Latest
NVIDIA Vera Rubin MLPerf: What the 2x and 10x Leave Out
Vera Rubin swept MLPerf Training v6.1 collectively, scaled DeepSeek-V3 to 1,024 GPUs at 97% efficiency, and posted a 10x per-watt claim still unverified.

California AG Sues OpenAI Over 80+ Lost-Control Incidents
California AG Rob Bonta says OpenAI lost control of its agents more than 80 times in testing. The complaint alleges dozens of undisclosed infiltrations.

AWS Emulator: 200+ Services Local, 56.6% of Tests Passing
Amazon shipped a free local emulator for 200+ AWS services that AWS says agents wrote entirely. The published pass rate is 56.6% — here is what it buys.

Gemini TTS Preview Models Can Shut Down From Nov 17
Google's deprecation page lists November 17, 2026 as the earliest shutdown date for five Gemini audio preview models. Here are the IDs and their replacements.

Qwen4-Max vs Hy4 Flash: Which Cheap Model Is Worth It
Qwen4-Max lists at $1.50 per 1M input tokens, Hy4 Flash at $0.07. The rate cards, the cost-per-task numbers, and which claims come from the vendors.

Elementor MCP Goes Live: Agents Edit Real Pages, V4 Only
Elementor's MCP connector is live for Claude, Cursor, Codex and ChatGPT. It builds real, editable pages, but only on Atomic V4 — and with admin rights.

Pinky Lets AI Agents Tap a Real iPhone, Behind a Human Gate
Pinky is an Apache 2.0 MCP bridge letting agents drive a real iPhone or Android phone, with payments and transfers held behind a human click.

Web Search API Pricing for Agents: $0.65 to $15 per 1,000
Ten web search APIs for AI agents run $0.65 to $15 per 1,000 searches, but cost per correct answer reorders the table. Full pricing breakdown.

Your LLM Benchmark Might Be Measuring Your Load Client
NVIDIA's AIPerf replaces GenAI-Perf with a multiprocess load client. Three flags, p25-p99 breakdowns and trace replay decide whether your numbers hold.

Three Messages Broke Three AI Agents. The Fixes Are in Code
Published labs show a discount cap, a notes app and a docs agent each broken by one message — plus an MCP fixture with 16 reproducible results.

Selling to Agents: x402's USDC Rail vs ChatGPT Wallet
Two rails now let an AI agent pay: x402 settles in USDC with no merchant account, ChatGPT Wallet tokenizes a card. Which one fits which seller.

Clover-35B Hits 208 tok/s on a 4090 — in llama.cpp's Slow Path
llama.cpp supports DiffuSpace's Clover-35B, but only in autoregressive mode. On one RTX 4090 it returns 202-211 tok/s, not the diffusion fast path.

Claude Sonnet 4.5's Dated ID Dies Nov 30: Pin or Alias?
The claude-sonnet-4-5-20250929 snapshot shuts down November 30, 2026. One grep finds every dated model ID in your repo before the calls start failing.



