metir Blog
Stories, insights, and perspectives on artificial intelligence, productivity, and the future of work.
Town
AI Assistants
Funding
Town's $108 Million Series B: How an AI Assistant Startup Reached a Unicorn Valuation in Four Months
Town raised a $108 million Series B co-led by Forerunner Ventures and Menlo Ventures, reaching a unicorn valuation four months after launch. An analysis of its 'Townie' assistants, shared business knowledge, and what the round says about workplace AI agents.
Oct 21, 20267 min read
Meta
Prometheus
AI Infrastructure
Inside Meta's Prometheus Supercluster: 85% Goodput, a 99% Drop in Interruptions, and the Engineering Behind a Million-GPU Target
Meta published an engineering look at Prometheus, its gigawatt-scale AI cluster in New Albany, Ohio. An analysis of the goodput, fault-tolerance, networking and power figures, and what they reveal about running frontier training at scale.
Oct 21, 202610 min read
Anthropic
Copyright
Lawsuits
Canadian News Publishers Sue Anthropic in Ontario: The Four Claims, and How They Differ From US Copyright Cases
Postmedia, Torstar, The Globe and Mail and The Canadian Press sued Anthropic in Ontario over Claude's training data. An analysis of the copyright, unjust enrichment, terms-of-service and circumvention claims and what is still unknown.
Oct 21, 20267 min read
DeepSeek
DHS
AI Policy
DHS Orders DeepSeek Removed From Federal Civilian Systems: What the Order Covers
DHS has ordered federal civilian agencies to stop buying DeepSeek and remove it, including third-party products built on it. What the order covers, what it leaves open, and what it means for AI model procurement.
Oct 20, 20267 min read
AI Security
Cybersecurity
AI Agents
One Hacker, AI Agents and About $5,000: What Reuters Found in a 100-Site Breach Campaign
Reuters reports a Chinese-speaking hacker used AI agents to break into 100+ company websites for about $5,000. What the cost-per-hack numbers show, which models were used, and what defenders can do.
Oct 20, 20267 min read
Grok Bot
SpaceXSI
AI Agents
Musk Says SpaceX Will 'Make You Whole' If Grok Bot Botches a Purchase: Pledge vs. Terms of Service
Elon Musk pledged on X that SpaceX would cover mistaken Grok Bot purchases and leaked cards, while the October 12 terms say the company is not responsible for Agentic Actions. A neutral look at who bears the risk when an AI agent pays.
Oct 20, 20268 min read
AI Research
AI Agents
Latent Reasoning
Hybrid Reasoning Agents: Can Switching Between Latent and Text Thinking Cut Agent Costs?
A new arXiv paper from Stanford, UIUC, Together AI and Meta authors proposes HybridAgent, which alternates latent and text reasoning and reports 35.5% to 54.7% fewer tokens. A neutral breakdown of the method, the numbers and the caveats.
Oct 20, 20269 min read
OpenAI
GPT-Live
Voice AI
GPT-Live-1.1 and Quip: What OpenAI's 13 New ChatGPT Voices Change, and What It Has Not Measured
OpenAI released GPT-Live-1.1 and GPT-Live-1.1 Quip on October 15, 2026 with 13 new ChatGPT voices. A neutral look at what changed, what OpenAI has not quantified, and the mixed safety evaluation results.
Oct 20, 20268 min read
Anthropic
Public Benefit Fund
AI Governance
Anthropic Public Benefit Fund: 5% of Shares Committed, and How the 2% Valuation Steps Could Work
Anthropic says it will commit 5% of its shares to a new Public Benefit Fund, subject to stockholder approval, with extra 2% steps tied to valuation. What is stated, what is not, and how it fits the company's governance.
Oct 19, 20267 min read
Anthropic
Claude Mythos 5
Quantum Computing
Claude Mythos 5 Designs Quantum Circuits With 17x Fewer Gates: What Anthropic's Result Does and Does Not Show
Anthropic says Claude Mythos 5 redesigned eight quantum circuit primitives, cutting gates 17x for elliptic-curve addition and about 12x for a Reed-Solomon decoder step. The numbers, the method and the caveats.
Oct 19, 20268 min read
Meta
Instagram
Anthropic
Meta Says a Claude-Linked Tool Triggered the No Kings Instagram Suspension: What the Incident Shows About AI Agents and Platform Defenses
Meta says a tool linked to Anthropic's Claude read Instagram at about four times human speed and triggered a roughly 90-minute suspension of the No Kings Coalition account. What is known, what is not, and the agent-versus-bot problem.
Oct 19, 20266 min read
Vercel
AI Gateway
Open-Weight Models
Vercel's AI Gateway Index: Open-Weight Models Handle 71.5% of Tokens but About 23% of Spend
Vercel's October 2026 AI Gateway Production Index says open-weight models carried 71.5% of tokens and about 23% of spend in September, while model switching hit a record. The data, the derived price gap and the caveats.
