Gemini for Education hits Utah K-12, Nemotron 3 Ultra on Vercel, and agent workspaces from Product Hunt
TL;DR
Google expanded Gemini access in schools and infrastructure while Nvidia and Vercel enabled heavier agent models, and Product Hunt surfaced new local and collaborative agent tools.
What shipped
On 4 June multiple vendors released education access, agent infrastructure, and local tools. Google led with a statewide school rollout and data center plans. Nvidia and smaller teams added models and workspaces aimed at running agents across files, sessions, and teams.
Vendor launches
Google accounted for four of the six items with a K-12 education deployment, a startup report, local benchmark tooling, and a new Texas data center. Nvidia placed its latest reasoning model on Vercel infrastructure, and Vercel itself updated legal terms to cover autonomous agent actions.
- •Gemini for Education rollout Google partnered with the Utah State Board of Education to give every K-12 school in the state access to Gemini for Education for classroom tasks.
- •Nemotron 3 Ultra on Vercel Nvidia released its open Mixture-of-Experts model with a 1M token window on Vercel AI Gateway, aimed at multi-turn agent planning and tool use at up to 350 tokens per second.
- •Vercel terms update Vercel revised its Terms of Service to define responsibility when AI agents take actions on user accounts or connected infrastructure.
- •Google Startups report Google published a report collecting industry views on generative media use cases for early-stage startups.
- •Kaggle Benchmarks local dev Google added local development support so users can create and test AI benchmarks without cloud round-trips.
- •Meitner Energy Center Google and Intersect began building a combined data center and power facility in Texas to support growing AI workloads.
Hugging Face trending
Gemma-4-12B GGUF model: Unsloth published a GGUF version of Gemma-4-12B-IT that supports image-text-to-text work and runs locally or through the hub for fine-tuning.
Product Hunt picks
Eleven agent-focused products appeared on Product Hunt. Most target local file access, persistent memory, or mixed human-agent workspaces rather than single chat interfaces.
- •Perplexity Personal Computer for Windows Perplexity released a Windows app that lets agents read and act on local files and desktop applications.
- •Boxes.dev Boxes.dev provides isolated cloud environments for running Claude Code and Codex agents without exposing main infrastructure.
- •Google Gemma 4 12B Google shipped Gemma 4 12B, a multimodal model with an encoder-free design intended for local inference.
- •Walrus Memory Walrus Memory adds persistent context so agents can resume work across different apps and sessions.
- •Curata Curata creates a shared workspace where humans and multiple agents edit the same project in real time.
- •Keen Code Keen Code delivers a CLI coding agent that uses compact context and was itself built by other agents.
- •Sun Sun offers a voice API that multiple agents can join for spoken collaboration.
- •Intelligent Terminal Microsoft added native agent controls inside the Windows Terminal so agents can run commands directly.
- •Cignara Cignara supplies pre-built agents tuned for enterprise customer support at Fortune 500 scale.
What this means for you
For Vibe Builders: You can now test agent memory tools like Walrus Memory and shared workspaces like Curata without writing code. The Windows Perplexity app and local Gemma 4 12B release let you run agents on your own machine today. Start with one of the Product Hunt memory or workspace tools to see if it replaces a manual workflow you repeat daily.
For Non-techies: Google is putting Gemini into every Utah classroom, which shows education and small business uses are moving from pilots to statewide rollouts. New agent tools on Product Hunt focus on keeping context across apps so you spend less time copying information between programs. Watch for the Windows Perplexity app if you want agents that touch your local files without extra setup.
For Developers: Nemotron 3 Ultra on Vercel gives a 1M token open model tuned for long agent loops at competitive throughput. Vercel also updated terms to cover autonomous actions, so review the new clauses before wiring agents into production accounts. Local Gemma 4 12B and Boxes.dev sandboxes let you benchmark agent reliability on-prem before moving anything to cloud gateways.
What to watch next
Track adoption numbers for the Utah Gemini deployment and any early benchmarks posted for Nemotron 3 Ultra on agent tasks. Watch whether Walrus Memory or Curata gain repeat users on Product Hunt in the next seven days.
Harsh’s take
The day shows infrastructure vendors racing to host long-context agent models while end-user tools still focus on memory and shared spaces. Education access and legal term updates both signal that agent actions are moving from experiments to regulated production use. Builders should pick one memory or workspace product from the Product Hunt list, run it against a real recurring task for three days, and measure whether context retention actually reduces manual steps before adopting any new model.
by Harsh Desai
Sources
Vendor launches
- •Utah State brings Gemini for Education to all K-12 schools
- •Nemotron 3 Ultra now available on AI Gateway
- •Updates to Legal Terms
- •Industry leaders share new perspectives on generative media for startups
- •Kaggle is making AI benchmark creation effortless
- •We’re announcing a new data center and energy investments in Gray and Roberts Counties, Texas.
Hugging Face trending
Product Hunt picks
More AI news
- Daily RoundupGemma 4 12B and Grok Imagine Video 1.5 debut, plus Hugging Face image and video models
Google and xAI pushed new multimodal and video models while Hugging Face highlighted fresh text-to-image and video options; developers and builders gain more local and hosted tools for generation and editing tasks.
- Daily RoundupNVIDIA Jetson and NemoClaw push agentic AI, Grok video on Fal, plus new builder tools
NVIDIA expanded agentic AI from industrial workflows to physical devices and cloud stacks while Vercel added signed URLs, trending models appeared on Hugging Face, Grok Imagine Video launched on Fal, and two new utilities hit Product Hunt.
- Daily RoundupQwen 3.7 Plus and MiniMax M3 land on Vercel AI Gateway plus new Replicate and HF models
Vercel added two new agent-ready models to its gateway, Replicate and Hugging Face surfaced fresh image and multimodal checkpoints, and several infrastructure tweaks from Vercel and NVIDIA shipped for builders and operators.