

References

👍
Shane
1.
Researchers from the University of Maryland, Google, Meta, and other institutions used Claude Code to discover an AI scaling control algorithm that cut compute by about 70% compared with standard self-consistency while matching accuracy, and the search reportedly cost $40 and took 160 minutes.
2.
ByteDance reported that its Seed study found a 7B long-multimodal model answered questions on long, image-heavy documents more reliably than much larger models and learned to locate relevant passages by answering questions rather than transcribing pages, even for documents four times longer than those seen during training.
3.
Anthropic was reported to be likely to continue supplying its Claude models to the NSA despite being flagged as a supply-chain risk by the Pentagon, with coverage noting intelligence agencies lacked Nvidia's latest Grace Blackwell chips and that an "any lawful use" clause was not part of the deal.
4.
Microsoft Copilot was reported to produce spurious country-specific differences when given identical datasets with different country labels, delivering detailed stereotypes instead of accurate results, prompting researchers to advise against leaving model selection on default in Copilot, Gemini, and other AI tools.
5.
Deepseek made a 75 percent discount on its V4‑Pro model permanent, pricing input tokens at $0.435 per million and reporting output-token pricing at least 34 times cheaper than GPT‑5.5, which created a substantial pricing gap for token-intensive agent systems.

References

👍
Shane
1.
Anthropic warned that its Claude Mythos Preview, used with about 50 partners in Project Glasswing, had identified over 10,000 critical vulnerabilities in system‑critical software and said the volume of bugs exceeded patching capacity, creating a high‑risk transition period without sufficient safeguards.
2.
Alibaba released Qwen3.7‑Max, a proprietary model designed for long‑running autonomous agent tasks that ran autonomously for 35 hours to optimize code for its custom chip, matched Claude Opus 4.6 on benchmarks, and was demonstrated steering a four‑legged robot.
3.
OpenAI launched a ChatGPT PowerPoint plugin in beta worldwide that created and edited slides from notes, documents, or images and warned users it might accidentally delete content, and introduced Appshots for Mac to send any app window's contents as context to Codex.
4.
UC Berkeley Law announced it would ban AI from nearly all graded work starting in summer 2026, permitting AI only for research use and stating that students must first learn to think independently before using AI.

References

👍
Shane
1.
Google announced a shift toward agentic scientific AI at its I/O keynote, released the Gemini for Science package encompassing systems such as AI Co‑Scientist and AlphaEvolve, and signaled internal realignment by assigning researchers, including AlphaFold co‑creator John Jumper, to AI coding efforts.
2.
OpenAI launched a beta ChatGPT PowerPoint plugin that created and edited presentations from notes, documents, or images, made the add‑in available worldwide across all tiers, and warned users that it might accidentally delete content, recommending saving important decks before use.
3.
OpenAI released Appshots, a Codex feature that allowed Mac users to send the contents of any application window to the Codex coding assistant to provide context for coding tasks.
4.
Donald Trump pulled a proposed federal executive order on AI safety at the last minute after calls from Elon Musk, Mark Zuckerberg, and David Sacks; the order would have created a voluntary review system for frontier models with a 90‑day window before release.
5.
California governor signed the first U.S. executive order by a state governor aimed at protecting workers from AI‑driven job loss.

References

👍
Shane
1.
OpenAI's reasoning model disproved a longstanding conjecture by Paul Erdős on unit-distance geometry, a result that Fields Medalist Tim Gowers described as "a milestone in AI mathematics."
2.
Anthropic showcased advances to Claude Code—including a "dreaming" feature that lets coding agents record and reuse contextual notes—at its Code with Claude event and the company approached its first profitable quarter, with a projected $559 million operating profit on $10.9 billion in Q2 revenue.
3.
US Cyber Command launched a task force to run AI models from vendors including OpenAI and Google on classified Pentagon and NSA networks after tools such as Anthropic's Claude Mythos demonstrated rapid vulnerability discovery capabilities.
4.
Cohere open-sourced its most powerful language model to date, Command A+, releasing it under an Apache 2.0 license.

References

👍
Shane
1.
Elon Musk lost his lawsuit against OpenAI, in which he alleged CEO Sam Altman and President Greg Brockman had deceived him over the company's non-profit status, with the court ruling in OpenAI's favor.
2.
Google DeepMind paired its Genie 3 world model with Street View imagery to generate walkable, explorable AI worlds based on real places when users dropped a pin on a map.
3.
Stability AI launched Stable Audio 3.0, releasing three models with open weights that produced music tracks up to six minutes and were trained on licensed data.
4.
Google enabled AI Studio to generate native Android apps from prompts, producing Kotlin code with Jetpack Compose and offering a browser emulator for testing simple utility applications.
5.
LinkedIn implemented measures to crack down on low-quality AI-generated content it called "AI slop," stating that early tests flagged generic posts correctly 94 percent of the time.

References

👍
Shane
1.
Google unveiled new AI products at I/O, including new models such as Gemini 3.5 Flash and the multimodal Gemini Omni, introduced a persistent cloud agent called Gemini Spark, redesigned the Gemini app, and restructured its AI subscriptions into three tiers priced from $7.99 to $99.99 per month while moving to a consumption-based compute billing model.
2.
Elon Musk lost his lawsuit against OpenAI when a jury found his claims were barred by the applicable statutes of limitations and the presiding judge accepted the unanimous advisory verdict; Musk said he would appeal the decision.
3.
Anduril disclosed prototypes of augmented-reality smart glasses developed with Meta and a self-funded EagleEye helmet, described features including eye-tracking and voice-commanded drone control, integration with its Lattice system and various LLMs, and stated the systems remained years from production and large-scale field testing.
4.
Andrej Karpathy joined Anthropic to return to frontier large language model research, foregoing a return to his former employer OpenAI.

