Artificial Intelligence on Pulse — Page 13

Anthropic Brings Claude Fable 5 Back Globally After US Lifts Export Controls

Commerce Secretary Howard Lutnick withdrew the June 12 export-control license requirement on Claude Fable 5 and Mythos 5 on June 30, ending a roughly three-week global suspension triggered when Amazon researchers flagged a jailbreak technique in Fable 5.

SpaceX Has an AI Device Prototype, Report Says — Then Musk Publicly Denies It

AIDenied by Musk

SpaceX Has an AI Device Prototype, Report Says — Then Musk Publicly Denies It

TechCrunch reported July 1 that SpaceX has been developing an AI-focused hardware device that 'sure sounds phone-ish,' citing internal sourcing on a project exploring a Starlink-connected, xAI-powered consumer device. Elon Musk publicly denied the specific report within hours, per The Verge, even as SpaceX's fresh public-market currency (following its roughly $2.1 trillion day-one IPO valuation) and its $60 billion all-stock acquisition of Cursor make a hardware-plus-AI push entirely plausible.

Restaurants Can Now Take Orders Directly From ChatGPT and Claude Through Square's New Integration

AINo added commission

Restaurants Can Now Take Orders Directly From ChatGPT and Claude Through Square's New Integration

Square launched an integration on July 1 letting restaurants accept orders placed directly through ChatGPT and Claude, alongside a companion Alexa+ voice-ordering feature, with no setup required for eligible sellers and no added marketplace commission beyond Square's standard processing fee. Orders route through Square's existing 'Order by Cash App' infrastructure straight into point-of-sale and kitchen-display systems, positioning the integration as a low-fee alternative to delivery-aggregator commissions that can run around 30%.

AI0% marketplace fee vs. 25-30% aggregators

Square Lets Restaurants Take Orders Directly From ChatGPT and Claude, No Marketplace Fee

Square Lets Restaurants Take Orders Directly From ChatGPT and Claude, No Marketplace Fee

Square launched a ChatGPT app and Claude plugin letting consumers discover restaurants and place orders directly inside those chat interfaces, automatically opted in for any US Food & Beverage seller with an activated Square Online Ordering profile, VentureBeat reported July 1. Restaurants pay only Square's standard processing fee of roughly 2.9% plus 30 cents per transaction, rather than the 25-30% cut delivery aggregators typically charge.

Square Lets Restaurants Take Orders Straight From ChatGPT and Claude, With No New Fees

AIZero new fees

Square Lets Restaurants Take Orders Straight From ChatGPT and Claude, With No New Fees

Square launched new ChatGPT and Claude integrations on July 1 that let consumers discover restaurants and place orders directly inside those AI assistants, with sellers opted in automatically at no additional setup cost and zero added marketplace fees beyond Square's standard online-ordering rate of roughly 2.9-3.3% plus $0.30 per transaction. Orders route instantly into a restaurant's existing Square Point of Sale and Kitchen Display System.

Cloudflare's New Default Policy Forces AI Crawlers to Pay Publishers for Content

AIDefault policy shift

Cloudflare's New Default Policy Forces AI Crawlers to Pay Publishers for Content

Cloudflare announced on July 1 that starting September 15, its default settings will block 'mixed-use' AI crawlers from scraping ad-supported content on new customer sites and existing free-tier sites unless the AI company pays, building on its 2025 Pay Per Crawl marketplace with a new 'Pay Per Use' model that compensates publishers when their content actually drives value in an AI answer. CEO Matthew Prince cited crawl-to-referral ratios as extreme as 73,000-to-1 for Anthropic and 1,700-to-1 for OpenAI, versus 14-to-1 for Google.

