OpenAI's new GPT-Live family powers ChatGPT voice mode with more natural, less interruptive conversational AI, with API access for developers coming soon. 🔧
The AI Daily
Daily · July 9, 2026
Happy Thursday. The model wars intensified today as SpaceXAI's Grok 4.5 launched at roughly half the price of Anthropic and OpenAI equivalents — a direct shot across their bows from a company flush with post-IPO capital. Meanwhile, OpenAI rolled out GPT-Live, a new conversational voice model series, and quietly signaled a broad GPT-5.6 public release is imminent. On the money side, inference chip startup SambaNova closed a $1B round at an $11B valuation, underscoring that the market believes cheap, fast inference is where the next fortune gets made.
🧠 Foundation models
go deeper →A sharp price war broke out today at the frontier model layer, with SpaceXAI's Grok 4.5 undercutting rivals at roughly half their cost, while OpenAI simultaneously debuted GPT-Live voice models and telegraphed an imminent broad GPT-5.6 release — compressing the competitive cycle dramatically.
OpenAI launches GPT-Live voice model series ahead of broad GPT-5.6 release
SpaceX's Grok 4.5 launches at half the price of rivals — here's why that could rattle Anthropic and OpenAI
Grok 4.5, the first model born from SpaceXAI's Cursor acquisition, targets coding and agents at sharply lower prices than Anthropic and OpenAI.
Separating signal from noise in coding evaluations
OpenAI's internal analysis flags reliability flaws in SWE-Bench Pro, casting doubt on a benchmark widely used to rank frontier coding models.
I Built a Self-Improving AI, and So Can You
Practical experiments show self-improving AI loops are accessible beyond frontier labs, signalling a democratisation of recursive model improvement.
🏗️ Infrastructure
go deeper →The data-center buildout continues at scale — from Meta's first major Canadian facility to a wave of new Texas campus filings — while SambaNova's $1B inference chip raise signals that custom silicon for fast, cheap inference is the next critical bottleneck investors are funding.
Plans filed for 100,000 sq ft data center outside Dallas, Texas
A Belltown Power-linked affiliate is targeting a Royse City, Texas data center launch next year, adding to a surging cluster of AI compute facilities in the region.
Meta is building its first big Canadian data center as AI expansion crosses the border
Meta's first large-scale Canadian data center reflects AI hyperscalers' urgent hunt for new power and land beyond the US. 💰
Prime Intellect raises $130M at $1B valuation for its AI training platform
Prime Intellect's Nvidia- and Intel-backed $130M raise bets on distributed AI training infrastructure reaching unicorn scale. 💰
AMD targets system-level AI infrastructure optimization as agentic workloads reshape enterprise compute
AMD argues that agentic AI demands heterogeneous, modular system design — not just faster chips — shifting the infrastructure conversation to workload-level efficiency.
Tensordyne targets AI inference market with logarithmic math and Juniper-derived rack architecture
Tensordyne's novel logarithmic arithmetic and rack-scale design aims to slash power consumption for real-time AI inference at the edge.
Edged eyes 725MW data center campus in Pennsylvania
Edged's Project Atlas — a scaled-back but still massive 725MW Pennsylvania campus — signals continued appetite for gigawatt-class AI compute capacity.
💰 Funding & deals
go deeper →Today's biggest funding signal was SambaNova's $1B Series F at an $11B valuation — a strong market bet on inference chips as the next chokepoint — while Lovable's reported $13.2B valuation talks show that AI-native app builders are still commanding extraordinary multiples.
Inference chip startup SambaNova valued at $11B in $1B funding round
SambaNova's General Atlantic-led $1B Series F at an $11B valuation cements inference silicon as one of the hottest bets in AI infrastructure financing. 🏗️
Lovable reportedly in talks to double its valuation to $13.2B
Lovable's rumoured $300M Menlo Ventures-led round at $13.2B would make it one of Europe's most valuable AI startups, validating no-code AI app builders as a category.
Argentum targets the capital stack as the missing layer in AI infrastructure buildout
Argentum argues that creative infrastructure financing — not just silicon and power — is the binding constraint on how fast AI data centers can actually be built. 🏗️
SK Hynix looking to raise $28bn with US IPO
SK Hynix's planned $28B US IPO would be a landmark listing for the HBM memory maker that is central to Nvidia's AI chip supply chain. 🏗️
🔧 Middleware & platforms
go deeper →Slack's Slackbot took a meaningful step toward becoming a true enterprise AI agent today, gaining the ability to query CRM data, generate charts, and trigger DocuSign — all from a chat message — illustrating how existing platforms are quietly building agentic layers atop foundation models.
