Flux 3 video with audio, Vercel MCP deploys code, and fresh agent tools
TL;DR
Vendors added deployment and evaluation tools while new models and agent products appeared across gateways and marketplaces.
What shipped
On 23 July several vendors released updates focused on agent workflows and model access. Vercel expanded its tooling for flags and deployments while new models reached gateways and marketplaces. Product Hunt and industry sources showed a wave of agent-focused launches and benchmarks.
Vendor launches
Vercel released multiple updates for its flags and deployment systems that target agent builders. Ling 3.0 Flash became available on AI Gateway with a focus on token-efficient runs. A Google report examined how teams adopt AI tools in daily work.
- •Vercel Flags evaluation view Vercel Flags added live per-minute charts on each flag detail page so teams can link performance shifts to specific version changes.
- •GitHub tools Eve extension Vercel released an installable extension that adds 42 GitHub tools with built-in auth and presets to any Eve agent.
- •Vercel CLI flag history The new vercel flags versions command prints revision history with author and environment details for quick audits.
- •Vercel MCP deploy tool The Vercel MCP server now deploys code to new or existing projects and returns a shareable URL directly from chat tools like Claude.
- •Ling 3.0 Flash on AI Gateway Ant Group's 124B-parameter MoE model reached AI Gateway with a 256K context window and free access through 3 August for agent runs.
- •Google ATLAS report Google published its first Activity Task Landscape and Adoption Study showing real usage patterns of its AI tools across teams.
Hugging Face trending
Two new models climbed the trending list on Hugging Face Hub. Both support standard inference paths and target different generation tasks.
- •Qwen3.6-35B-A3B-Uncensored-Genesis-Hermes-V5-GGUF LuffyTheFox released a 35B image-text-to-text model on the Hub built with the hermes library for fine-tuning and inference.
- •antares-1b fdtn-ai published a 1B text-generation model on the Hub built with the transformers library for quick local or hosted runs.
Product Hunt picks
Eleven new agent and companion products appeared on Product Hunt. Most focus on meeting support, testing, or local model safety for builders who avoid heavy coding.
- •AI Eyes A permission-first real-time sensing layer for AI companions that adds live environmental input.
- •Empathy Engine A human-in-the-loop support agent that monitors customer signals during live chats.
- •SwiftScale Software A QA agent that tests applications using plain-language instructions instead of scripts.
- •Mufal An undetectable AI copilot designed for live meeting transcription and notes.
- •ReExplain A tool that records what a user knows and surfaces gaps for further learning.
- •Plow Mac App A Mac application that runs GPT-5.6 agents locally using OpenClaw and Hermes runtimes.
- •Chimlo A notch-based tracker that monitors Codex and Claude Code activity on MacBooks.
- •HOL Guard A dedicated firewall product built to monitor and contain AI agent actions.
- •Fikry A deliberately mis-trained model that demonstrates the effects of bad data on outputs.
- •Drawsy An AI workspace that combines canvas drawing with live code execution.
- •AgentLoop An agent runner that spins up fresh Codex workers and critics on every cycle.
Other
Databricks and AWS published production guides for agent retrieval and evaluation. Ling-3.0-flash also appeared on OpenRouter with zero-cost access for testing.
- •AWS trade assistant case study Jefferies built an AI trade assistant that reduced front-office manual steps using Amazon Bedrock.
- •Amazon Bedrock agentic retrieval AWS released guidance on agentic retrieval patterns for its managed knowledge bases.
- •Strands and AgentCore evaluation blueprint AWS shared a production evaluation framework for agents using Strands and AgentCore.
- •Databricks frontier data agent A Databricks post compares a frontier data agent against general coding agents on quality and token cost.
- •Ling-3.0-flash on OpenRouter The 124B MoE model reached OpenRouter with 262K context and free pricing for input and output.
- •Databricks self-serve infrastructure Databricks described its internal vending-machine system for provisioning agent workloads.
- •Free Ling-3.0-flash tier OpenRouter added a free tier of Ling-3.0-flash so builders can test the model without usage fees.
Industry news
Black Forest Labs shipped Flux 3 with native audio support for short videos. AMD announced a new rack-scale AI system to compete with Nvidia offerings.
- •Flux 3 video model Black Forest Labs launched Flux 3, a multimodal model that generates up to 20-second videos with native audio.