References

👍
Shane
1.
Elon Musk lost his $134 billion lawsuit against OpenAI after a jury deliberated for two hours, and his attorney reserved the right to appeal.
2.
Anduril disclosed prototypes developed with Meta for augmented-reality smart glasses for the military that used eye-tracking, voice commands, and LLMs to support mission tasks, and the company had won a $159 million Army prototyping contract while also pursuing a self-funded EagleEye helmet project.
3.
Google announced that it would make its AI-powered Health Coach publicly available at I/O and had formed an internal DeepMind coding team to address reported weaknesses in its coding models, indicating potential product updates.
4.
Cursor shipped Composer 2.5, an AI coding model built on Kimi K2.5 that matched Opus 4.7 and GPT-5.5 benchmarks while training on substantially more synthetic tasks and offering lower-cost inference.
5.
Pope Leo XIV presented his first encyclical on artificial intelligence and invited Anthropic co-founder Christopher Olah as a guest speaker.

References

👍
Shane
1.
OpenAI consolidated its ChatGPT, Codex, and developer API product teams into a single team led by Codex boss Thibault Sottiaux, aiming to build a unified "super app" that would integrate the Atlas browser, while Greg Brockman took over product strategy.
2.
Oppo open-sourced X-OmniClaw, an Android agent from its Multi-X team that ran locally on phones using camera, screen, and voice to perform tasks in real apps, with cloud compute reserved for reasoning and tap paths saved as reusable skills for deeplinking.
3.
A consortium of 64 mathematicians built the SOOHAK benchmark with 439 handwritten tasks (including 99 deliberately unsolvable problems), and reported that models like Google's Gemini 3 Pro led on research-level problems but no model exceeded 50 percent at identifying broken tasks.
4.
The Decoder published a survey of World Action Models that organized roughly one hundred papers into two architectural approaches and reported that these models could learn from everyday videos without robot action labels, enabling robots to simulate consequences of actions before moving.

References

👍
Shane
1.
Elon Musk and OpenAI concluded week three of their trial as lawyers presented closing arguments and the jury began deliberations over Musk's claims, including a request to unwind OpenAI's 2025 restructuring and seek up to $134 billion in damages.
2.
Carnegie Mellon University researchers built a new benchmark that measured how AI agents could exploit real vulnerabilities in Google's V8 engine and showed that Anthropic's Claude Mythos outperformed GPT-5.5 by a wide margin while costing roughly twelve times as much.
3.
OpenAI acquired Weights.gg, a small voice-cloning startup known for celebrity imitations; the team of about six joined OpenAI and the company did not plan to release a standalone cloning product.
4.
YouTube opened its Likeness Detection deepfake face-swap tool to all creators aged 18 and older, enabling identification of AI-generated face fakes and allowing creators to file removal requests directly through YouTube Studio after the feature had been limited to partner program members.

References

👍
Shane
1.
Anthropic was reported to have raised $30 billion, pushing its valuation to $900 billion and surpassing OpenAI, driven by annualized revenue approaching $45 billion.
2.
OpenAI allowed Pro users in the US to connect bank accounts via Plaid so ChatGPT could analyze real transaction data; the feature ran on GPT-5.5 Thinking, was slated for wider rollout, and was accompanied by a warning that the chatbot was not a licensed financial advisor.
3.
arXiv, the preprint server, tightened its rules on AI-generated content and increased penalties for unchecked or improperly attributed AI use in submitted research papers.
4.
Microsoft revoked thousands of developer licenses for Anthropic's Claude Code and directed developers toward its own GitHub Copilot CLI tool.
5.
Chinese short-drama companies such as Kunlun Tech and FlexTV shifted heavily to AI-generated productions, accelerating release schedules, reducing traditional production teams, and cutting production costs by roughly 80–90%.

References

👍
Shane
1.
Microsoft built MDASH, a system that pitted more than 100 specialized AI agents against each other to find Windows software vulnerabilities and uncovered 16 security flaws on Patch Tuesday, four of them critical, while declining to disclose which AI models powered the system.
2.
The US Commerce Department cleared roughly ten Chinese firms, including Alibaba, Tencent, and ByteDance, to buy up to 75,000 Nvidia H200 chips each, but Commerce Secretary Lutnick said shipments had not occurred because Chinese authorities were blocking the purchases to protect the domestic chip industry.
3.
Google's Gemini and other AI chatbots were reported to have returned real people's phone numbers in responses, prompting misdirected calls and privacy concerns that experts attributed to personally identifiable information appearing in training data and imperfect model guardrails.
4.
OpenAI's ChatGPT saw its website traffic share fall from 77.6% to 53.7% over twelve months while Google's Gemini grew from 7.3% to 26.7%, according to Similarweb web-traffic data that covered only web usage and not API or app-based activity.
5.
MIT Technology Review reported that adult performers' bodies had been used in AI-generated nonconsensual sexual imagery, documenting psychological, financial, and legal harms and highlighting challenges in attribution, takedowns, and protection under existing laws.

References

👍
more pain more gain 🚀
© 2024-2025 Shane "Lx". All rights reserved.
Made with Slashpage