Serial Founder Bhavin Turakhia Bets $30M of His Own Money on an AI Alternative to Microsoft Office

AI$30M self-funded

Serial Founder Bhavin Turakhia Bets $30M of His Own Money on an AI Alternative to Microsoft Office

Indian serial entrepreneur Bhavin Turakhia is self-funding $30 million into Neo, a new enterprise software company betting that workplace productivity tools designed before the AI era need to be rebuilt from scratch rather than upgraded with bolted-on chatbots. Turakhia, who has previously co-founded Directi, Radix, Titan and banking software firm Zeta — largely bootstrapping each with his own capital before bringing in outside investors — told TechCrunch he sees the AI shift as significant enough to justify a ground-up rebuild of office software.

Autonomous Vehicle Hype Is Back — Humble Robotics Is Bringing It to Freight

AI$24M disclosed

Autonomous Vehicle Hype Is Back — Humble Robotics Is Bringing It to Freight

Humble Robotics, a cabless, fully autonomous electric freight hauler startup founded by Eyal Cohen, emerged from stealth in April 2026 with $24 million in funding and is now the subject of renewed investor interest in autonomous vehicles, TechCrunch reported July 1. Cohen — a veteran of Otto (acquired by Uber) and Pronto — says vision-model-based perception has finally caught up to the vision for self-driving trucks, building what he calls the simplest possible robotics platform rather than retrofitting a human-driven truck design.

Anthropic Brings Claude Fable 5 Back Globally After US Lifts Export Control Order

AIGlobal access restored

Anthropic Brings Claude Fable 5 Back Globally After US Lifts Export Control Order

The US Department of Commerce cleared Anthropic to restore global access to Claude Fable 5 and Mythos 5 starting July 1, after suspending both models in mid-June under an export control directive citing national security. Access resumes on the Claude Platform, Claude.ai, Claude Code, Claude Cowork, and will be re-enabled on AWS, Google Cloud and Microsoft Foundry, with up to 50% of weekly usage included free through July 7.

AI~1,500 hours/week saved

Morgan Stanley Cut Its Riskiest Reconciliation Job in Half by Making Its Agents Less Autonomous

Morgan Stanley Cut Its Riskiest Reconciliation Job in Half by Making Its Agents Less Autonomous

Morgan Stanley deployed an internal agentic system called FIXR to handle profit-and-loss reconciliation -- one of banking's most accuracy-critical, deadline-driven workflows -- cutting the process from up to six hours to two-to-three hours per book, saving roughly 1,500 hours per week across about 100 controllers, VentureBeat reported June 30. Managing Director Todd Johnson said the gains came from deliberately limiting agent autonomy and keeping humans tightly in the loop rather than maximizing how much the system operates independently.

Anthropic Launches Claude Sonnet 5 at a Steep Discount as It Races Toward a Blockbuster IPO

AI60% intro discount

Anthropic Launches Claude Sonnet 5 at a Steep Discount as It Races Toward a Blockbuster IPO

Anthropic released Claude Sonnet 5 on June 30 at introductory pricing of $2 per million input tokens and $10 per million output tokens through August 31 — roughly 60% cheaper than standard rates — delivering performance close to Opus 4.8 at a fraction of the cost. The launch comes weeks after Anthropic confidentially filed a draft S-1 following a $65 billion Series H that valued the company near $965 billion.

AI~60% discount vs. flagship

Anthropic Launches Claude Sonnet 5 at a Steep Discount as It Races Toward an IPO

Anthropic Launches Claude Sonnet 5 at a Steep Discount as It Races Toward an IPO

Anthropic launched Claude Sonnet 5 on June 30, pricing it at $2 per million input tokens and $10 per million output tokens through August 31 — roughly 60% cheaper than its flagship Opus 4.8 model — while claiming near-Opus-level performance. The move lands weeks after Anthropic confidentially filed IPO paperwork with the SEC on June 1 and the same week Together AI raised $800 million specifically on the thesis that cheaper, open models are eating into closed frontier-lab pricing.

