Vercel AI Gateway data opens, OpenAI prompting guide, and agent tools for builders

What shipped

On 13 July multiple vendors released production tooling and research updates. Vercel led with open data on its AI Gateway alongside deployment controls, while OpenAI and independent labs focused on practical prompting and autonomous agents. The releases point to wider availability of usage data and simpler ways to run agents in real applications.

Vendor launches

Vercel shipped the largest set of updates, opening production traffic data from its AI Gateway and adding CLI controls for flags and deployment policies. Google added campaign grouping and Waze personalization features tied to Gemini. The changes give teams clearer visibility into model usage and tighter control over releases.

• •AI Gateway leaderboards Vercel made its production traffic leaderboards open with downloadable data and shareable charts covering trillions of tokens. Teams can now compare model adoption against the July Production Index without custom queries.
• •AI Gateway Index July 2026 Open-weight models reached 29 percent of routed token volume while price per token stayed flat. The report shows 29 percent month-over-month growth in traffic through the gateway.
• •Waze Gemini updates Waze added motorcycle mode and reduced voice prompts alongside new Gemini features that adjust navigation behavior based on rider preferences.

Hugging Face trending

Two new 9B models from empero-ai appeared on the Hugging Face trending list. Both support GGUF and transformers formats, giving builders immediate options for local text and image-text inference.

• •Qwythos-9B-v2-GGUF empero-ai released a GGUF-quantized image-text-to-text model that is now trending on the Hugging Face Hub. Users can download and run it locally for multimodal tasks.
• •Qwythos-9B-v2 empero-ai published the base 9B text-generation model on Hugging Face with transformers support. It offers fine-tuning and inference options for developers testing smaller open models.

Product Hunt picks

AWS released a new UI for generative AI inference recommendations in SageMaker and shared a case study of an agentic solution built with Bedrock. A separate post discussed AI as an accessibility tool for neurodivergent users.

• •SageMaker inference UI AWS added a visual interface inside SageMaker AI that recommends inference settings for generative models. Teams can compare cost and latency options before deployment.
• •AI accessibility post AWS published guidance on using AI tools to support users whose cognitive styles differ from standard interfaces. The post frames AI features as practical accommodations rather than optional extras.
• •Bluesight Bedrock agent Bluesight built an agentic workflow on Amazon Bedrock that automates medication inventory checks. The case study shows how the setup reduced manual review steps in a healthcare setting.

Other

OpenAI published a prompting guide aimed at everyday users, Rich Sutton launched Oak Lab to pursue self-learning agents, and executives from OpenAI and Microsoft commented on data-center plans and model distribution risks. Anthropic research on model internals also drew coverage.

• •OpenAI prompting guide OpenAI released a four-block framework (goal, context, format, constraints) that tells users to describe the desired result instead of step-by-step instructions. The guide covers both chat and code models.
• •Oak Lab launch Turing Award winner Rich Sutton started Oak Lab to build agents that learn continuously from their environment rather than relying on static training runs.
• •Anthropic research review MIT Technology Review analyzed Anthropic's latest papers on model internals and noted what the findings do and do not demonstrate about model capabilities.
• •Nadella AI warning Microsoft CEO Satya Nadella cautioned companies about relying on proprietary models from a small number of labs, citing risks of hidden dependencies.

Industry news

• •Sam Altman’s space data center trash talk is what most experts already believe Responding to Musk accusing him of being a scammer, Altman said, "homeboy you're the one sellling [sic] public market investors on short-term space datacenters."
• •Should AI help you get away with killing your spouse? What does a world of total user-aligned AI actually look like?
• •OpenAI's new prompting guide tells users to stop overthinking and start with the result OpenAI has released a prompting guide for everyday users, not developers. Instead of rigid formulas, it suggests four optional building blocks: goal, context, format, and constraints. The core advice is simple: describe the result you want, not the steps to get there. It's the first time OpenAI covers both Chat and Codex in a single framework.
• •Turing Award winner Rich Sutton founds Oak Lab to build AI agents that learn on their own Richard Sutton, 2024 Turing Award winner and co-founder of modern reinforcement learning, has launched a new startup called Oak Lab in Toronto. He calls current deep learning methods "weak and inefficient" and wants to build AI agents that learn continuously from their environment.
• •What Anthropic’s latest AI discovery does, and doesn’t: show This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Anthropic: currently the world’s most valuable AI company, with a nearly $1 trillion valuation: has a reputation for publishing strange and heady research. It’s looking into whether AI models can feel pain, for example,…
• •Satya Nadella has issued a shocking warning to companies using AI Of all the debates raging about the potential downsides of AI, there is one worry causing the most hand-wringing among AI enthusiasts in Silicon Valley, that the giant AI labs that sell proprietary models are somehow acting like Trojan horses.

What this means for you

For Vibe Builders: You can now pull live production rankings from Vercel AI Gateway data and test the new 9B Qwythos models on Hugging Face without writing custom pipelines. OpenAI's result-first prompting guide gives a quick way to improve outputs from both chat and code models. Oak Lab's focus on self-learning agents points to future tools that may reduce manual fine-tuning.

For Non-techies: OpenAI's new prompting guide shows how to get better answers by describing the outcome you want instead of listing steps. AWS case studies demonstrate AI handling inventory checks and accessibility needs in daily operations. These updates make it easier to try production-grade agents without deep technical setup.

For Developers: Vercel opened its AI Gateway traffic data and added CLI management for flags and deployment policies, letting you script production controls. The July index shows open-weight models at 29 percent of volume, useful for deciding when to test local 9B models. Oak Lab's reinforcement-learning direction and Anthropic research provide signals on where agent reliability work is heading.

What to watch next

Track Oak Lab announcements and any follow-up from Rich Sutton on self-learning agents. Watch for updates to the Vercel AI Gateway Production Index next month and new model releases from empero-ai on Hugging Face.

Sources

Vendor launches

• •AI Gateway leaderboards, now with open data and shareable charts
• •Configure which sources can create deployments with Deployment Policies
• •Manage Vercel Flags targeting rules from the CLI
• •Web Analytics and Speed Insights are now more cost-efficient
• •Open-weight models surge to 29% of volume, price per token flattens
• •Optimize your reach and frequency across campaigns with video campaign groups.
• •Waze rolls out new customization features and more Gemini updates

Hugging Face trending

• •Qwythos-9B-v2-GGUF by empero-ai trends on HuggingFace
• •Qwythos-9B-v2 by empero-ai trends on HuggingFace

Product Hunt picks

• •TailMux

Other

• •Launching UI for generative AI inference recommendations in Amazon SageMaker AI
• •When your brain works differently, AI isn’t a luxury: it’s accessibility
• •Building an agentic AI solution at Bluesight with Amazon Bedrock

Industry news

• •Sam Altman’s space data center trash talk is what most experts already believe
• •Should AI help you get away with killing your spouse?
• •OpenAI's new prompting guide tells users to stop overthinking and start with the result
• •Turing Award winner Rich Sutton founds Oak Lab to build AI agents that learn on their own
• •What Anthropic’s latest AI discovery does, and doesn’t: show
• •Satya Nadella has issued a shocking warning to companies using AI