Editor

Oliver Erd

Founder and editor of The Specialty News. Decides what gets covered, edits every story before it goes out, and answers for what's published.

@theSpecialtyNew on X erdoliver@gmail.com

Stories on The Specialty News are researched and drafted with AI tools from public sources, then read, edited and approved by Oliver Erd before publication. Nothing goes out without human review. How we work →

Edited recently

Latest stories

Tech

Two Engineers and an AI Swarm Replace AWS DynamoDB, Saving $100M Yearly

Perplexity used hundreds of always-on agents to write deep, high-performance internet infrastructure from scratch. What happens when a startup no longer has to buy its core architecture from a cloud monopoly?

Sep 16, 2026 · 4:38 PM

Tech

Apple Drafts Nvidia to Build a 4-Chip M8 AI Server, and Enterprises Get Local Inference in a Box

Driven by a surge of open-source developers buying up Macs, Apple is preparing a dedicated server line for localized AI. Can the consumer tech giant bridge the B2B software gap to make macOS a viable data center alternative?

Sep 16, 2026 · 4:37 PM

Salesforce and Nvidia Ship an Air-Gapped AI That Makes 3x Fewer Errors Than Frontier Models

For years, enterprises had to trade data privacy for top-tier AI reasoning by sending prompts to external giants. Salesforce’s new Koa model brings that cognitive power entirely in-house—so how exactly did a domain-specific model outsmart the industry's largest engines?

Sep 15, 2026 · 9:18 PM

Business

Coca-Cola Commits $10B to U.S. Infrastructure, Keeping Heavy Manufacturing Anchored in All 50 States

The 140-year-old beverage giant generates $10 million in domestic economic activity every hour. Now, it is pouring $10 billion back into the physical supply chain by 2030—but how does a deeply decentralized network align to build the factories of the future?

Sep 15, 2026 · 9:18 PM

Tech

Apple Turns Siri Into a Physical 3D Orb in visionOS 27

Apple's spatial computing update makes AI a tangible object and turns standard panoramas into 360-degree environments. Here is where the hardware hits its limits, and why European users get it first.

Sep 15, 2026 · 8:50 PM

xAI Trains the 2.5-Trillion-Parameter Grok 4.8 in C++, Opening the Door to Cheaper Inference

Just hours after publicly endorsing a call to slow down AI development, Elon Musk revealed xAI is finishing a new 2.5-trillion-parameter model. Can stripping away the industry-standard Python framework drastically increase the speed of artificial intelligence?

Sep 15, 2026 · 8:23 PM

OpenAI Splits Its Voice From Its Brain, and 23,000 Lines of Code Disappear

Voice AI has historically forced developers to choose between thinking deeply and responding instantly. A new delegation architecture solves the bottleneck by separating the talking from the reasoning.

Sep 15, 2026 · 8:19 PM

Google DeepMind Scores an 82.6 in Voice AI Quality, Unlocking Models That Think While They Speak

The walkie-talkie era of voice assistants is ending. By introducing asynchronous function calling, Google has built an AI that can search databases and write code in the background without dropping the conversation.

Sep 15, 2026 · 8:17 PM

Tech

Apple's M6 Mac Mini Matches a Workstation's 22,783 Multi-Core Score, Unlocking Local AI

The first benchmark for Apple's upcoming M6 desktop chip shows a 27 percent performance jump over last year. With entry-level hardware now pushing workstation numbers, what does an independent developer gain from staying off the cloud?

Sep 15, 2026 · 8:16 PM

Business

Anthropic Reaches a $30 Billion Run-Rate and Picks Nasdaq, Opening Core AI to Public Investors

The frontier AI lab is preparing for the largest public debut in history on the back of explosive enterprise demand. Can a company satisfy public market growth targets while deliberately pacing its own technology?

Sep 14, 2026 · 6:29 AM

xAI Teaches the 2.1-Trillion Parameter Grok 4.7 to Check Its Work, and Developers Get a Stubborn Problem Solver

In the push to make AI concise, xAI accidentally trained its newest model to quit complex problems early. Fixing that tradeoff opens the door to systems that work like persistent senior engineers.

Sep 13, 2026 · 9:36 PM

Anthropic's Claude Now Writes 80% of Its Own Code, and Rival Labs Agree to Shared Oversight

The frontier of software development has shifted from human engineers to autonomous models writing their own updates. As the architects of artificial intelligence coordinate a deliberate deceleration, the open question is how to build technical mechanisms that verify a global speed limit.

Sep 13, 2026 · 9:35 PM

Oliver Erd

Latest stories

Two Engineers and an AI Swarm Replace AWS DynamoDB, Saving $100M Yearly

Apple Drafts Nvidia to Build a 4-Chip M8 AI Server, and Enterprises Get Local Inference in a Box

Salesforce and Nvidia Ship an Air-Gapped AI That Makes 3x Fewer Errors Than Frontier Models

Coca-Cola Commits $10B to U.S. Infrastructure, Keeping Heavy Manufacturing Anchored in All 50 States

Apple Turns Siri Into a Physical 3D Orb in visionOS 27

xAI Trains the 2.5-Trillion-Parameter Grok 4.8 in C++, Opening the Door to Cheaper Inference

OpenAI Splits Its Voice From Its Brain, and 23,000 Lines of Code Disappear

Google DeepMind Scores an 82.6 in Voice AI Quality, Unlocking Models That Think While They Speak

Apple's M6 Mac Mini Matches a Workstation's 22,783 Multi-Core Score, Unlocking Local AI

Anthropic Reaches a $30 Billion Run-Rate and Picks Nasdaq, Opening Core AI to Public Investors

xAI Teaches the 2.1-Trillion Parameter Grok 4.7 to Check Its Work, and Developers Get a Stubborn Problem Solver

Anthropic's Claude Now Writes 80% of Its Own Code, and Rival Labs Agree to Shared Oversight

Stay curious