Google's Gemini Omni Flash Hits the API, Turning Enterprise Video Production Into a Conversation

AI$0.10/sec video

Google's Gemini Omni Flash Hits the API, Turning Enterprise Video Production Into a Conversation

Google rolled out Gemini Omni Flash — the first model in its new multimodal 'Omni' family — to developers via the Gemini API, Google AI Studio and the Gemini Enterprise Agent Platform. The model generates and conversationally edits 720p video through natural-language prompts at $0.10 per second of output, putting a 10-second clip at roughly a dollar, and every clip carries Google's SynthID watermark.

GitHub Copilot's Usage-Based Billing Closes Its First Full Cycle — and Enterprise Bills Are Landing 25-40% Higher

AI25-40% bill hike

GitHub Copilot's Usage-Based Billing Closes Its First Full Cycle — and Enterprise Bills Are Landing 25-40% Higher

June 30 marked the close of the first full billing month under GitHub Copilot's new usage-based pricing (launched June 1). Early enterprise reports show average bills 25-40% higher than the flat-fee equivalent, forcing engineering leaders to grapple with per-request AI cost governance for the first time.

Morgan Stanley Cut Its Riskiest Reconciliation Job in Half by Making Its AI Agents Less Autonomous

AI50% reduction

Morgan Stanley Cut Its Riskiest Reconciliation Job in Half by Making Its AI Agents Less Autonomous

Morgan Stanley deployed an internal agentic system called FIXR into profit-and-loss reconciliation, one of banking's most accuracy-critical, deadline-driven workflows, and cut the manual work in half — not by giving the system more autonomy, but by deliberately limiting it. Human controllers stay tightly in the loop, and their judgment calls get converted into fixed, repeatable rules the system applies going forward rather than being left to model discretion.

Google Ships Nano Banana 2 Lite, a Faster and Cheaper Image Generator

AINew model tier

Google Ships Nano Banana 2 Lite, a Faster and Cheaper Image Generator

Google released Nano Banana 2 Lite -- internally Gemini 3.1 Flash-Lite -- as a faster, cheaper image-generation model available through the Gemini API, according to VentureBeat. The launch fits Google's pattern of shipping lightweight, low-cost model variants to win high-volume, price-sensitive workloads the same week Anthropic discounted Claude Sonnet 5, intensifying the price war across every layer of the AI stack.

The DeepMind Trio Who Built a Poker AI Are Now Making Money for Quant Hedge Funds

AITalent migration story

The DeepMind Trio Who Built a Poker AI Are Now Making Money for Quant Hedge Funds

TechCrunch profiled a startup founded by three former DeepMind researchers who built poker-playing AI systems and are now applying the same game-theoretic, imperfect-information decision-making techniques to quantitative trading for hedge funds, in a story published June 30. The team's background in adversarial, hidden-information games like poker translates directly into modeling markets where other participants' information and intentions are unknown.

Meituan Open-Sources LongCat-2.0, a 1.6T Near-Frontier Model Trained Entirely on Chinese Chips

AI1.6T parameters

Meituan Open-Sources LongCat-2.0, a 1.6T Near-Frontier Model Trained Entirely on Chinese Chips

Meituan open-sourced LongCat-2.0, a 1.6-trillion-parameter agentic coding model trained entirely on Chinese-made chips, marking a significant milestone in China's push toward AI self-sufficiency amid continued export restrictions on advanced Western AI hardware. The release positions LongCat-2.0 as near-frontier in capability while proving domestic chip infrastructure can now train models at genuinely large scale.

AIAI contributions banned

Godot Game Engine Bans AI-Authored Code Contributions

Godot Game Engine Bans AI-Authored Code Contributions

The open-source Godot game engine announced it will no longer accept code contributions authored by AI, citing concerns that heavy AI users submitting pull requests can't be trusted to understand their own code well enough to maintain or fix it. The policy is one of the most explicit rejections yet by a major open-source project of AI-generated contributions, running counter to the broader industry push toward AI-assisted coding.

