Daily AI Briefing
The 120 stories that mattered in AI, curated and summarized from dozens of sources by AIBriefs.
Claude Fable 5, a Mythos-class model with safety safeguards, is now generally available, outperforming all prior Anthropic models on benchmarks. It is available as an orchestrator in Perplexity's Computer for Pro and Max users, and can autonomously complete tasks such as beating Pokémon FireRed using only vision.
Opus 4.8 builds on Opus 4.7 with sharper judgment and improved benchmarks, available at the same price. New features include effort control, dynamic workflows in Claude Code, and fast mode at 3x lower cost.
MiniMax M3 ranks #7 on Code Arena Frontend and #8 on Artificial Analysis Intelligence Index. Priced at $0.07 per run vs $3.39 for Opus with similar bug-catching, open weights dropping soon.
The 11.95B-parameter model runs locally on laptops with 16GB RAM, supporting text, audio, image, and video inputs via a novel encoder-free architecture. Early benchmarks show it nearly matches larger 26B models while enabling agentic workflows and on-device processing.
China is preparing a $295 billion plan to fund a nationwide AI infrastructure buildout, including data centers and computing power. The initiative comes as the US-China AI race intensifies, with Beijing seeking to accelerate domestic capabilities.
Moonshot AI's Kimi reached a $30B pre-money valuation, a sixfold increase in six months. The company launched a new fundraising round following a reported $2B raise in May.
Harness-1, a 20-billion parameter open-source search agent, was unveiled by researchers from UIUC, UC Berkeley, and Chroma. It outperforms GPT-5.4 on recalling relevant information using a state-externalizing harness.
FrontierCode tasks each required 40+ hours from open-source maintainers. The benchmark measures whether code would actually be merged, addressing findings that over half of SWEBench results are unmergeable slop.
Apollo Global Management finalized a $35 billion debt deal to acquire Google TPU chips for Anthropic. Google's financial backstops underpinned the transaction, which is among the largest AI infrastructure financings.
Fully autonomous drones have killed human soldiers for the first time, according to a New Scientist report. The incident marks a significant milestone in the use of AI in warfare.
Neura Robotics raised up to $1.4 billion from Nvidia, Amazon, and other investors. The company develops humanoid robots for the physical AI space.
President signed an executive order titled 'Promoting Advanced Artificial Intelligence Innovation and Security' on June 2, 2026. The order aims to foster AI innovation and enhance security protocols.
Uber is capping employee AI tool spending at $1,500/month per tool after exhausting its 2026 AI budget by April. President Andrew Macdonald said there’s no clear link between token consumption and useful features, making the trade-off harder to justify.
A German court held Google liable for false statements in AI Overviews in a case brought by two publishers. The ruling could impact all AI search engines and chatbots that poorly paraphrase source links.
Ideogram 4.0 is now the #1 open-weight image model on Design Arena, featuring structured JSON prompting, multilingual text rendering, and native 2K resolution. The model is available on Hugging Face with weights and inference code via Diffusers. It was trained from scratch and is not a fine-tune of existing models.
Bloomberg video profiles Anthropic, now valued at $965 billion, covering its growth and strategy. It offers an inside look at the AI startup's trajectory.
Apollo and Blackstone are providing capital to fuel the AI industry boom. The move signals growing interest from alternative asset managers in AI infrastructure and startups.
Microsoft announced a series of generative AI models at its Build developer conference, positioning them as cheaper alternatives to OpenAI and Anthropic. The move comes after reliance on Anthropic's Claude forced Microsoft to raise GitHub Copilot prices and cap usage.
ChatGPT app crossed 600 million monthly active users for the first time, according to Similarweb. The milestone highlights the rapid growth of OpenAI's flagship product.
The company raised $200 million in funding. It is part of a crop of startups using AI to build more sophisticated robots for US manufacturing.
Anthropic is expanding its security-focused Claude Mythos model to 150 organizations across 15 countries. The deployment targets critical infrastructure sectors including power, water, and healthcare, where a cyberattack could affect up to 100 million people.
NVIDIA's new Vera CPU delivers 80% faster agentic task completion than x86 processors, targeting AI factories handling reinforcement learning and tool execution. It extends the AI platform from GPU to CPU, enabling faster agent loops and higher throughput.
The deal is the latest in a series of major partnerships between tech giants and Corning, which has become a key player in AI infrastructure. Financial terms were not disclosed, but the agreement is expected to support Amazon's expanding AI data center footprint in the U.S.
