Generative AI pivots, edge AI breakthroughs, AI standards, ethics…

Generative AI Revolution

Meta’s “Avocado” model may go paid, ending its open‑source push

Meta is reportedly developing a new large‑language model codenamed Avocado, which insiders say could be offered behind a paywall rather than as open‑source software. The Verge notes that this marks a sharp reversal from Mark Zuckerberg’s earlier “open source is the path forward” stance, while Engadget confirms the model’s internal code name and hints at a 2026 release timeline. Analysts view the move as a strategic bid to monetize Meta’s AI investments after the underwhelming rollout of Llama 4. The Verge Engadget

Tiiny AI unveils the world’s smallest AI supercomputer for edge workloads

Startup Tiiny AI introduced the “AI Pocket Lab,” a handheld device capable of running 120‑billion‑parameter models on‑device, leveraging the latest ARM v9.2 cores. Wccftech highlights how the pocket‑sized supercomputer could democratize edge AI by eliminating costly cloud‑based inference, positioning it as a game‑changer for developers seeking low‑latency, privacy‑preserving AI. Wccftech

Mistral AI launches Devstral 2, an open‑weights coding model that rivals proprietary options

French AI firm Mistral AI released Devstral 2, a 123‑billion‑parameter open‑weights model that scores 72.2 % on the SWE‑bench Verified benchmark, placing it among the top performers for autonomous coding agents. Ars Technica reports that Mistral also delivered the accompanying Vibe CLI, enabling developers to interact with the model directly from the terminal and automate multi‑file changes. TechCrunch

New “RSL” licensing spec forces AI scrapers to pay for web content

The Really Simple Licensing (RSL) 1.0 specification, now an official standard, lets publishers embed machine‑readable rules that require AI crawlers to compensate for harvested content. The Register explains how RSL extends the classic robots.txt model, while The Verge notes backing from Yahoo, Ziff Davis and O’Reilly, signaling the first coordinated effort to monetize AI data extraction. Ars Technica Adweek

Artificial Intelligence Landscape

Major AI firms excluded from the Agentic AI Foundation standards group

The Agentic AI Foundation (AAIF), formed under the Linux Foundation, brings together AWS, OpenAI, Google, Microsoft and others to define interoperable AI‑agent standards, but Meta is notably absent. Computer World reports that Meta’s focus on a proprietary Avocado model explains its decision to sit out the consortium, raising concerns about future ecosystem fragmentation. Computer World

IEEE rolls out two new AI‑ethics certifications for professionals and products

IEEE’s CertifAIEd program now offers separate certifications to assess ethical compliance of AI systems and the competence of individuals overseeing them. IEEE Spectrum details the framework’s pillars—accountability, privacy, transparency and bias mitigation—and notes its alignment with the EU AI Act, positioning it as the sole internationally recognized AI‑ethics credential. The Guardian

Google pilots AI‑generated article summaries on Google News pages

Google is testing AI‑crafted “article overviews” on select publishers’ Google News listings, aiming to surface concise insights that help readers decide which stories to read. TechCrunch reports participation from outlets such as The Guardian and The Times of India, while the pilot also raises questions about editorial control and algorithmic bias. TechCrunch

OpenAI hires former Slack CEO as chief revenue officer to accelerate enterprise adoption

OpenAI announced the appointment of Denise Dresser, ex‑CEO of Slack, as its new chief revenue officer. The Verge explains that Dresser will spearhead global revenue strategy and help “put AI to work in day‑to‑day operations” for businesses, leveraging her experience launching AI‑enhanced collaboration tools at Slack. Harvard Business Review

Technology Trends Shaping 2025

Deloitte warns AI agents have failed to deliver on promised productivity gains

Deloitte’s 2025 Tech Trends report finds that autonomous AI agents have not achieved mainstream adoption, citing integration complexity and limited ROI as primary barriers. ZDNet highlights the consultancy’s call for clearer business cases and better alignment with existing workflows to unlock the technology’s potential. Zdnet

Model Context Protocol (MCP) aims to create a universal “internet for AI agents”

Anthropic‑originated MCP (Model Context Protocol) is gaining traction as a standard for AI agents to access tools and data across the web in a consistent format. The Verge describes MCP as a three‑letter key to a new AI‑centric internet layer, promising interoperability that could finally justify the massive AI‑investment spend. The Verge

State attorneys‑general issue warning letters to major AI developers over harmful outputs

Dozens of U.S. state attorneys‑general have collectively warned companies like OpenAI, Anthropic and Apple to implement safeguards against “sycophantic and delusional” AI outputs. Gizmodo reports the letter’s demand for risk‑mitigation measures, while the Financial Times and 9To5 Mac note that the same coalition is pressing for broader consumer‑protection regulations. Gizmodo The Verge Phoronix

Digital Transformation Momentum

AI drives record‑high dividend and buyback program at GE Vernova

GE Vernova’s latest earnings call credited artificial intelligence as a key driver of soaring demand for natural‑gas‑fired power equipment, enabling the company to raise its dividend, expand share buybacks and lift its 2028 earnings outlook to $52 billion. The Boston Globe emphasizes that AI‑enabled design and operations are reshaping the energy sector’s growth trajectory. bostonglobe.com