For those that are new around here, this is an extension of my weekly newsletter where I highlight new and innovative AI products that are worth exploring.
Hey {{FirstName | hey}}!
Happy Wednesday!
Here’s your news in newsletter!
In this Issue:
What a Wild July for Industry News
New Topic Byte!
The (dramatically reduced) Raise Recap
What a Wild July for Industry News
AI Use Is Exploding, So Are the Bills - J.P. Morgan flags how exploding token usage is catching enterprise budgets off-guard as multi-step agents consume massive amounts of compute. They predict a shift toward cheaper "good enough" models to manage costs, treating AI like a metered utility. 💵📉
Even if raw API token prices drop, running autonomous enterprise agents can still end up costing more due to multi-step reasoning loops and redundant retries. Builders need to focus on prompt architecture and context caching rather than relying on cheap inference alone. 📊⚙️
A wild roundup of market moves: OpenAI offers federal equity options, Anthropic deals with export restrictions on top models, and Meta hunts for neocloud compute partners. The big tech chess game for compute and government backing is getting wilder by the day! ♟️🚀
Apple has reportedly filed a massive lawsuit against OpenAI, claiming the AI giant infringed on trade secrets and iPhone IP to build its own hardware-software integrations. This marks a dramatic split between two tech behemoths. 🍏💥
OpenAI dropped GPT-5.6 as the new preferred model in Microsoft 365 Copilot, delivering higher reasoning efficiency per token. The update includes an "Ultra" tier that coordinates four parallel agents for complex workstreams. 🧠⚡
Meta pulled its new "Muse Image" feature in under 72 hours after users and SAG-AFTRA slammed it for using public Instagram face data by default without opt-in consent. A quick reminder that privacy missteps still trigger instant public outrage! 📸🛑
The Canadian and Korean AI Safety Institutes signed a landmark partnership to establish interoperable evaluation frameworks and combat synthetic media risks. International safety standard-setting is officially moving into high gear! 🇨🇦🇰🇷
Illinois passed sweeping legislation targeting non-consensual AI replicas, deepfakes, and algorithmic bias in hiring. States are stepping up to regulate tech labs faster than federal lawmakers can agree! ⚖️🔨
Tencent released Hunyuan-3 (Hy3) under an open Apache license, beating rivals like GLM-5.2 across general benchmarks at dirt-cheap API rates. Open weights out of Asia continue to put immense pressure on Western pricing models. 🇨🇳🔓
Towards AI: Micro-Agent Router Architecture put out a practical guide for SaaS builders on using micro-agent routers to dynamically send tasks to the cheapest, fastest model capable of doing the job. Intelligent routing is becoming the secret weapon for keeping LLM margins intact. 🔀💡
Legal scholars from the UN warn that treating binding hard law and flexible soft governance as separate issues creates dangerous regulatory loopholes. True accountability requires blending statutory enforcement with adaptive industry frameworks. 📜⚖️
Anthropic's Claude models are now available inside Microsoft Foundry, allowing enterprise developers to build cross-platform Copilots effortlessly. Multimodal integrations are getting seamless! ☁️🤖
Anthropic hosted a deep-dive session on how product teams can build robust, continuous evaluation pipelines for autonomous agents. They emphasize that catching edge-case failures early is the key to scaling agentic products. 🧑💻🎯
A New Topic Byte - For the Builders and Employees
Recently, I’ve had several conversations with teams asking variants of the same questions: “What about Open Source models? How are they actually useful for my business?” It’s a crucial thematic I’m gladly seeing come up as more and more companies navigate the shift from the quick API experiments phase towards core production infrastructure. In my latest piece, A Pragmatic Look at In-Housing AI: Why It’s a Business Strategy, Not a Tech Trend, I break down why there’s a real conversation to have around relying solely on frontier model APIs and how in-housing open-source models give businesses true operational independence, and much more.
The (Dramatically Reduced) Raise Recap 📈💰
Tempus AI agreed to acquire cancer genomics firm Personalis for $1.5 billion, . It proves healthcare AI is where high-stakes deals are going down right now! 🏥
SambaNova completed the first close of a massive $1 billion financing round at an $11 billion valuation. They are building full-stack hardware to keep up with enterprise AI demand! ⚡
Together AI raised $800 million in fresh capital, leaping to an $8.3 billion valuation. The open-source model hosting space is scorching hot right now! ☁️
CuspAI closed a huge $450 million Series B to build an AI "materials foundry". They are using artificial intelligence to discover new physical materials for clean energy and computing. 🧪
Prime Intellect locked down $130 million to build decentralized, open-source AI training networks. They want to aggregate compute power globally so anyone can train frontier models! 🌐
Lyzr ($100 Million Series B)Lyzr raised $100 million in a wild meta-move—they actually used their own AI agent ("SivaClaw") to run investor outreach and field questions! 🤖💸Read more: https://www.briefs.co/news/lyzr-secures-100m-after-using-its-own-ai-agent-to-handle-fundraising-outreach/
Valarian raised $50 million to help enterprises secure their internal communication and data pipelines from AI leaks. Keeping confidential data safe is big business! 🔒
Bespoke Labs raised $40 million to streamline AI post-training, while Gritt exited stealth with $34 million to build solar-plant construction robots! ☀️🤖
Pie landed $23.7 million to bring AI tools to Main Street businesses, while LinqAlpha secured $22 million to expand its financial market intelligence engine. 📈
Early-stage robotics startups like CarbonSix, Luxonis, Macrodata Labs, and Queue Six Robotics all locked down fresh capital to scale computer vision and automated arms. 🦾
Nvidia-backed Fireworks AI achieved a $1.75 billion valuation, while Indian startup Emergent reached $1.5 billion in funding discussions! 🚀
SK Hynix pulled off a historic $26.5 billion US listing—the largest foreign IPO in American stock market history—to fund its high-bandwidth memory (HBM) chip expansion.
Moonshot AI (creators of the Kimi chatbot) is dismantling its VIE structure to prep a Hong Kong IPO targeting a $30B+ valuation after releasing its Kimi K3 model. 🇨🇳
Stay well, and see you next week!
-✌🏽 Sam
Cue the legal stuff: The views and opinions expressed in this newsletter are solely my own and do not reflect the views, positions, or policies of my employer or any organizations I am affiliated with. All content is for informational purposes only and to express opinion.