The agenda covers safety, youth protection, workforce transition, and global standards. It calls for proactive governance and international cooperation.
The model ESMC was trained on 2.8 billion protein sequences. Lab-validated binders designed by ESMFold2 achieved high affinity in days.
The 1,400-acre Project Jupiter site in New Mexico is part of a $50 billion infrastructure bet. The scale of AI energy demand has never been tested before, per Bloomberg.
OpenAI, Anthropic, and other AI labs sent a letter to lawmakers calling for stricter oversight of synthetic DNA sequences to prevent bioweapons development. The letter highlights risks of AI being used to design dangerous biological agents.
OpenAI outlines a federal framework for governing frontier AI, focusing on safety, resilience, and national security. The blueprint proposes a U.S. governance approach for advanced AI systems.
Hermes Desktop app from Nous Research is now available on macOS, Windows, and Linux, offering a GUI for the Hermes Agent. It has 140K GitHub stars in 3 months and is the most used agent on OpenRouter.
Usage limits doubled until July 5 for all paid plans. The promotion applies to the 5-hour rate limits, allowing users to tackle bigger projects at no extra charge.
Microsoft's CEO declared the company will pivot from OS and apps to AI agents, marking a major strategic shift. The statement positions agent-based AI as the next computing platform.
Stargate, a collaboration of OpenAI, Oracle, and SoftBank, breaks ground in Abilene, Texas, backed by President Trump. Local leaders cite new investment and infrastructure improvements. Goldman's top bankers now focus on AI data center financing.
TSMC CEO says customer demand is so high the company can only support so much, even with its US factory buildout. The world's largest chipmaker is struggling to keep up with AI demand, affecting the entire AI supply chain.
OpenAI launches new Codex plugins, sites, and annotations tailored for analysts, marketers, designers, investors, and other teams. The update extends Codex's AI capabilities beyond developers to support a wider range of workflows.
NVIDIA and Microsoft announce expanded partnership at Microsoft Build, including RTX Spark and DGX Station for Windows. NVIDIA's GPU-accelerated Microsoft Fabric and open models on Microsoft platform are also highlighted.
The UK government is investing billions in a state-backed AI supercomputer to foster homegrown chip startups. The initiative aims to reduce the country's reliance on US technology in the AI sector.
Step 3.7 Flash is a 196B total / 11B active MoE model with built-in 1.8B ViT, running locally on 128GB RAM. On SWE-Bench Pro it scores 56.26%, beating DeepSeek V4 Flash.
SB 315 requires frontier AI firms to submit safety plans, independent testing results, and report critical incidents within 72 hours (or 24 hours if imminent risk). OpenAI and Anthropic supported the bill, which governor J.B. Pritzker said he will sign.
Bots now account for over 50% of global web traffic, surpassing human traffic for the first time, according to Cloudflare Radar data. The milestone underscores the increasing prevalence of automated crawlers and AI agents online.
Intel plans to ship an AI chip by end of 2026 that uses cheaper memory and cooling than rival Nvidia and AMD offerings. Kevork Kechichian, Intel's data center head, says the chip will be both cheaper and run cooler.
Nvidia's new Vera AI chip is being adopted by Anthropic and OpenAI. CEO Jensen Huang highlighted the partnerships at a recent event, signaling strong demand for Nvidia's next-generation hardware.
Anthropic's Claude for Legal officially launches, featuring over 90 AI agents, 12 main plugins, and MCP connectors to legal tech. The platform automates key legal tasks and integrates with existing tools.
The 1-petaflop RTX Spark is designed to run AI agents like OpenClaw or Hermes Agent securely, with sandboxes co-developed with Microsoft. PCs from ASUS, Dell, HP, Lenovo, Microsoft, and MSI will ship this fall. Nvidia CEO Jensen Huang envisions a future where PCs act autonomously: 'You ask — and the PC does the work.'
SPACE introduces source-free concept erasure for MLLMs. Visual-Noise Guided In-Context Distillation offers an alternative unlearning approach using noisy visual prompts.
Google has released Magenta Realtime 2, a music generation model, on HuggingFace. The model is gaining traction with 52 likes and 311 downloads.
ByteDance, parent of TikTok, is considering spending up to $70 billion on AI infrastructure, marking one of the largest corporate AI investments. The plan, still under internal review, would fund data centers and compute resources.
Google's NotebookLM receives new agentic features and advanced reasoning to assist with complex research projects. The update enhances the tool's ability to synthesize and analyze information.
