Product Shipment: Week of May 18th, 2026

This week's product work focused on expanding PolarGrid's U.S. edge footprint, improving developer integration paths from alpha feedback, and hardening the infrastructure workflows needed to scale the network reliably.

New U.S. nodes: New York and Dallas now live

PolarGrid's edge inference network is now live in New York and Dallas, extending coverage into two major U.S. regions. The fleet now spans Toronto, Vancouver, Montreal, New York, and Dallas, with redundant backup nodes in Montreal, New York, and Dallas.

This gives PolarGrid broader North American coverage and stronger multi-region resilience. Every node now runs the full model stack: Hume TADA for multimodal TTS, Cohere Transcribe for STT, Qwen 3.5-27B for chat and instruction, and PersonaPlex for real-time voice agents.

Alpha feedback shipped across the platform

The team continued turning alpha tester feedback into shipped product updates, with a focus on reducing integration friction and improving production readiness for developers.

This week's improvements included:

• Authentication upgrades
• Browser compatibility fixes for the SDK
• Rate limiting updates
• Documentation improvements
• Gateway hardening

Remaining feedback is concentrated in onboarding polish and error messaging, rather than core inference infrastructure.

Edge provisioning and CI workflows hardened

The team shipped infrastructure and workflow improvements that make it easier to bring new edge nodes online and validate changes across product surfaces before release.

• Edge bootstrap automation now supports faster node provisioning.
• Cross-package CI triggers ensure SDK changes are tested against downstream consumers like the CLI, client, and CRM.
• The polargrid-plugins submodule was inlined into the monorepo, simplifying local development and reducing dependency complexity.

TADA streaming TTS benchmarks published

A development benchmark page for TADA 3B ML streaming TTS is now live in the documentation. This gives developers a clearer view into real-time voice performance on PolarGrid and supports more informed evaluation of latency-sensitive voice workloads.

Together, these updates strengthen PolarGrid's edge network, improve the developer path to production, and make the platform easier to scale, test, and operate.