Sponsored by
Hello readers,
Welcome to the AI For All newsletter! Today, we’re talking about how AI is remixing concrete, the advantages of using AI in vendor procurement, and more!
AI in Action: Getting Concrete
Concrete is the most widely used construction material on Earth, and the United States pours roughly 400 million cubic yards of it every year. But for all that volume, mix design has traditionally been a slow, intuition-driven process — engineers testing formulations by trial and error, constrained by decades-old methods and a heavy reliance on imported cement. Meta, working with the University of Illinois at Urbana-Champaign and Amrize, the largest cement and concrete manufacturer in North America, has built an AI model that changes that calculus considerably.
The model, called BOxCrete, uses Bayesian optimization to navigate the enormous space of possible concrete formulations. Rather than testing mixes randomly, the AI learns from existing lab data, proposes high-potential candidates, and refines its predictions with every new result — compressing months of trial-and-error into a fraction of the time. The practical proof came at Meta's data center in Rosemount, Minnesota, where the AI-optimized mix reached full structural strength 43% faster than the original formula while also reducing cracking risk by nearly 10%, all using domestically sourced materials. The mix has since been qualified for use across additional sections of the building.
Meta released BOxCrete as open-source software under an MIT license, meaning any producer can download and use it without licensing fees. Pennsylvania-based Quadrel, an enterprise platform serving the ready-mix industry, has already embedded the framework into its daily mix design and quality control workflows. The models improve continuously as field results come in. An industry that has moved slowly for decades now has a freely available AI tool that gets better the more it's used — and the data behind it is the most comprehensive open-source concrete performance dataset publicly available.
🔥 Rapid Fire
Microsoft is moving all GitHub Copilot subscribers to token-based billing
Microsoft paused new Student, Pro, and Pro+ signups
Debunking the Mythos hype and fear — how everyone fell for it again
Analysis: Four Horsemen of the AIpocalypse
Anthropic’s service availability falls below the standard 99.99%
Anthropic degrades the performance of Opus 4.6 and 4.7
Opus 4.7 performs worse than 4.6 and burns more tokens
Anthropic briefly removed Claude Code from the Pro subscription
Mythos was held back due to capacity constraints, not capabilities
AI compute demand is inflated by OpenAI and Anthropic
Amazon did not invest $25 billion in Anthropic
The deal is for $5B with up to $20B more in the future maybe
Anthropic will hand the money right back to Amazon in yet another circular deal — Anthropic also agreed to spend $100B on AWS over the next decade (Anthropic does not have this money)
NVIDIA is selling years’ worth of GPUs in advance and warehousing them
Companies are exceeding budgets for inference by “orders of magnitude”
Subsidized LLM subscriptions are coming to an end
Read the Microsoft story from above
Anthropic is already moving enterprise customers to API rates
This will be at the cost of growth and customer retention
Spellbook CEO Scott Stevenson said, “The reason many AI startups are crushing revenue records is because they are using a dishonest metric. The biggest funds in the world are supporting this and misleading journalists for PR coverage.” >50% of enterprise AI startups are using “contracted ARR” to inflate their revenues according to Stevenson.
Analysis: The Hater’s Guide to Private Credit
Sona Asset Management said of private credit, “Private credit is full of cruddy loans issued by companies that couldn’t get financed anywhere else, and assembled haphazardly into rattlebags, rather than carefully constructed portfolios, by dealmakers (cosplaying as managers) who get paid to shovel client money out as fast as they can.”
Some shareholders question if Sam Altman should lead OpenAI through IPO
Altman has potential conflicts of interest
Altman asked OpenAI to fund startups he invested in
SoftBank seeks $10 billion margin loan backed by OpenAI shares
This is on top of SoftBank’s recent $40B bridge loan
Some have suggested that SoftBank is out of assets to sell
OpenAI makes yet another fake billion-dollar deal it can’t afford to pay
Oracle, Nebius data centers for OpenAI, Microsoft are way behind schedule
Data center company Fermi’s CEO abruptly departs amid construction delays
Anthropic secretly installs spyware when you install Claude Desktop
Top lawyer at prestigious law firm falls victim to AI hallucinations
What 2,000 SaaS Companies Reveal About Growth in 2026
Is your growth in-line with your peers in B2B SaaS & AI?
Benchmark yourself against actual billings data for Maxio’s 2000+ global customers, alongside firsthand company perspectives to understand how growth varied by company size, business model, and strategic focus.
Key takeaways from the report:
Average growth across 2,000 companies
Growth by revenue band
AI-led vs AI-enhanced. Who performed better?
📖 What We’re Reading
As many as 64% of procurement leaders anticipate that artificial intelligence, particularly generative AI, will transform their roles over the next five years.
Despite this, only around 4% of procurement teams have adopted AI at scale so far, highlighting that the technology is still in its formative stages.
Here’s how artificial intelligence can enhance your vendor procurement success rates and how they can drive supply chain efficiency at scale for adopters.