OKX Opens a Marketplace Where AI Agents Hire and Pay Each Other

AIAgent marketplace

OKX Opens a Marketplace Where AI Agents Hire and Pay Each Other

Crypto exchange OKX launched OKX AI, a marketplace where autonomous AI agents can hire one another, settle payments in stablecoins and build portable on-chain reputations -- opening to developers after a closed beta with 50 early service providers. It builds on OKX's Agent Payments Protocol, which lets agents negotiate, set up escrow and run pay-per-use billing without human intervention, a bet that 'agentic commerce' becomes a trillion-dollar market.

Meituan Open-Sources LongCat-2.0, a 1.6T Agentic Coding Model Trained Entirely on Chinese Chips

AI1.6T parameters

Meituan Open-Sources LongCat-2.0, a 1.6T Agentic Coding Model Trained Entirely on Chinese Chips

Meituan open-sourced LongCat-2.0, a 1.6-trillion-parameter mixture-of-experts coding model with a one-million-token context window, releasing it under the permissive MIT license on GitHub and Hugging Face. The release unmasked 'Owl Alpha,' the anonymous stealth model that had quietly topped global developer charts on OpenRouter for two months -- and revealed that the entire system was trained and served on a 50,000-card domestic compute cluster with no Nvidia or AMD accelerators. It is the clearest evidence yet that China can build a near-frontier AI stack end to end despite US export controls.

TIDAL Cracks Down on AI Music by Cutting Off Monetization

AIAI music demonetized

TIDAL Cracks Down on AI Music by Cutting Off Monetization

TIDAL, the streaming service owned by Block, will stop paying royalties on AI-generated music -- demonetizing tracks its systems detect as AI-made without banning them outright. The policy is one of the clearest moves yet by a major platform to protect human artists' royalty pool as a flood of synthetic tracks threatens to dilute streaming payouts.

OpenAI Teases Dedicated Hardware for Its Codex Coding Agent

AIHardware tease

OpenAI Teases Dedicated Hardware for Its Codex Coding Agent

OpenAI is teasing new hardware built around Codex, its AI coding agent, signaling ambitions to move beyond software and into purpose-built devices for developers. The hints fold into OpenAI's broader hardware push -- the secretive 'io' effort it built with former Apple design chief Jony Ive -- and suggest the company sees agentic coding as a flagship use case worth its own physical form factor.

DeepSeek and Peking University Open-Source DSpark, Speeding LLM Inference by Up to 85%

AIUp to 85% faster

DeepSeek and Peking University Open-Source DSpark, Speeding LLM Inference by Up to 85%

DeepSeek and Peking University open-sourced DSpark, a speculative-decoding framework that boosts per-user LLM generation speed by 60-85% -- and up to 661% throughput under tight latency constraints -- without hardware upgrades or model retraining. Released under the MIT license and already live in DeepSeek's V4-Flash and V4-Pro production models, it is another Chinese efficiency breakthrough aimed at slashing the cost of serving AI.

Google Makes Gemini's Personalized AI Image Generation Free for All U.S. Users

AIFree for U.S.

Google Makes Gemini's Personalized AI Image Generation Free for All U.S. Users

Google opened Gemini's personalized AI image generation -- powered by its 'Nano Banana' model and able to pull context from a user's Gmail, Google Photos, YouTube and Search -- to all U.S. users for free, after previously gating it behind Plus, Pro and Ultra subscriptions. With Gemini past 750 million monthly active users, making a premium, data-personalized feature free is an aggressive distribution play against OpenAI and a bet that personal context is Google's structural edge.

Lawmakers Move to Ban AI Companies From Selling Your Health and Location Data

AIData-privacy bill

Lawmakers Move to Ban AI Companies From Selling Your Health and Location Data

Democratic lawmakers introduced the Health and Location Data Protection Act, which would bar companies -- including AI firms -- from selling or sharing Americans' sensitive health and precise-location data. Backed by Senator Elizabeth Warren and Representative Mary Gay Scanlon, the bill targets a data-broker economy that AI's data hunger has made more lucrative and more dangerous.

