AI BreakfastAI Breakfast You read. We listen. Let us know what you think by replying to this email. Sound familiar? Over 4 million people have had the same lightbulb moment. Morning Brew is a free daily newsletter that breaks down what's happening in business, finance, and tech — clearly, quickly, and with enough personality to make it the best email in your inbox. No yelling. No filler. Just the news, finally making sense. Try it for free OpenAI teases new hardware for Codex, dropping July 15 OpenAI engineers just cut inference costs for logged-out ChatGPT traffic by more than half, dropping the number of active Nvidia GPUs needed to power guest users down to just a few hundred. The precise optimization techniques remain a secret, but the hardware savings give the company crucial breathing room while data center expansions move slowly. OpenAI engineers also fixed long-standing crashes within the data pipelines powering ChatGPT search. A rare bug affecting Rockset, the core data system behind search and plugins, was traced back to faulty Azure hardware and an 18-year-old race condition in GNU libunwind. By analyzing thousands of core dumps, engineers found that Linux signals were corrupting memory during C++ exception handling. OpenAI swapped in libgcc's unwinder and pushed the fix upstream. On the AI capabilities front, OpenAI launched GeneBench-Pro, a new benchmark with 129 synthetic problems designed to test if AI can handle complex computational biology. It forces models to navigate ambiguous data in genomics and cancer research rather than follow static scripts.
OpenAI is launching physical developer hardware. Teaming up with custom keyboard maker Work Louder, the company teased a compact, square macro pad tied to its Codex coding assistant, dropping July 15. A teaser video on 𝕏 showed a device based on Work Louder’s Creator Micro 2, featuring programmable keys, dials, and joystick controls. Unlike OpenAI’s separate consumer hardware project with Jony Ive, this device is built purely to accelerate engineering workflows by mapping complex AI coding shortcuts to dedicated physical buttons. New Claude Sonnet 5 performance matches Opus 4.8 but costs more per task Anthropic just dropped Claude Sonnet 5 as its new default mid-tier model, and while it brings massive upgrades for heavy coding, early testers are pointing out a pretty wild financial catch. The headline specs look great: a 1M token context window, real-time cyber safeguards, and five distinct reasoning effort settings designed for complex agentic workflows. However, independent benchmarking shows Sonnet 5 can actually end up costing more per completed task than Anthropic’s premium flagship, Opus 4.8. Even with promotional launch pricing slashed to $2 per million input tokens and $10 per million output tokens through August, the sheer volume of generated text means power users might see higher bills. At the same time, Anthropic is branching out into specialized research and infrastructure. The company launched Claude Science in beta for macOS and Linux, an environment that hooks Claude models directly into high-performance computing clusters and 60 distinct scientific databases to automate biomedical data pipelines. For engineers looking to deploy autonomous setups, Anthropic published a new development framework for building multi-agent, proactive loops inside Claude Code. To tie it all together, enterprise teams can manage these loops via the new Claude apps gateway, a self-hosted stateless Linux container that handles identity auth and budget tracking across AWS and Google Cloud. Furthermore, Claude Opus 4.8 and Claude Haiku 4.5 reached general availability within Microsoft Foundry.
Meanwhile, the US Department of Commerce officially lifted export controls on Anthropic’s top-tier systems, paving the way for the return of Claude Fable 5 and Mythos 5 after a multi-week ban over jailbreak concerns. To appease regulators and monitor who is using the systems, Anthropic may introduce strict KYC identity verification requirements for Fable 5 access. Google clears Gemini 3.5 Pro for July Google Cloud is adding specialized "large quantitative models" from SandboxAQ to its marketplace, letting researchers combine Gemini’s workflow skills with math models trained directly on physics equations rather than internet text. Data teams are also getting TabFM, a foundation model built to execute zero-shot predictions on databases without traditional training, which Google plans to include directly into BigQuery SQL. For regular users, Gemini Spark is morphing into a full-fledged macOS desktop agent, letting users automate local file management, connect third-party apps like Canva, and eventually trigger multi-step workflows remotely from a phone. The creative stack got a double dose of media tools, too: a lightning-fast Nano Banana 2 Lite image model launched on OpenRouter to spit out 1K images in four seconds, while a new Gemini Omni Flash video model dropped on a stateful Interactions API to allow iterative chat-style video editing for ten cents a second. To cap it off, NotebookLM is getting mobile-focused, turning dense research papers into 60-second, vertical, TikTok-style explainer clips for Pro subscribers. Over on the flagship side, Gemini 3.5 Pro is officially clear for a July launch without government interference. While high cybersecurity scores triggered federal restrictions for Anthropic's Fable 5 and OpenAI's GPT-5.6, Google's model escaped review by staying below those unwritten hacking capability thresholds. Frontier models and product moves
Meta limits use of Claude and Codex over fears of AI model distillation
Meta limits use of Claude and Codex over fears of AI model distillationDeepSeek plans a mid-July launch for the full V4 series to fix preview complaints
DeepSeek plans a mid-July launch for the full V4 series to fix preview complaintsEthan Mollick maps the striking performance gap separating open and closed weights
Ethan Mollick maps the striking performance gap separating open and closed weightsWix subsidiary Base44 launches a custom model to lower vibe coding costs
Wix subsidiary Base44 launches a custom model to lower vibe coding costsCline launches a cheap subscription to access top open-weight networks like GLM-5.2
