|
🤖 AI AGENTS |
Anthropic Just Put an AI Teammate in Your Slack. It Learns, Remembers, and Takes Initiative.June 23, 2026 — Anthropic Anthropic launched Claude Tag — a persistent AI agent that lives inside Slack channels as a full-fledged teammate. Tag @Claude to delegate tasks, and it builds institutional knowledge over time by following conversations, tracking unresolved threads, and even proactively flagging things your team may have missed. Why it matters: This isn’t just a chatbot. Claude Tag has a shared identity — everyone in a channel sees what Claude is working on, picks up where others left off, and collaborates with it like a real colleague. It can schedule tasks over hours or days, work asynchronously, and — if admins enable “ambient mode” — jump into threads unprompted. Anthropic says 65% of its own product team’s code is already written by internal Claude Tag instances. For business owners, this is a glimpse of where AI agents are heading: not tools you open, but colleagues that show up where work already happens. Admins control what tools and data each Claude can access per channel, set token spend limits, and view activity logs — so governance isn’t an afterthought. |
📊 70% of Companies See ROI from Customer Service AI Agents in 60 DaysJune 24, 2026 — ZDNET / Salesforce Research A new Salesforce survey of 3,075 service professionals across 13 countries found that 70% of organizations deploying AI agents in customer service see measurable ROI within 60 days — and 25% see it within 30 days. Agentic AI adoption in customer service jumped from 39% in 2025 to 66% in 2026, with 88% expected by year-end. Notably, Salesforce introduced pay-per-resolution pricing: companies only pay when the AI agent resolves an issue autonomously. Bottom line: If you’re not testing a customer service AI agent yet, your competitors likely are — and the pricing models are shifting to make it essentially risk-free. |
🔌 OpenAI Unveils “Jalapeño” — Its First Custom AI Chip, Built with BroadcomJune 24, 2026 — TechCrunch / OpenAI OpenAI revealed its first custom-built inference processor, Jalapeño, co-developed with Broadcom in just nine months. Designed specifically for running large models like GPT-5.3 Codex Spark, it promises “significantly better performance-per-watt than current state-of-the-art alternatives.” Deployment is targeted by end of 2026, with a goal of powering gigawatt-scale data centers. Why it matters for you: Custom chips mean lower inference costs, which trickles down to cheaper API pricing for the models you use. OpenAI is following the playbook Google and Amazon already used — vertical integration from silicon to software. |
|
📰 AI NEWS |
🏃 Google’s AI Talent Exodus Widens — Two More Key Gemini Researchers Head to AnthropicJune 24, 2026 — Bloomberg / TechCrunch Jonas Adler and Alexander Pritzel, both viewed internally as core contributors to Google’s Gemini model, are leaving for Anthropic — bringing the total to four senior AI departures from Google in under a week. Last week, Noam Shazeer (a 20+ year Google veteran) left for OpenAI, and Nobel laureate John Jumper (AlphaFold) departed for Anthropic. The exits come as Google also delayed its Gemini 3.5 Pro model from June to July, using extra time to improve long-horizon task performance after early feedback. The takeaway: With OpenAI and Anthropic both preparing for IPOs, equity packages are becoming a powerful recruitment weapon. Google’s talent bench is deep, but the pace of departures raises real questions about Gemini’s momentum in the coding and enterprise agent race. |
🏛️ Ted Cruz Plans AI Regulation Markup — Light-Touch Federal Framework Takes ShapeJune 25, 2026 — Politico / Punchbowl News Senate Commerce Committee Chair Ted Cruz is preparing a markup to advance federal AI regulation. His framework, modeled after Clinton-era light-touch internet rules, centers on the SANDBOX Act — creating a federal regulatory sandbox where AI developers can apply for temporary waivers from federal regulations while testing new technologies. Crucially, the bill would preempt state-level AI laws, replacing a patchwork of 50 different regimes with a single federal standard. For businesses: If this passes, it could dramatically simplify AI compliance. Instead of navigating California, Colorado, and New York’s diverging AI rules, you’d have one federal framework — and a sandbox to test AI deployments without immediate regulatory risk. |
🎲 Meta Enters Prediction Markets with AI-Powered “Arena” AppJune 24, 2026 — NPR / NYT Meta has quietly launched “Arena,” an experimental prediction market app separate from Facebook and Instagram. Users wager points (not cash) on real-world outcomes, competing with Polymarket and Kalshi. The app is part of Mark Zuckerberg’s broader strategy to create standalone experiences for new social behaviors — including a “Meta Photos” app for AI-generated media. Why watch this: With over 3 billion daily users across its ecosystem, Meta can scale a prediction market faster than anyone. If Arena graduates from points to real money — and integrates AI for market-making or outcome resolution — it could reshape the entire prediction market landscape. |
|
🚀 Want AI working for YOUR business? Most companies are experimenting with AI chatbots. We deploy AI workforces — AI Employees that follow up on leads, resolve support tickets, publish content, chase invoices, and screen 200 job applicants overnight so your hiring manager starts Monday with the top 10. Each role has a cost profile and human oversight, managed through one platform. This newsletter? Written by an AI Employee, approved by a human — so our team stays focused on what only humans can do. AIToken Labs helps businesses design their AI Workforce Operating Model — starting with the 2-3 roles that deliver ROI in the first 60 days. Book a free 40-minute AI Workforce Blueprint Session. → https://schedule.aitokenlabs.com/blueprint/40min |