Anthropic and Gov. Newsom Strike Deal to Give California Government Claude at Half Price

AI~50% gov discount

Anthropic and Gov. Newsom Strike Deal to Give California Government Claude at Half Price

Anthropic and California Governor Gavin Newsom announced an agreement giving all state agencies and local governments discounted access to Claude -- reported at roughly half price -- plus training and support. The deal positions California as a model for 'responsible' government AI adoption even as the Pentagon has frozen Anthropic out over its safety conditions.

Wix's Base44 Launches Its Own Model, Base1, in a Bet on Defensibility

AIBase1 model

Wix's Base44 Launches Its Own Model, Base1, in a Bet on Defensibility

Base44, the vibe-coding platform Wix acquired for $80 million a year ago, began rolling out its own large language model, Base1, trained on a dataset generated from tens of millions of real user interactions on its platform. The move is a direct play for defensibility -- owning the model gives Base44 control over compute and inference spend and a structurally stronger margin profile, as the debate intensifies over whether companies built atop someone else's frontier model can endure.

Arena, the AI Leaderboard Everyone Uses, Is Now a $100M Business

AI$100M ARR

Arena, the AI Leaderboard Everyone Uses, Is Now a $100M Business

Arena, the crowdsourced model-evaluation platform born as Chatbot Arena at UC Berkeley, hit $100 million in annualized run-rate revenue just eight months after launching its paid 'AI Evaluations' service. Its free public leaderboard -- built on more than 10 million head-to-head user votes -- has become the industry's default benchmark, and the company has parlayed that trust into a $1.7 billion valuation.

Cursor Launches a Mobile App to Steer Your Coding Agent on the Go

AIMobile agent control

Cursor Launches a Mobile App to Steer Your Coding Agent on the Go

Cursor, the AI coding tool now owned by SpaceX after its record $60 billion acquisition, released a mobile app that lets developers direct, monitor and approve their coding agents from a phone. The launch reflects how AI software development is shifting from line-by-line editing toward delegating and supervising autonomous agents that work in the background.

AI· Jun 27, 2026

SoftBank's CEO Isn't Alone in Doubting Elon Musk's Orbital Data Center Hype

AIOrbital compute debate

SoftBank's CEO Isn't Alone in Doubting Elon Musk's Orbital Data Center Hype

As Elon Musk pitches data centers in orbit to power the next wave of AI compute, skepticism is mounting -- and SoftBank's Masayoshi Son, himself one of AI's biggest spenders, is among the doubters. Critics question the physics, economics and cooling realities of running AI clusters in space, casting the idea as visionary marketing more than near-term infrastructure.

Jun 27, 2026

AI· Jun 27, 2026

Onsemi to Acquire Synaptics for $7B All-Stock Deal, Expanding Edge AI and Robotics Silicon

AI$7B

Onsemi to Acquire Synaptics for $7B All-Stock Deal, Expanding Edge AI and Robotics Silicon

Onsemi agreed to acquire Synaptics in an all-stock transaction valued at approximately $7 billion, announced late June, giving it a broader footprint in edge AI, human-interface chips, IoT and automotive silicon. The deal is one of the largest semiconductor M&A transactions of 2026 and repositions Onsemi against Nvidia's edge ambitions.

Jun 27, 2026

AI· Jun 27, 2026

Claude Code Turned Every Engineer Into Three -- Now Companies Need More Product Thinkers

AI~3x output

Claude Code Turned Every Engineer Into Three -- Now Companies Need More Product Thinkers

A widely shared analysis argues that AI coding tools like Claude Code have multiplied individual engineer output several-fold, shifting the binding constraint in software organizations from writing code to deciding what to build. The implication: the scarce role is no longer the coder but the product thinker who can direct all that newfound capacity toward something valuable.