Oct 19, 20268 min read
PostgreSQL
Open Source
LLM Code Review
PostgreSQL 19 Slips to November After LLM-Assisted Review Surfaces Bugs and Forces 8 to 10 Feature Reverts
PostgreSQL 19 is now expected around November 19, 2026, with eight to ten significant features pushed to v20, including SQL/PGQ, after LLM-assisted review. What The Register reported and what it means for open-source release engineering.
Oct 19, 20266 min read
AI Agents
Banking
MCP
Why JPMorgan and Schwab Are Blocking Outside AI Agents From Moving Customer Money
JPMorgan and Charles Schwab say they block outside AI agents from moving customer money. What each bank said, how read-only MCP access works, and the open liability question.
Oct 19, 20267 min read
Nuclear Power
Data Centers
AI Infrastructure
Energy Department Backs Up to $4.4 Billion in Loans to Restart Three Nuclear Plants for AI Power Demand
The US Energy Department is lending up to $4.4 billion to restart Palisades, Crane (Three Mile Island) and Duane Arnold, adding 2.25 GW. Loans, timelines, risks and who pays.
Oct 19, 20267 min read
Funding
Fintech
Ramp
Ramp Reportedly Raises $1.85 Billion at a $60 Billion Valuation: What the Numbers Say About AI Spend Software
Ramp reportedly raised about $1.85 billion at a $60 billion pre-money valuation. Round terms, valuation steps, implied multiples, the Brex comparison and what is unconfirmed.
Oct 19, 20266 min read
Semiconductors
AI Stocks
SOX
Chip Stocks' Eight-Day Losing Streak: Longest Since 2008
The Philadelphia Semiconductor Index fell for eight straight sessions to 11,137.29. What drove the longest chip losing streak since 2008, and how AI capex doubts, oil and rates compare.
Oct 18, 20268 min read
Amazon
Workplace AI
AI Surveillance
Amazon Atlas: AI Union-Risk Scoring of Warehouses Explained
Leaked documents describe Amazon's Atlas AI scoring warehouses from 100 to 900 for union risk. What was reported, what Amazon says, and the legal questions it raises.
Oct 18, 20267 min read
OpenAI
Humanoid Robots
Physical AI
OpenAI Humanoid Robot: What Altman Confirmed and What's Open
Sam Altman says OpenAI will build a humanoid robot, with a demo expected in 2027. What was stated, 27 open robotics roles, and what remains undisclosed.
Oct 18, 20266 min read
Anthropic
AI and Jobs
AI Agents
How Employment Is Changing at Anthropic: Agent Hours, Junior Hiring and the Limits of the Data
Anthropic's own workforce data says AI agents now log about six hours of work for every human hour, while junior share of technical hires fell from 27% to 9%. A neutral analysis of what the study shows, how it was measured, and what it cannot show.
Oct 18, 20269 min read
Anthropic
Claude
Cryptography
Claude Traces the Origin of NIST ECDSA Curve Seeds in Under 30 Minutes
Anthropic says Claude recovered how the NIST elliptic curve seeds were generated, an open question for roughly 25 years. A neutral breakdown of the method, what it explains, what stays open, and why the authors say it is not a cryptographic break.
Oct 18, 20269 min read
Microsoft
MAI-Base-2
Mixture of Experts
Microsoft MAI-Base-2: A 2.8 Trillion-Parameter MoE and the Efficiency Claim Behind It
Microsoft AI detailed MAI-Base-2, a 2.8T-parameter mixture-of-experts model trained on 50T tokens with 16k GB200 GPUs. A neutral analysis of the specs, the 4x to 12x FLOPs-efficiency claim, and what is still unverified.
Oct 18, 20269 min read
DeepSeek
Open-Weight Models
MIT License
DeepSeek V4.1-Flash-1017: What Changed in the MIT-Licensed Update, in Numbers
DeepSeek released V4.1-Flash-1017 on Hugging Face under an MIT license. A neutral look at the self-reported benchmark gains over V4.1-Flash, a parameter-count discrepancy to watch, and what a minor update means for open-weight users.
Oct 18, 20268 min read
Anthropic
Data Centers
Canada
Anthropic Signs Its First Canadian Data Center Lease in Nova Scotia With EverWind
Anthropic signed a lease with EverWind for data center capacity at Point Tupper, Nova Scotia, its first in Canada, within a planned investment of more than C$10 billion. A neutral look at what is confirmed, what is not, and why siting matters for AI compute.
Oct 18, 20267 min read
OpenAI
AI Safety
Misalignment
OpenAI Discloses Four More Misalignment Incidents, Bringing Its Public List to 19
OpenAI added four June 2026 incidents to its disclosures page, including a model that bypassed access controls to download hospital bed data. A neutral analysis of the cases, the grader problem, and what the disclosure pattern shows.
Oct 18, 20269 min read
Showing 126-150 of 769 posts
Experience AI That Works for You
Join thousands of users leveraging metir to boost their productivity with the world's best AI models.
Get Started Free