Cline launches a cheap subscription to access top open-weight networks like GLM-5.2New Runway update allows its built-in agent to use Google's Omni Flash
New Runway update allows its built-in agent to use Google's Omni Flash Agents and the agentic stack𝕏 launches hosted MCP to feed real-time data to AI agents
𝕏 launches hosted MCP to feed real-time data to AI agentsThinking Machines and Bridgewater fine-tune a custom AI to replicate expert financial judgment
Thinking Machines and Bridgewater fine-tune a custom AI to replicate expert financial judgmentCognition builds Devin Fusion to slash autonomous engineering costs by 35%
Cognition builds Devin Fusion to slash autonomous engineering costs by 35%Morgan Stanley fixes risky reconciliation work by limiting AI autonomy
Morgan Stanley fixes risky reconciliation work by limiting AI autonomyNvidia builds a GPT for physics simulations to control virtual movement
Nvidia builds a GPT for physics simulations to control virtual movementAndrew Ng outlines the loops changing engineers into AI product managers
Andrew Ng outlines the loops changing engineers into AI product managersOpenAI and Nvidia developers are forcing LLMs to speak like cavemen to cut costs (paywalled)
OpenAI and Nvidia developers are forcing LLMs to speak like cavemen to cut costs (paywalled)Vercel raises backend deployment caps by twenty times to accommodate massive AI models
Vercel raises backend deployment caps by twenty times to accommodate massive AI models Business, labor, and institutionsAmazon's AWS commits $1 billion toward a new unit for embedded AI engineers
Amazon's AWS commits $1 billion toward a new unit for embedded AI engineersChamath Palihapitiya raises $135M Series A for his AI coding startup, takes CEO role
Chamath Palihapitiya raises $135M Series A for his AI coding startup, takes CEO roleMarc Andreessen is appointed to the DOD's Defense Policy Board
Marc Andreessen is appointed to the DOD's Defense Policy BoardArena, the AI leaderboard everyone uses, is now a $100M business
Arena, the AI leaderboard everyone uses, is now a $100M businessMiddle powers challenge Silicon Valley with a massive open-source AI coalition
Middle powers challenge Silicon Valley with a massive open-source AI coalitionDeloitte tells its own consultants: AI is coming for the billable hour
Deloitte tells its own consultants: AI is coming for the billable hourAmazon replaces human HR with AI, leaving workers stranded in loops
Amazon replaces human HR with AI, leaving workers stranded in loopsApple Vision Pro exec is reportedly leaving for OpenAI
Apple Vision Pro exec is reportedly leaving for OpenAI𝕏 Money is reportedly launching to Premium+ subscribers with 6% APY
𝕏 Money is reportedly launching to Premium+ subscribers with 6% APYBitcoin miner and AI firm Ionic Digital files for a Nasdaq direct listing
Bitcoin miner and AI firm Ionic Digital files for a Nasdaq direct listingSummer heat waves trigger a live stress test for global AI data centers
Summer heat waves trigger a live stress test for global AI data centersData centers risk massive downtime from bacteria growing in cooling fluid
Data centers risk massive downtime from bacteria growing in cooling fluidLeaked project reveals Meta sent thousands of crisis prompts to rival AI bots
Leaked project reveals Meta sent thousands of crisis prompts to rival AI bots Hardware and infrastructureChina's Meituan builds a trillion-parameter AI using only domestic chips
China's Meituan builds a trillion-parameter AI using only domestic chipsFirefly Aerospace taps Nvidia Jetson to run edge AI in lunar orbit
Firefly Aerospace taps Nvidia Jetson to run edge AI in lunar orbit Science and medicineMeta's Brain2Qwerty v2 decodes full words directly from raw brain signals
Meta's Brain2Qwerty v2 decodes full words directly from raw brain signalsNew ConlangCrafter AI model can imagine and build entirely new languages
New ConlangCrafter AI model can imagine and build entirely new languages Media and creative AINetflix secures family consent to replicate Gene Wilder's voice with AI
Netflix secures family consent to replicate Gene Wilder's voice with AIElevenLabs rolls out SynthID support
ElevenLabs rolls out SynthID support Developer toolsCursor launches native iOS app in public beta
Cursor launches native iOS app in public betaOpenClaw is now on iOS and Android
OpenClaw is now on iOS and AndroidAgibot robots hit 99% success during a six-day live factory demo
Agibot robots hit 99% success during a six-day live factory demoApptronik, the $5.5 billion robotics startup, built a school for humanoids
Apptronik, the $5.5 billion robotics startup, built a school for humanoidsFigure 03 humanoids land a BMW logistics contract with new tactile hands
Figure 03 humanoids land a BMW logistics contract with new tactile handsUBTECH launches 50 realistic humanoid robots designed for home companionship
UBTECH launches 50 realistic humanoid robots designed for home companionshipNew Reflect v1.0 software grants humanoids mid-mission language control
New Reflect v1.0 software grants humanoids mid-mission language controlDavid Sacks warns US policy drops local AI models into purgatory
David Sacks warns US policy drops local AI models into purgatoryEnterprise AI race shifts from model capability to workflow control
Enterprise AI race shifts from model capability to workflow control Cursor for iOS runs cloud or desktop coding agents from your phone to build software anywhere. Foresight by Lightning Rod delivers calibrated, benchmark-verified AI forecasts for developers building agents and decision tools. Oxlo.ai provides unified API access to over 35 frontier AI models with predictable monthly subscriptions. v0 Design Systems 2.0 imports your existing code, components, and patterns to build and iterate apps in chat. AgentPeek displays live Claude Code and Codex agent sessions in your Mac notch for easy local monitoring. Thank you for reading today’s edition. Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter. Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on X! Thinking of starting your own newsletter? AI Breakfast readers who sign up with Beehiiv receive a 14-day free trial and 20% off for 3 months. Keep Reading AI Breakfast Curated weekly analysis of the latest AI projects, products, and news