- •ChatGPT health advice tiers OpenAI began rolling out Health in ChatGPT where paid users receive answers from the stronger GPT-5.6 Sol model.
- •AMD Helios rack system AMD unveiled its Helios AI rack-scale system that begins shipping to customers later this year.
What this means for you
For Vibe Builders: You can now point chat tools at Vercel MCP to ship code without leaving the conversation and test Ling 3.0 Flash for free on OpenRouter. Product Hunt listings give ready-made agents for meetings and testing that run locally or with minimal setup. Add HOL Guard or Plow to keep experiments contained while you ship.
For Non-techies: ChatGPT health answers now differ by subscription tier so check your plan before relying on medical queries. Flux 3 brings audio to short AI videos while new Mac apps let you run models safely in the background. Watch for free access windows on models like Ling 3.0 Flash before paid tiers appear.
For Developers: Vercel added CLI history inspection and live flag evaluation charts that tie directly to version changes for faster debugging. Databricks and AWS released evaluation blueprints and agentic retrieval patterns you can benchmark against current stacks. Test the free Ling-3.0-flash endpoint on OpenRouter before committing tokens to production runs.
What to watch next
Track how teams adopt the new Vercel deploy tool in the next few days. Watch OpenRouter and AI Gateway pricing changes for Ling 3.0 Flash once the free period ends. Monitor AMD Helios early benchmarks against existing Nvidia rack deployments.
Harsh’s take
The day split between infrastructure tweaks from Vercel and model access expansions elsewhere. Many launches still require manual wiring even when marketed as one-click. Builders should pick one new tool such as the Vercel MCP deploy flow and run a full end-to-end test this week before the next wave arrives.
by Harsh Desai
Sources
Vendor launches
- •Evaluation metrics for Vercel Flags
- •GitHub tools are now an installable eve extension
- •Inspect feature flag history with Vercel CLI
- •Vercel MCP can now deploy code
- •Apply now for the Google for Startups Gemini Startup Forum.
- •Ling 3.0 Flash is now available on AI Gateway
- •GeForce NOW Sets Sail With ‘Path of Exile: Curse of the Allflame’ Joining the Cloud
- •Understanding the AI economy
- •NVIDIA AI Supercomputer Comes Online at Naval Postgraduate School
- •Drive with Halo on Waze
Hugging Face trending
- •Qwen3.6-35B-A3B-Uncensored-Genesis-Hermes-V5-GGUF by LuffyTheFox trends on HuggingFace
- •antares-1b by fdtn-ai trends on HuggingFace
Product Hunt picks
- •AI Eyes
- •Empathy Engine
- •SwiftScale Software
- •Mufal
- •ReExplain
- •Plow Mac App
- •Chimlo
- •HOL Guard
- •Fikry
- •Drawsy
- •AgentLoop
Other
- •Building trade assistant: How Jefferies optimized front office trading operations with AI
- •Building multi-Region visualizations with Highcharts in Amazon Quick
- •Agentic retrieval for Amazon Bedrock Managed Knowledge Base
- •Evaluating AI Agents: A production blueprint with Strands and AgentCore
- •Why A Frontier Data Agent Outperforms General Coding Agents in Quality and Cost
- •Ling-3.0-flash now available on OpenRouter (262k context, $0.00/M in, $0.00/M out)
- •Provisioning for the Agentic Era: How Databricks Built a Self-Serve Infrastructure Vending Machine
- •Ling-3.0-flash (free) added on OpenRouter
Industry news
- •Meta launched a new AI optimism ad set to a song about human extinction
- •Flux 3 generates videos with native audio up to 20 seconds long
- •Meta’s New Feel-Good AI Ad Uses a Song About the World Ending
- •ChatGPT will give you worse health advice if you don't pay
- •AMD takes on Nvidia with its Helios AI rack-scale system
More AI news
- LaunchRunway launches Runway Media Router through Runway Dev platform
Runway launched Runway Media Router on its Runway Dev platform. The router provides API access to third-party image, video and audio models.
- FeatureDependabot adds three-day cooldown before version updates
Dependabot now defaults to a three-day cooldown that delays version update pull requests.
- FeatureHermes-agent enables smart approvals by default and adds user-defined deny rules
Hermes-agent now defaults to smart LLM approvals for flagged commands and evaluates each independently. It adds user-defined deny rules that block commands even in yolo mode with explanatory feedback via /deny.