Gemini Spark is Google's always-on AI agent that accesses personal Gmail, Docs, and Calendar to automate tasks. In a hands-on test, it generated a full party itinerary and guest list from a single prompt, but misidentified the author's partner as a 'close friend'. The agent was introduced at Google I/O as a response to the OpenClaw agent trend.
OpenAI breaks ground on a 1GW data center in Michigan as part of the Stargate project. The project aims to expand AI infrastructure, create jobs, and support communities.
eToro's AI agent Tori now uses SpaceXAI models to analyze real-time market sentiment from X. The integration allows Tori to track live signals and market moods, available via the API console.
Anthropic confidentially submitted an IPO draft S-1 to the SEC. Its annualized revenue run rate is nearing $50B and it is on track for its first profitable quarter. OpenAI trails with lower revenue and no profitability yet, as both companies face pressure from massive capital expenditures.
Salesforce's investment in Anthropic is valued at about $5 billion. The deal marks a major bet on Anthropic's AI technology by a leading enterprise software company.
Super Micro Computer is seeking $7 billion in an equity deal to fund purchases of AI equipment. The deal aims to capitalize on surging demand for AI servers.
Benchmark evaluates top ASR models on bilingual code-switched speech. Results reveal performance gaps in handling mixed-language conversations.
Leaked messages from Google's internal 'Memegen' forum show employees criticizing AI code generation, calling it a bottleneck machine. Leadership has boasted that 75% of new code is AI-generated, but developers push back.
A tweet reports Microsoft as achieving 1.75x speedup by making LLMs forget intermediate reasoning traces. The technique removes unstructured internal monologue to reduce latency.
Security researchers at Blue41 discovered a vulnerability in Bunq's financial AI assistant that can be triggered by a €0.01 bank transfer. The exploit could allow attackers to compromise the AI's behavior.
Broadcom announced the largest set of Spring security fixes on Monday, driven by an increase in AI-powered attacks targeting the Java ecosystem. The update addresses multiple CVEs and underscores how AI is rewriting software security rules.
Bloomberg reports Magnetar Capital is launching a fund that uses AI bots to replace human stock analysts. The fund represents a significant push toward automation in quantitative investing.
A person was wrongfully arrested after an AI system misidentified them. The individual is now seeking justice through legal action.
Numan built Nu, an AI health coach using LangGraph, operating safely outside medical device regulation. The system uses agent evaluations and automated improvement loops for compliance and quality.
The round was led by Norwest, with participation from S Capital VC, Cerca Partners, and Oceans Ventures. Snowflake Ventures also invested strategically.
On-device AI agents are limited by DRAM capacity, capping model sizes. Apple's new architecture overcomes this, enabling larger models and more capable agents without cloud dependency.
Glean's Work AI Index 2026 reports that 87% of workers use AI, saving 13 hours per week, but only 13% say their organization is performing well. Rebecca Hinds discusses the 'hidden human labor' of babysitting AI outputs and the gap between individual gains and organizational maturity.
Ramp is launching Applied AI Solutions, deploying its engineers inside enterprise finance teams to build custom AI tools. The company argues off-the-shelf finance AI can't match bespoke solutions. The model is inspired by the forward deployed engineer role popularized by OpenAI and Anthropic.
Kimberly Powell, VP of Healthcare and Life Sciences, presented NVIDIA's vision for AI-driven healthcare transformation at GTC 2026. The address covered open foundation models, agentic AI, and robotics for clinical applications.
GitHub Copilot's new app offers flexibility to select from multiple model providers, unlike Codex which is locked to OpenAI models. The app is now available for download.
In a Pragmatic Engineer podcast episode, Kelsey Hightower demonstrates Claude taking actions in the AWS console. The video highlights Claude's agentic capabilities in a cloud environment.
Boris Cherny, head of Claude Code at Anthropic, says he no longer directly prompts the assistant. His new approach is to write loops that instruct Claude, reflecting a shift in AI developer tooling methodology.
Column explores the bipartisan and cross-sector alliances forming around AI regulation, as seen at the Second Annual AI Honors. The event included Sens. Rounds and Warner, NVIDIA co-founder Chris Malachowsky, and other figures.
Igor Babuschkin, co-founder of xAI, has unveiled a new startup focused on personalized AI. The announcement was made via Bloomberg, with no further details on funding or product disclosed.
Addverb, an Indian robotics startup backed by Mukesh Ambani, is seeking $100 million in funding to expand operations and compete with Chinese manufacturers. The company has developed a wheeled humanoid robot called Elixis-W.
