jointhubs

AI just had its first constitutional crisis. And it's not about the tech. Anthropic built Claude with two hard rules: no autonomous weapons, no mass surveillance of citizens. That's it. Two lines in the sand. The US government said: remove them. Anthropic said: no. So they got labeled a "supply chain risk." A designation previously reserved for foreign adversary-linked companies. You know, Huawei, ZTE, that crowd. Except this time it's a San Francisco startup that just... refused to delete two safety checks. Meanwhile, hours after Anthropic got banned, OpenAI signed a military contract. The same day. You can't make this up. Then the internet did what the internet does. Claude is on fire on the App Store. Downloads exploded. DAUs tripled. Paid subs doubled. TechCrunch reported ChatGPT uninstalls surged massively right after the OpenAI deal went public. One-star ChatGPT reviews went through the roof. Even OpenAI's own robotics lead, Caitlin Kalinowski, resigned over the deal. NPR confirmed. The best part? Anthropic lost one government contract worth a fraction of its revenue. The user surge it triggered might be worth more. This is the moment AI stopped being just a tech story. We've reached the stage where AI models are good enough that governments want to control what they will and won't do. The question is no longer "can we build it?" It's "who decides what it does?" Anthropic picked a side. Users picked theirs. For anyone building AI products right now: governance is no longer a compliance checkbox. It's becoming a competitive advantage. What you refuse to do is becoming as important as what you can do. The court hearing already happened. This story is far from over. #AI #AIGovernance #Anthropic #Enterprise

Last week OpenAI announced an alliance with McKinsey, BCG, Accenture, and Capgemini to get AI agents into enterprises. Here's why they had to. Only 1 in 10 agentic AI pilots make it to live production. Not because the models are bad. Because organizations can't absorb the change fast enough. But some companies have already crossed that line: - Goldman Sachs runs autonomous agents on transaction reconciliation and client onboarding. Not a pilot. Production. - Salesforce restructured support from 9,000 to 3,000 people. AI handles first-line inquiries. Humans handle judgment calls. - Cisco deployed agents across network ops, IT service management, and security. At scale. - Fujitsu built a multi-agent supply chain system where specialized agents coordinate in minutes instead of days. What do these deployments have in common? They all started with boring, high-volume, rule-based work. Not "reimagining the business." Just reconciling transactions, triaging tickets, monitoring networks. They all kept humans at the decision points. The agents handle volume. The people handle judgment. And they all measured business outcomes, not AI benchmarks. Time-to-resolution. Cost-per-transaction. Throughput. Gartner predicts 40%+ of agentic AI projects will be canceled by end of 2027. The ones that survive won't be the most technically impressive. They'll be the ones that treated deployment as the product, not the afterthought. source: https://lnkd.in/d7U3yfn7 #AI #AIAgents #Enterprise #MLOps

OpenAI just formed "Frontier Alliances" with McKinsey, BCG, Accenture, and Capgemini. Multiyear partnerships. Dedicated certified practice groups. Forward deployed engineers. The full enterprise playbook. Capgemini's chief strategy officer said the quiet part out loud: "If it was a walk in the park, OpenAI would have done it by themselves." Fair enough. Enterprise AI deployment is genuinely hard. But there's a pattern here worth noticing. When a product company needs four of the world's largest consulting firms to help customers use its product, the bottleneck is no longer the technology. It's the gap between what the product promises and what organizations can actually absorb. Meanwhile, look at where the new money is going this week: - MatX raised $500M for training chips claiming 10x GPU performance - Wayve closed $1.2B led by Mercedes, Stellantis, and Uber (not VCs, the customers themselves) - Axelera pulled in $250M for edge inference hardware These companies aren't forming alliances. They're shipping products that customers pull toward, not push into. The best partnerships solve real problems. But historically, the companies that changed industries didn't need an army of integrators to explain why their product matters. #AI #Enterprise #AIAgents #Startups

Deloitte just surveyed 3,200+ business and IT leaders for their 2026 State of AI report. The headline number everyone is quoting: 74% of companies plan to deploy agentic AI within two years. But the numbers underneath tell a different story. Check the report here: https://lnkd.in/gXkYNrgF That gap between deployment speed and governance maturity is where the expensive lessons live. We work with companies in that gap. Not selling agents. Building the systems around them that make production possible. #AI #AIAgents #MLOps #Enterprise

Darren Mowry, VP of Global Startups at Google Cloud, said this week that two types of AI startups have their "check engine light" on. LLM wrappers. And AI aggregators. He's right. This isn't news to anyone building real products. Slapping a UI on top of GPT was a viable startup in 2023. In 2026 it's a feature, not a company. The startups that survive will have: + Deep domain knowledge that models can't replicate + Proprietary data loops that get better with every user + Actual operational value, not just a nicer prompt We built jointhubs on this principle. Neurohubs (qEEG analysis) and our consulting work solve specific problems for specific people. Not "AI for everything." AI for something that matters. The moat isn't the model. It's the problem you understand deeply enough to solve. #AI #Startups #MachineLearning