TechLatest

Posted on Jun 16 • Originally published at Medium on Jun 12

TechLatest AI & Tech Weekly #20

#newsandupdates #weeklynews #technologynews #technews

Welcome to this week’s edition of TechLatest AI & Tech Weekly 👋

Here’s a curated roundup of our latest blogs, notable product launches, and the most interesting AI & ML updates from June 08–June 14, 2026.

AI/ML News Roundup: June 08–June 14, 2026

Key highlights from this week’s AI developments include frontier model advancements with agentic capabilities, massive funding rounds reshaping valuations, and practical product launches for developers and enterprises. These updates emphasize autonomous agents, infrastructure scaling, and open-weight benchmarks relevant to builders and researchers.

Open-Source AI, AI Agents & Developer Releases

xAI Launches Grok Build Plugin Marketplace

xAI introduced the Grok Build Plugin Marketplace, expanding Grok’s developer ecosystem. Launch partners include MongoDB, Vercel, Sentry, Chrome DevTools, Cloudflare, and Superpowers. The marketplace allows developers to connect external tools and services directly into Grok workflows. The release aims to make Grok more useful for software development and automation tasks. Source

Nous Research Ships Hermes Agent Profile Builder

Nous Research released the Hermes Agent Profile Builder, a unified dashboard for creating and managing AI agents. The platform combines identity configuration, memory, skills, MCP servers, and agent profiles in a single workflow. It is designed to simplify the setup of persistent, personalized AI agents. The release strengthens the broader Hermes ecosystem for agent development. Source

Databricks Open-Sources OmniGenT

Databricks released OmniGenT, an open-source meta-harness for orchestrating AI agents across multiple ecosystems. The framework enables teams to compose, govern, and share agents built with tools such as Claude Code, Codex, and Pi. OmniGenT aims to standardize agent management and interoperability. The project focuses on enterprise-scale multi-agent development and deployment. Source

Anthropic Disables Claude Fable 5 and Mythos 5

Anthropic announced the shutdown of Claude Fable 5 and Mythos 5 following a reported U.S. government order. The company stated that access to both models would be discontinued while complying with regulatory requirements. The move has sparked discussions around AI governance and model availability. Anthropic has not indicated when or if the models will return. Source

Google Releases Gemini SQL2

Google announced Gemini SQL2, powered by Gemini 3.1 Pro, for advanced text-to-SQL generation. The system reportedly achieved 80.04 on the BIRD Single-Model Leaderboard, a widely used benchmark for database query generation. Gemini SQL2 focuses on converting natural language requests into accurate SQL queries. The release targets enterprise analytics, data engineering, and business intelligence use cases. Source

Moonshot AI Releases Kimi K2.7 Code

Moonshot AI introduced Kimi K2.7 Code, an updated coding-focused model in the Kimi family. The model reportedly scored 21.8 on Kimi Code Bench v2, improving over the previous K2.6 release. K2.7 Code is optimized for programming assistance, code generation, and developer workflows. The update continues Moonshot’s push into AI-powered software engineering tools. Source

Cohere Releases North Mini Code

Cohere launched North Mini Code, a 30B open-weight Mixture-of-Experts model with only 3B active parameters per token. Designed specifically for agentic coding and software engineering tasks, it supports a 256K context window and runs on a single H100 GPU. The model is available under Apache 2.0 and supports tool use and reasoning workflows. Cohere positions it as an efficient coding model for self-hosted deployments. Source

Zyphra Releases Zamba2-VL

Zyphra unveiled Zamba2-VL, a family of vision-language models combining Mamba-2 and Transformer architectures. The hybrid design significantly reduces time-to-first-token while maintaining strong multimodal performance. The models support image understanding and text generation workloads with improved responsiveness. Zyphra highlights efficiency gains as a key advantage over conventional Transformer-only systems. Source

Hugging Face & Open-Source Ecosystem

Hugging Face Serge launched on June 12, 2026, as an open-source (Apache-2.0), GitHub-native AI code review system that reviews pull requests using OpenAI-compatible models and repository-owned review policies, available via GitHub Action, GitHub App webhook, and a staged web app mode.
Cohere open-sourced North Mini Code , a specialized code model with 30B total parameters (3B active), announced in the June 11, 2026 daily AI summary.
Google’s DiffusionGemma model is now open-source and available for free use on NVIDIA platforms , released in the June 11, 2026 daily roundup, with text generation speeds over 500 tokens per second.
Apache Burr , an Apache-licensed open-source framework for building reliable AI agents with stateful workflows and observability, was released in this period and is aimed at helping developers create more robust AI systems.