Jun 27, 2026

OpenAI Unveils GPT-5.6 Sol, Terra and Luna -- but Only for Government-Vetted Preview Partners

AIGated frontier launch

OpenAI Unveils GPT-5.6 Sol, Terra and Luna -- but Only for Government-Vetted Preview Partners

OpenAI unveiled a new GPT-5.6 model family -- code-named Sol, Terra and Luna -- but said access is limited to a small set of preview partners disclosed to the US government, the same gating regime now applied to Anthropic's most capable models. It is the clearest sign yet that frontier-model releases are passing through a national-security filter before reaching the broader market.

DeepSeek Open-Sources DSpark, Claiming 60-85% Faster Generation

DeepSeek released DSpark, a 'semi-parallel' speculative-decoding module for its DeepSeek-V4 Flash and Pro models, and open-sourced DeepSpec, a full codebase for training and evaluating speculative-decoding draft models. The company claims generation runs 60-85% faster on Flash and 57-78% faster on Pro at the same throughput -- a fresh reminder that the open-weight challengers keep narrowing the efficiency gap.

Cerebras Runs OpenAI GPT-5.6 Sol at 750 Tokens per Second, Setting a New Frontier-Model Speed Record

AI750 tok/sec

Cerebras Runs OpenAI GPT-5.6 Sol at 750 Tokens per Second, Setting a New Frontier-Model Speed Record

Cerebras Systems will deploy OpenAI's GPT-5.6 Sol on its wafer-scale WSE-3 chips at up to 750 tokens per second in July — roughly 10x faster than any Nvidia GPU deployment of a frontier model in production. The partnership is a strategic proof point for Cerebras ahead of its planned 2026 IPO.

AIGPT-5.6 (Sol/Terra/Luna)

OpenAI Launches GPT-5.6 -- But the Government Decides Who Gets to Use It

OpenAI Launches GPT-5.6 -- But the Government Decides Who Gets to Use It

OpenAI unveiled its most capable models yet -- GPT-5.6 Sol, Terra and Luna -- but at the Trump administration's request limited the initial release to a small group of trusted partners whose identities are shared with the government. OpenAI complied while publicly objecting, warning that 'this kind of government access process should not become the long-term default.' The move mirrors restrictions placed on Anthropic's frontier models and marks the first time the US government is effectively vetting who can access America's leading AI systems.

OpenAI Poaches Uber's India Chief to Lead Its Biggest Market Outside the US

OpenAI has hired Uber's India head to run its operations in the country, signaling how seriously the company is treating India -- its largest user base outside the United States. The move pairs a frontier AI lab with a seasoned local operator as OpenAI races to convert massive Indian usage into revenue and entrench itself before rivals.

The 'Software Factory' Myth: AI Is Helping Companies Ship Bugs Faster

A widely shared analysis argues that most enterprises adopting AI coding tools to build a 'software factory' are really just shipping bugs faster: AI accelerates code production, but downstream testing, review and CI/CD don't scale with it, so defects and incidents climb. Data cited from Faros AI shows developer throughput up sharply -- but incidents and bugs rising even faster.

A New Agentic Memory Framework Uses 118K Tokens Per Query -- LangMem Burns 3.26M

A newly detailed agentic memory framework reportedly answers queries using about 118,000 tokens, versus roughly 3.26 million for the widely used LangMem approach -- a ~28x efficiency gain. As AI agents take on longer, multi-step tasks, how they store and retrieve memory is becoming a decisive factor in both cost and capability.

Adobe Acquires Topaz Labs to Bring Best-in-Class AI Image Enhancement Native to Creative Cloud

AI~$800M est.

Adobe Acquires Topaz Labs to Bring Best-in-Class AI Image Enhancement Native to Creative Cloud

Adobe agreed on June 25 to acquire Topaz Labs, the pro-grade AI photo and video upscaling and denoising firm favored by photographers and filmmakers. Terms weren't disclosed but industry estimates put the deal in the $700M-$1B range, folding Topaz's Gigapixel, DeNoise and Video AI models directly into Photoshop, Lightroom and Premiere.

