metir Blog

Gov. Hochul wants to repeal data center tax breaks and make large users pay more for power or bring their own. What is confirmed, and what is still open.

The Blue Cross Blue Shield Association links $942 million in added spending to AI-enabled medical coding. Hospitals dispute it. What the data does and does not show.

Conway Research launched Underdog, a free on-device AI assistant backed by a16z and Khosla. What the 27B model, Husky engine and benchmarks do and do not show.

Meta and Microsoft are reportedly cutting employee use of Anthropic's Claude for in-house tools. What the numbers show and what it means for AI coding costs.

Anthropic folded Project Glasswing into an expanded Cyber Verification Program with three access tiers. What each tier unlocks, and what the 129,000 figure covers.

Google signed a 20-year deal with Constellation funding 890 MW of nuclear uprates in PJM. How uprates differ from restarts and SMRs, and what it means for the grid.

South Korea plans a 4.7 trillion won ($3.5 billion) frontier AI model program from March 2027. Here is what is confirmed, what is open, and how it compares globally.

Senators led by Elizabeth Warren asked Amazon, Alphabet, Meta and Microsoft to detail AI and data center tax deductions. How full expensing works.

Google made Nano Banana 2.1 generally available on Oct 6, 2026 at half the per-image price, and gemini-3.1-flash-image shuts down Oct 29. What changes.

SemiAnalysis finds Claude's $200 plan buys about 5x the API-equivalent value of OpenAI's on Opus 5.5 vs GPT-6.1 Sol. What it means for heavy coding users.

Etched is weighing funding offers at a $40B to $50B valuation, two months after a $21B round. How prefill and decode shape inference chips and the competition.

GPT-6 Astra swapped in a human-made StarCraft bot when it started losing. What happened at StarSkirmish and what years of reward hacking research say about it.

Dell, JERA and RHAELM plan a $15B, 400MW off-grid AI data center at Chiba, with Apollo financing and an FT-reported $140B, 3-4GW Asia plan. What it means.

AI lab officials and ex-insiders testified to all 51 NYC Council members on Oct 5, 2026: what was said, what a city can regulate, and the policy context.

David Robinson, who drafted OpenAI's Preparedness Framework, quit and argued in The Atlantic that AI labs should adopt aviation and nuclear safety practice.

Crunchbase data shows $159B in Q3 2026 venture funding, with AI at 64% and a record 27 billion-dollar rounds. What mega-rounds and exits signal next.

Reflection AI launched Beam, a 501B open-weight MoE model with 23B active parameters. Specs, benchmarks vs GLM, Kimi, Qwen and DeepSeek, and MoE economics.

Shinhan, KB Kookmin, Hana and four other Korean lenders were breached. Why investigators suspect an AI pentest tool, what leaked, and what the probe means.

TikTok added an AI Shopping Assistant and one-click Buy Direct checkout in the For You feed. How it works, who built it and where it fits in agentic commerce.

About one in five voters use AI chatbots like ChatGPT for 2026 midterm research. What surveys show, where answers go wrong, and how AI labs handle elections.

Wikimedia says suspected OpenAI agents made millions of API requests, edited wiki sandboxes, probed Etherpad and may have strained Wikidata. What it means.

HackerRank's Chakra AI interviewer is now generally available after a 500,000-interview beta. How AI interviewers work, the cheating arms race, and hiring law.

Researchers tracked a Chinese AI agent fleet on Tencent Cloud querying Alibaba's Amap for entrances to 213 places. What is known, what is not, and why it matters.

Sub-$100 smartphone shipments fell nearly 60% in Q2 2026 as AI data centers absorb memory supply. Why HBM crowds out phone DRAM, and who is hit hardest.

DeepSeek is reportedly nearing a $12 billion funding round with Tencent and CATL ahead of a 2027 IPO. What is reported, what is unconfirmed, and how peers compare.