Slack's Slackbot can now pull your CRM data, generate charts, and send DocuSigns — all from a chat message.
⚡ Salesforce finally unified its $27.7B Slack acquisition with its CRM platform, turning Slackbot into a cross-enterprise agentic assistant for every workspace user. 📱
Our approach to government and national security partnerships
OpenAI's published framework for responsible government and national-security AI use sets guardrails that will shape how frontier models enter defence and intelligence markets. ⚖️
📱 Application solutions
go deeper →AI's workforce impact sharpened into focus today, with Indian IT facing an estimated 25,000–35,000 silent layoffs by year-end as automation takes hold, while on the product side Meta revealed prototype always-on AI smart glasses — a wearable form factor that could redefine ambient AI compute.
35,000 tech roles on the chopping block as AI takes the wheel at IT
⚡ Indian IT staffing firms project up to 35,000 AI-driven 'silent layoffs' in 2026, the clearest on-the-ground signal yet of automation displacing tech labor at scale. 🇮🇳
Meta is reportedly working on smart glasses that would be recording all the time
Meta's prototype 'super sensing' glasses that continuously capture audio and images could make ambient AI perception a mainstream consumer product.
This startup thinks robotics is about to have its ChatGPT moment
General Intuition is training physical-AI foundation models on millions of hours of video-game data, betting that synthetic data can unlock a step-change in robot intelligence. 🧠
Helping K–12 educators build practical AI skills
OpenAI Academy and the Walton Family Foundation are bringing hands-on AI Skills Jams to K–12 teachers, embedding AI literacy into the US public-school pipeline.
Google's deepfake detector system used to debunk McConnell hoax pic
⚡ Google's deepfake detection tool's first high-profile real-world use — debunking an AI-generated political hoax — validates its production readiness for misinformation defence. ⚖️
⚖️ Policy & legal
China dominated the policy front today with two distinct moves: a security warning about backdoor vulnerabilities in Anthropic's Claude Code, and reported plans to selectively permit top domestic AI firms to purchase limited quantities of Nvidia H200 chips — a targeted loosening of self-imposed restrictions that signals Beijing is recalibrating its technology strategy.
China warns about AI risks with Anthropic's Claude Code
China's official warning that Claude Code versions contain backdoors capable of exfiltrating data to remote servers will accelerate enterprise security scrutiny of Western AI coding tools. 📱
China plans to let top AI firms buy limited Nvidia H200 chips, the Information reports
Beijing's reported plan to allow selective H200 purchases by elite domestic AI firms suggests a pragmatic recalibration of chip self-sufficiency policy under competitive pressure. 🏗️
NEWSLETTER: China weighs silicon curtain around sought-after AI models
China is reportedly considering restricting exports of its own advanced AI models, mirroring US chip export controls in a potential tit-for-tat silicon curtain.
🇮🇳 India lens
India's AI story today is primarily a cautionary labour tale: staffing firms are quantifying thousands of 'silent layoffs' in tech as AI automation accelerates, even as Indian edtech and deep-tech startups continue raising fresh capital, suggesting a bifurcated economy where AI displaces some while funding others.
Elevate Education Bags ₹170 Cr To Scale AI-Led Higher Education Platform
Elevate Education's ₹170 Cr Series D round backs an Indian AI-first higher-education platform rebranded from Sunstone, targeting skilling in an AI-disrupted job market. 💰
Novyte Banks On Agentic AI To Solve The Optimisation Puzzle In Materials Science
Indian deep-tech startup Novyte is deploying agentic AI to compress the slow, iterative cycle of advanced materials discovery — a niche but high-value scientific application. 📱
📖 Beyond the headlines
Separating signal from noise in coding evaluations — OpenAI's self-published takedown of SWE-Bench Pro reliability is a rare admission from a frontier lab that the benchmarks used to justify model releases — and thus pricing, procurement, and competitive positioning — may be fundamentally unreliable. For any strategist making buy-vs-build decisions based on coding model leaderboards, this is essential reading before the next contract is signed. Read →