Liquid AI's Tiny LFM2.5-230M Beats Models 4x Its Size and Runs Anywhere

Liquid AI released LFM2.5-230M, its smallest model yet, which the company says outperforms models four times its size at data extraction while being small enough to run 'anywhere' -- including phones, laptops and edge devices. The release advances the counter-narrative to ever-larger models: that efficient, specialized small models can win on the tasks enterprises actually run at scale.

OpenAI's Updated GPT-5.5 Instant Gets Better at Shopping and Complex Constraints

OpenAI shipped an updated GPT-5.5 Instant that is better at shopping, handling complex constraints, and understanding user intent -- and it's already live in the API. The release sharpens OpenAI's fast, low-latency tier for agentic and commerce use cases, even as the company's more powerful new GPT-5.6 family sits behind a government-vetted gate.

Meta Revives Facebook's Creator Studio as a Standalone AI Companion App

Meta has relaunched Facebook's Creator Studio as an AI companion app, repositioning a dormant publishing dashboard into an AI-powered assistant for content creators. The move folds Meta's generative tools into a dedicated creator product, part of a broader push to embed AI across its apps and keep creators inside its ecosystem rather than defecting to rival tools.

Anthropic Accuses Alibaba of Largest-Ever Attempt to Distill Claude's Capabilities

AI28.8M exchanges

Anthropic Accuses Alibaba of Largest-Ever Attempt to Distill Claude's Capabilities

Anthropic accused Alibaba of illicitly extracting capabilities from its Claude models in what it called the largest known attack of its kind, according to a letter seen by Reuters. The campaign ran from April 22 to June 5, generating more than 28.8 million exchanges with Claude across nearly 25,000 fraudulent accounts, in what Anthropic describes as a distillation effort to accelerate China toward its frontier 'Mythos' capabilities.

AI Was Supposed to Kill Engineering Jobs -- New Data Suggests They're the Most Resilient

Despite predictions that AI coding tools would gut software engineering employment, new data suggests engineering roles are among the most resilient to AI disruption, according to TechCrunch. The finding complicates the dominant narrative -- reinforced this week by SpaceX's $60B Cursor deal -- that agentic coding is rapidly replacing human developers.

Google Keeps Losing AI Researchers to Rivals as the Talent War Intensifies

AI researchers are continuing to depart Google for its rivals, according to TechCrunch, extending a steady exodus of senior talent from the company that pioneered the transformer. The brain drain -- toward OpenAI, Anthropic, startups and well-funded new labs -- underscores how the scarcest resource in AI is not compute or capital but the small pool of people who can build frontier systems.

AIFirst custom inference chip

OpenAI Unveils 'Jalapeño,' Its First Custom AI Chip, Built With Broadcom for Inference at Scale

OpenAI Unveils 'Jalapeño,' Its First Custom AI Chip, Built With Broadcom for Inference at Scale

OpenAI revealed Jalapeño, its first in-house silicon -- a chip designed with Broadcom and purpose-built for running AI models (inference) rather than training them. OpenAI says early results show 'significantly better performance-per-watt' than current state-of-the-art alternatives, marking its most concrete step yet to reduce a near-total dependence on Nvidia GPUs.

Mistral Launches OCR 4, Turning Document Extraction Into a Full Enterprise AI Play

France's Mistral released OCR 4, an upgraded document-understanding model that pushes beyond plain text extraction into a full enterprise data-extraction stack. The move positions Europe's leading AI lab to compete directly with Google, AWS and Azure for the unglamorous but enormous market of turning documents into structured, machine-usable data.

Alibaba's Model Never Trained as an Agent -- Yet Beat Agent Benchmarks Across Seven Tests

Alibaba researchers showed a model that was never explicitly trained for agentic tasks but still improved agent performance across seven benchmarks. The result challenges the assumption that strong agentic behavior requires dedicated, expensive agent-specific training -- a potentially significant efficiency unlock.

