Daily AI intelligence for Iru.
Monday, May 11, 2026
xAI Ships Custom Voice Cloning to API
docs.x.ai · May 11

xAI added voice cloning to its API today, letting developers clone a voice from a short audio clip and deploy it across Text-to-Speech and Voice Agent endpoints.

Bottom line
xAI is pairing aggressive inference pricing with voice infrastructure, positioning for enterprise voice-agent workloads before OpenAI can lock in that surface.
Gemini Generates PDF, Word, Excel Files In-Chat
designcompass.org · May 11

Google Gemini can now produce PDF, Word, Excel, Google Docs, Sheets, and Slides files directly inside a chat session without requiring a copy-paste step to a separate app.

Bottom line
Gemini is absorbing a workflow previously owned by Copilot in Word and Excel, and it works across both productivity ecosystems at once.
Cursor Launches SDK for External Agent Builders
forgenex.com · May 11

Cursor released an SDK this week that lets developers build agents using the same runtime powering Cursor's native capabilities, opening its editor infrastructure to third-party tooling.

Bottom line
Cursor is making a platform bet: if agents run on its runtime, it owns the execution layer regardless of which model wins.
Nvidia Vera Rubin Confirmed for July Customer Shipments
wccftech.com · May 11

Nvidia confirmed Vera Rubin will ship to Microsoft, Google, Amazon, Meta, and Oracle in July, with mass production in the second half of 2026, squashing circulating rumors of design setbacks.

Bottom line
A July Vera Rubin delivery to all five major hyperscalers compresses the window before significantly cheaper inference capacity arrives in the market.
Nvidia nvCOMP Cuts LLM Checkpoint Costs $56K Monthly
bitrss.com · May 11

Nvidia's nvCOMP GPU compression library reduces LLM training checkpoint sizes 25-40%, saving teams up to $56,000 monthly on checkpoint storage and up to $222,000 monthly at larger scale.

Bottom line
Checkpoint compression is one of the few infrastructure levers that reduces AI training costs without touching model quality or hardware procurement.

Recaps circulating today of Anthropic's mid-market enterprise firm, finance agent suite, and SpaceX compute deal reveal a strategy to capture the segment between hyperscaler deployments and SMB SaaS before OpenAI can consolidate its PwC CFO automation channel.

← Back to latest