Sponsored by
A year ago, the race seemed simple. Build the biggest model. Ship the fastest. Win the headlines.
This week suggests the race has quietly changed.
Gartner reported today that worldwide spending on AI platforms and models will reach $64 billion in 2026, up 63 percent from last year. But the more interesting detail is where that money is moving. The fastest-growing segment is not general-purpose AI. It is domain-specific models, purpose-built tools designed around particular industries and workflows, up 210 percent this year. Enterprise AI budgets are facing greater scrutiny around cost, efficiency, and measurable outcomes. Organizations are not spending less. They are spending more carefully.
The question is no longer simply whether a product uses AI. It is whether it produces enough value to justify what it costs.
That shift is visible in China too.
On July 16, Moonshot AI released Kimi K3, a 2.8-trillion-parameter open-weight model with a one-million-token context window, the largest open-weight model ever shipped. In blind developer testing it ranked first in the Frontend Code Arena, beating Claude Fable 5. On broader independent evaluations it placed fourth among all frontier models globally, on par with Claude Opus 4.8 and ahead of GPT-5.6 Sol. Full open weights drop July 27. Demand overwhelmed Moonshot's compute capacity within hours of launch, forcing the company to pause new subscriptions temporarily.
One benchmark does not rewrite the industry. But the trajectory does.
A downloadable, self-hostable model is now credibly inside the frontier conversation. The gap between the best closed Western models and what any developer can access, modify, and run themselves has shrunk to months. When powerful intelligence becomes widely available, raw capability stops being a differentiator. The advantage moves somewhere else: into execution, into workflow design, into understanding a customer's real problem well enough to build something they actually rely on.
Intuit illustrates what that looks like in practice. Rather than pursuing maximum automation, the company deliberately designed its agents to keep humans in judgment roles. Agents handle the assembly work: matching transactions, surfacing recommendations, automating defined tasks. People retain control over the decisions that have real financial consequences. The goal, as Intuit's team described it, was to embed AI in natural workflows rather than bolt it on top of them. In a domain where a mistake costs a real business real money, that design philosophy is not cautious. It is smart.
Sometimes progress looks like adding capability. Real progress often comes from knowing what not to automate.
The physical footprint of AI is also becoming impossible to ignore. Over the weekend, opponents of rapid data center expansion organized 142 demonstrations across 42 states, raising concerns about water consumption, electricity demand, environmental impact, and whether communities receive any real benefit from the infrastructure being built around them. You do not have to agree with every protester to recognize the signal. AI no longer exists only inside a browser. It requires land, water, electricity, hardware, public support, and enormous capital. Every technological breakthrough eventually meets the limits of the real world.
The organizations that thrive in the next decade will not be the ones chasing every new model release. They will be the ones that identify meaningful problems, measure outcomes honestly, keep things as simple as they need to be, and build products people can actually rely on.
The novelty phase of AI is fading. The value era has begun.
The AI Agent You Can Trust
The best assistants don't multitask their attention across a hundred tools. Neither does Catch. It's an AI agent that focuses on one thing — the admin work you'd rather not touch — and does it exceptionally well.
Scheduling, flights, restaurants, follow-ups, vendors, clients. You hand it over; Catch handles the back-and-forth and comes back with it done.
No context-switching. No dropped balls. Just your admin, quietly cleared — so your focus stays on the work only you can do.
Meet the agent built for admin, and it'll be ready to work before your next meeting.
Get started at catchagent.ai — and give your attention back to what matters.
Signals to Watch
AI spending is entering an accountability phase. Gartner's report today shows spending shifting toward domain-specific models at 210 percent growth. The era of funding AI experiments without defined success criteria is ending. Boards want proof, not pilots.
Open models are reshaping the economics of access. Kimi K3's frontier-tier performance and immediate demand surge show how quickly international open-model development is pressuring proprietary platforms. When the best tools are downloadable, differentiation shifts entirely to execution.
In high-stakes work, trust is part of the product. Intuit's experience shows that human oversight and user control matter as much as automation in domains where mistakes have real consequences. Half of small businesses describe AI as helpful. Nearly a quarter have not used it at all. That gap is almost always about trust, not access.
Good signals are worth sharing. If this issue made you think differently about where AI is headed, forward it to someone who would benefit from the perspective.
And if you are not already subscribed, join the Gritletter community for weekly insights on AI, technology, business, and the forces shaping what comes next.