Xiaomi's HarnessX Rewrites Its Own AI Scaffolding Mid-Task -- and Smaller Models Gain the Most

Xiaomi unveiled HarnessX, a system that lets an AI agent rewrite its own scaffolding -- the prompts, tools and control logic around the model -- in the middle of a task. The standout finding: smaller, cheaper models benefit the most, suggesting clever orchestration can substitute for raw model size.

New Paper Argues Microsoft Exaggerated Its 'Majorana 1' Quantum Breakthrough

A new paper argues that Microsoft overstated the quantum-computing claims it made roughly a year ago with its Majorana 1 chip, which the company touted as a breakthrough toward topological qubits. The dispute reignites long-running skepticism among physicists about whether Microsoft actually demonstrated the exotic particles its approach depends on.

AI· Jun 23, 2026

Anthropic Launches Claude Tag, a Persistent AI Teammate That Learns Your Company

Anthropic launched Claude Tag, replacing its old Slack app with a persistent AI teammate that learns a company's context over time, monitors channels, and works autonomously rather than only responding when summoned. It's a bid to move Claude from an on-demand assistant to an always-on coworker embedded in the flow of work.

Jun 23, 2026

AI· Jun 23, 2026

Krea Releases Krea 2 Raw and Turbo as Open Weights: Enterprise Image Gen in 2 Seconds

Krea released Krea 2 in Raw and Turbo variants as open weights under a custom license, claiming enterprise-grade AI image generation in roughly two seconds. Putting a fast, high-quality model into developers' hands as open weights pressures closed incumbents on both speed and cost.

Jun 23, 2026

Nvidia Says Its Rubin Data Center Design Runs Hotter to Use Far Less Water

Nvidia says its next-generation Rubin data center design runs at higher temperatures and relies on liquid cooling to dramatically cut water consumption, according to The Verge. The move addresses mounting scrutiny of AI's environmental footprint -- though critics note lower water use doesn't fully solve AI's resource problem.

SpaceX Signs a $6.3B Compute Deal to Rent GPUs to Open-Source Lab Reflection AI

AI$6.3B

SpaceX Signs a $6.3B Compute Deal to Rent GPUs to Open-Source Lab Reflection AI

SpaceX has agreed to supply Reflection AI -- an open-weight AI lab founded by ex-Google DeepMind researchers -- with Nvidia GB300 capacity at its Colossus 2 data center near Memphis, for $150 million a month from July 2026 through 2029, a $6.3 billion contract. The deal sits alongside SpaceX's even larger compute agreements with Anthropic ($1.25B/month) and Google ($920M/month), cementing the rocket company as a major AI-infrastructure landlord.

OpenAI Launches Initiative to Find and Patch Open Source Bugs

OpenAI has launched a new initiative to use its AI models to automatically find and help patch security bugs in open source software, according to TechCrunch. The effort positions AI as a defensive tool for the software supply chain that underpins much of the internet.

Alibaba's AI Video Model Climbs to No. 2 Globally as Sora and Seedance Fade

Alibaba's AI video generation model has risen to No. 2 in global rankings, overtaking OpenAI's Sora and ByteDance's Seedance, according to VentureBeat. The leap underscores how quickly Chinese labs are closing -- and in some categories leading -- the gap with US frontier players in generative media.

Sakana's New 'Fugu' System Hits Frontier Performance by Auto-Synthesizing Models

Japanese lab Sakana AI unveiled Fugu, a multi-model system that automatically synthesizes and combines specialized models to reach frontier-level performance without relying on a single giant model. The approach revives Sakana's evolutionary, model-merging philosophy as a counterpoint to the brute-force scaling pursued by the largest US labs.

'SaaS Isn't Coming Back': Crunchbase Argues Agentic AI Is Replacing the Model

A Crunchbase News analysis argues that traditional SaaS is being structurally displaced by agentic AI -- software that acts on outcomes rather than charging per seat. The piece frames a generational shift in how business software is built, priced and valued.