
Every week a new model drops. New APIs. New keys. New rate limits. More chaos for developers. Gatewayz fixes it. One API for 10,000+ models. One line of code away.

🦙 Building Modelz, the first DEX that turns AI models into liquid, tradeable tokens and streams API revenue back to holders. t.me/alpacanetworkai

Every week a new model drops. New APIs. New keys. New rate limits. More chaos for developers. Gatewayz fixes it. One API for 10,000+ models. One line of code away.

Open AI needs people who can afford to keep maintaining it. Hugging Face bringing oMLX's creator on board gives that work sustained backing while the project stays Apache 2.0. Funding the work after launch deserves as much attention as launch day.
Super happy to welcome @jundotkim, oMLX creator, to Hugging Face! 🥳 We are invested in Local AI and support MLX since @awnihannun and @angeloskath released it in Christmas 2023. oMLX gets stability and speed. It remains a community project led by Jun. huggingface.co/blog/omlx

Super happy to welcome @jundotkim, oMLX creator, to Hugging Face! 🥳 We are invested in Local AI and support MLX since @awnihannun and @angeloskath released it in Christmas 2023. oMLX gets stability and speed. It remains a community project led by Jun. huggingface.co/blog/omlx

Issuing a token is one part of building a market. Buyers also need clear ownership rights, custody and a dependable way to settle the trade. Pontes is worth watching because it addresses that settlement layer. The next test is how much activity follows.

Pontes connects market DLT platforms with the Eurosystem's TARGET Services. An initial group has completed onboarding. More participants and features will follow, with full implementation expected by 2028. This is a staged rollout.

Tokenized finance got a new settlement connection today. The Eurosystem launched Pontes, allowing eligible institutions to settle wholesale tokenized-asset transactions in central bank money. The cash side of the trade is getting an upgrade.

Jev reached nearly 13% of paid teams on Vercel AI Gateway in its first day, according to Vercel. That's a reason to watch specialized models closely. The next signal I'd want: how many teams keep using it a month later.
Jev was adopted faster than any other model in AI Gateway history. In the first day, @typesafeai reached ~13% of teams, 2x the GPT-5.6 family and 6x Fable 5.1.

Open models become more useful when more people can afford to run them. Releasing the optimization code matters: other teams can inspect it, test it, and build on the work. Lower compute barriers give more researchers room to experiment.
Biologists use specialized open-source models for tasks like modeling the structure of molecular systems, designing drug-like molecules, and predicting the effects of genetic mutations. But these models are often expensive to run, potentially limiting their impact. In our latest Science Blog, we share how Claude was able to optimize inference for more than 30 open-source models, making them 4x faster on average, partly by writing custom software for GPUs. We’re open sourcing all of the optimization code. Read more: anthropic.com/research/claud…

AI agents will need a way to pay for useful work. Arc's mainnet launch makes that conversation more concrete: software buying data, compute, or services within a defined budget. The interesting question: what will agents actually pay for?
Arc Mainnet is live. The Economic OS for the internet is online and onchain. An open internet platform for the world's financial markets, real-time value movement, and agentic economic activity.

The AI stack is getting rebuilt in real time. New models. New agents. New interfaces. New infrastructure. Modelz tracks what is emerging. Gatewayz helps you build across it. Modelz → modelz.io Gatewayz → gatewayz.ai

AI news this week (Sep 9 – Sep 16): 1) OpenAI launched the Agents API and brought GPT-Live-1 to the API, pushing agent workflows and real-time voice further into developer infrastructure. 2) DeepSeek released V4.1-Flash, a faster and cheaper multimodal model built on a new architecture with a much smaller KV cache. 3) Mistral partnered with Cloudera to bring sovereign AI into enterprise data environments, including private cloud, public cloud, on-prem, and air-gapped deployments. 4) Anthropic published its September threat intelligence report, detailing how malicious actors tried to misuse Claude across cyber operations, surveillance, scams, and weapons-related workflows. 5) Positron AI raised $875M at a $5B valuation to build inference chips designed for large models and long-context workloads. 6) Nvidia is reportedly in talks to invest in Anthropic’s massive IPO, another sign that model companies and compute suppliers are becoming deeply tied together.

For crypto and AI, the questions remain practical: What does the product do? Who controls it? What rights does ownership actually provide? The final rules matter. So does building something people want to use.

Our take: clearer, workable rules could make it easier for teams to plan and institutions to participate. The benefit would be less uncertainty around building useful products. A vote alone cannot create demand or guarantee token prices.

The sponsors' latest draft includes proposed protections for software developers, rules addressing conflicts of interest, and new stablecoin provisions. These details matter because regulation shapes who can build, operate, and participate.

CLARITY is back in focus today. The Senate has a procedural vote scheduled for 2:15 p.m. ET on whether to advance the bill. This is a step toward debate, not final passage. Here's why crypto builders should still care.

Give your next AI agent a very specific job. Find the bug. Organize the research. Catch the missed detail. Pick a problem you actually have, then build something you’ll want to use again tomorrow. What are you working on this week?