How Commerce.com and Cloudflare are getting storefronts ready for AI agents
Almost 100,000 Commerce.com domains now expose shopping tools to agents via WebMCP on Cloudflare, with merchants keeping control.
AI Daily Recap
41 articles · 5 categories
The finishable daily brief
What happened in AI — Oct 21, 2026
Wednesday, Oct 21, 2026
41 articles · 5 categories
read top to bottom · then stop
In 30 seconds
- DeepSeek open-sources DeepInfer, its production inference engine.
- U.S. bans DeepSeek from federal civilian agencies.
- Commerce.com opens ~100,000 domains to agents via WebMCP on Cloudflare.
- Ant Group and vLLM release Humming quantized inference kernels.
- OpenAI models and Codex arrive in Google Cloud's Model Garden.
- Google makes SynthID Detector globally available.
DeepSeek's DeepInfer release dominated the day: an open-source production inference engine reportedly serving 155T tokens a day, released as the U.S. barred DeepSeek from federal civilian agencies.
Agent plumbing kept maturing: WebMCP storefronts at Commerce.com scale, cross-harness memory, and NVIDIA's expanded Agent Toolkit.
Agent interfaces, memory and harnesses 5 items
Builders are standardizing how agents reach apps and share state across coding harnesses.
Open-source, cross-harness memory for AI agents
Open-source memory layer shared across Claude Code, Codex and Cursor, so switching harnesses doesn't lose context.
Claude Mods: Flipping the Harness Inside Out
Claude Mods flips the agent harness to face the human instead of the machine.
Article Series: Agentic AI Architecture
InfoQ series frames agentic AI architecture as a distinct software architecture, with expert-written articles.
Using AI to Delete Code: Agentic Tech Debt Cleanup at Uber
Uber describes using agents to delete dead code and clear tech debt at scale.
DeepSeek open-sources its inference engine 3 items
DeepSeek released DeepInfer, the engine behind its production serving, with reported scale and margin figures.
[AINews] DeepSeek DeepInfer: Open Source SOTA Inference running 155T toks/day/cluster making $22m/day at ~80% margin for a 81 day payback
Reports DeepInfer serving 155T tokens a day per cluster at roughly 80% margin, framed as open-source SOTA inference.
DeepSeek open-sources DeepInfer, the engine serving 155T tokens a day
DeepSeek open-sources DeepInfer, the engine it says serves 155T tokens a day.
DeepSeek Open-Sources Framework Behind Its High-Margin AI Inference
The Information covers the release of the framework behind DeepSeek's high-margin inference.
Inference kernels and enterprise agent stacks 3 items
Open inference tooling and vendor agent platforms both shipped infrastructure for production deployment.
Announcing Humming: Fast, Flexible Kernels for Quantized Inference
Ant Group and the vLLM community release Humming, a quantized inference kernel library with dense and MoE results and vLLM integration.
NVIDIA Agent Toolkit Expands, Grounding Enterprise Agents in Company Data
NVIDIA expands its open-source Agent Toolkit to ground enterprise agents in company data within defined boundaries.
OpenAI models are coming to Google Cloud
OpenAI frontier models and Codex workflows come to Gemini's Model Garden via Google Cloud Marketplace.
DeepSeek ban and AI supply-chain risk 2 items
The U.S. barred DeepSeek tools from federal civilian agencies, and China threatened countermeasures.
US bans DeepSeek across federal agencies, putting AI supply chains under scrutiny
The federal DeepSeek ban puts AI supply chains under scrutiny for enterprise buyers.
US bans DeepSeek AI tools from federal civilian agencies
DeepSeek AI tools are banned from federal civilian agencies.
Security, provenance and disclosure 2 items
Disclosure coordination and watermark detection both moved toward cross-vendor tooling.
Patch the Planet: global response coordination as agents uncover critical vulnerabilities
OpenAI's Patch the Planet links researchers, maintainers and incident responders to handle critical open-source vulnerabilities found by agents.
Google SynthID Detector Simplifies Watermark Detection and Expands Support to Other AI Providers
Google DeepMind makes SynthID Detector globally available, with support for content from OpenAI, NVIDIA and other partners.
You are caught up for this edition