Meta has partnered with Reliance Industries to build its first AI data center in India. The collaboration combines Meta's AI expertise with Reliance's extensive infrastructure and market presence.
AI coding tool adoption rose from 76% in 2024 to 90% in 2025, with teams now deploying up to 1,000 times per month. The rapid code generation has exposed bottlenecks in testing, review, and pipeline infrastructure.
China-based entities made over half of state-sponsored cyberattacks on tech firms for AI assets, according to CrowdStrike. The cybersecurity firm warns of Beijing's escalating efforts to catch up with the US in AI through cyber espionage.
A bipartisan bill co-sponsored by Sens. Dave McCormick (R-Pa.) and John Hickenlooper (D-Colo.) would create a National Commission on Robotics. The commission aims to unify federal robotics policy to support economic and military interests.
Average review time dropped from ~5 minutes to ~90 seconds. Bugbot now finds 0.62 bugs per review (up from 0.56) and costs 22% less per run.
Former xAI engineer Devin Kim sues xAI and SpaceX, alleging he was fired after flagging safety concerns about Grok days before SpaceX's IPO. No comment from xAI yet.
Gemini 3 Flash Preview can produce speaker labels, timestamps, emotion tags, and language detection with translation in one API call. The talk covers speech generation and live music capabilities, demonstrating the underlying audio understanding layer.
Decart's Oasis 3 generates real-time photorealistic driving environments for autonomous vehicle testing. The model is now available via API, though the article notes some caveats regarding fidelity.
Triomics, an oncology AI company helping cancer centers manage clinical information, raised $22 million in Series B funding led by Battery Ventures. The round will scale its platform across cancer centers nationwide and deepen life sciences partnerships.
SCAIL-2 animates a reference character with a driving video, supporting character replacement and multi-character animation. It is available on Hugging Face as an open-source model.
An MIT study found that AI assistants improved fake news detection during use, but users' independent ability to identify misinformation declined afterwards. The research raises concerns about over-reliance on AI for critical thinking.
dots.tts is a 2B-parameter continuous autoregressive TTS model released by RedNote (Xiaohongshu) under Apache 2.0. It models speech in a continuous latent space without codec quantization.
Blog post details implementing Kolmogorov-Arnold Networks on FPGAs for high-speed ML inference. The method leverages hardware acceleration for KANs.
OpenAI has invested in Fielding, an AI startup that automates compliance and underwriting tasks for financial institutions. The company aims to streamline regulatory processes and risk assessment using AI.
Microsoft CEO Satya Nadella argues that AI agents should be managed like human employees, with identities, permissions, and audit trails. The approach aims to ensure accountability and governance in enterprise AI deployments.
Einride shares surged on its first day of trading on the Nasdaq following a SPAC merger, though exact percentage not specified. The company is a leader in autonomous freight trucking technology.
D-Matrix says its AI chip is 10 times faster than a GPU and bypasses memory bottlenecks, entering full production. The Microsoft-backed startup targets Nvidia's dominance in inference workloads.
The widely cited $14B non-GAAP loss target for 2026 misses ~$7B to $10B in equity compensation. The median GAAP net loss estimate lands at $25B-$26B, roughly 80% higher.
ByteDance's Bernini-R model is now available on HuggingFace, with 48 likes and 57 downloads. The model is hosted under the ByteDance organization.
The AI coding agent startup raised a $7 million seed round from prominent angel investors. It aims to give companies more control over their AI coding tools, betting against lock-in with major model makers.
Mistral CTO Timothée Lacroix discusses how open-weight models are closing the gap with proprietary AI and accelerating enterprise adoption. The conversation covers model customization, production deployment, and Mistral's frontier AI approach.
KKR's Raj Agrawal argues that AI power demands may far exceed current market expectations. The investment firm sees energy infrastructure as a critical bottleneck for AI growth.
OpenAI's report describes PRC-linked groups using AI to target US tech debates, data center narratives, tariffs, and spread false claims about ChatGPT. The operations aim to influence American discourse on AI policy.
Sapient researchers trained a foundation model from scratch for approximately $1,500, dramatically undercutting the millions typically required. They developed HRM-Text, which replaces standard Transformers to reduce compute costs.
Lindy, an AI agent startup, switched from Anthropic's Claude to DeepSeek's models, reducing inference costs by millions. The move highlights the challenge of rising inference costs for AI agents, as seen with GitHub's shift to usage-based billing.