Frontier Model Advancements & Agentic Capabilities

Anthropic released Claude Fable 5 on June 9, 2026 , a publicly available, safeguarded Mythos-class frontier model with enhanced safety features, performing exceptionally well on long-term, complex software engineering, knowledge work, and research tasks.
Claude Fable 5 was jailbroken on June 10, 2026 , and on June 12, 2026 , the U.S. Commerce Secretary issued an export control directive forcing Anthropic to suspend global access to both Fable 5 and Mythos 5 for any foreign national, anywhere — the first time a frontier model was pulled by government order.
Claude Mythos 5 was launched alongside Fable 5 on June 9, 2026, for restricted trusted-access cybersecurity and research use, with both models priced at $10/M input tokens and $50/M output tokens.
Claude Design introduced a complete Agent runtime environment built on browsers and virtual machines, with 45 tools and 24 built-in skills , operating through an Agentic Loop , giving developers a powerful framework for building and deploying AI agents.
MetaMask published an Agent Wallet on June 8, 2026 , letting AI agents execute on-chain trades across EVM chains and DeFi primitives under mandatory security checks, with early access opening that day.

Practical Product Launches for Developers & Enterprises

Contentstack launched its Agentic Experience Platform (AXP) with Agent OS and Agent Accelerator on June 9, 2026 , announcing general availability of Agent OS — an autonomous agent layer spanning content, data, and real-time personalization — to move enterprises from AI experimentation to operational impact.
CircleCI released an MCP Server on June 12, 2026 , connecting AI coding assistants (Cursor, Claude Code, Windsurf, VS Code, Amazon Q Developer, Kiro) to CircleCI pipeline, build, log, test, and workflow data via the Model Context Protocol.
Descope MCP Server was announced on June 8, 2026, as part of Descope’s AI Launch Week, a hosted remote MCP server letting AI assistants inspect and manage Descope identity projects with read-first controls and human-approved write elevation.
Circle published a June 12, 2026 use-case guide for Circle MCP , a connector that links Circle community data and actions into external AI assistants through the Model Context Protocol, available on Business plans and above.
Dayflow became a context-enhancing tool for Mac AI in this period, recording minute-by-minute Mac device behavior to provide complete context to AI; it is open-source and uses the Gemini 3.5 Flash Lite API for free analysis.

Governance, Ethics & Regulation

London Tech Week 2026 (June 9–12, 2026) included policy announcements across the AI Adoption Summit and the UK’s AI Hardware Plan , announced with a £1.1 billion funding package , including a £750 million national AI supercomputer at the University of Edinburgh (operational by 2030) and a £150 million fund for UK hardware companies.
The AI Hardware Plan created a £150 million fund through the British Business Bank, £120 million for a new AI Hardware Innovation Programme (including at least £20 million for ARIA’s Scaling Inference Lab), and £80 million for skills (bursaries, Centre for Doctoral Training in Chip Design, PhD support).
The U.S. White House EO on AI innovation and security, along with a bipartisan Great American AI Act discussion draft, converged in mid-June 2026, establishing federal AI governance with provisions on frontier-system transparency, independent verification, and AI whistleblower protections.

Infrastructure & Hardware

Meta signed a deal with Reliance to build a 168-megawatt AI data center in India , announced on June 10, 2026 , marking Meta’s first such partnership in the country.
The UK’s AI Hardware Plan (announced Monday, but policy coverage active during June 9–12 Tech Week) includes a national AI supercomputer at the University of Edinburgh with UK-designed chips , intended as a crucial part of the AI Research Resource (AIRR) and operational by 2030.
The Model Context Protocol (MCP) adoption continues to expand, with CircleCI MCP Server , Descope MCP Server , and Circle MCP all launched or documented in this period, enabling AI assistants to securely connect to CI, identity, and community data.

Blogs We Published This Week

Git for Agent Memory: Why You Should Treat Hermes Skills Like Code

This article explains why Hermes Skills should be managed like software projects rather than static prompts. It explores version control, collaboration workflows, skill evolution, rollback strategies, and Git-based memory management. You’ll learn how treating agent skills as code improves reliability, maintainability, and long-term learning. A practical guide for anyone building self-improving AI agents.

Git for Agent Memory: Why You Should Treat Hermes Skills Like Code

OpenClaw or Hermes? Choosing the Right AI Agent Stack in 2026

A detailed comparison between two of the most talked-about AI agent frameworks in 2026. The article covers architecture, memory systems, skills, deployment models, gateways, migration paths, and real-world use cases. It helps developers understand where Hermes excels and when OpenClaw is the better choice. Ideal for teams evaluating their long-term AI agent strategy.

OpenClaw or Hermes? Choosing the Right AI Agent Stack in 2026

Build an ML Model That Actually Ships: A 6-Step Visual Walkthrough

Most machine learning tutorials stop at training a model — this guide focuses on getting models into production. Through a simple visual framework, it walks readers through problem definition, data preparation, model training, evaluation, deployment, and monitoring. The article emphasizes practical engineering decisions that turn experiments into real products. Great for both beginners and practitioners.

Build an ML Model That Actually Ships: A 6-Step Visual Walkthrough

Anthropic Cybersecurity Skills — Full Tutorial

This comprehensive tutorial explores Anthropic’s Cybersecurity Skills framework and how it enhances AI-driven security operations. It covers structured incident analysis, threat intelligence workflows, MITRE ATT&CK mapping, NIST alignment, and repeatable security playbooks. Readers learn how to equip AI agents with professional-grade cybersecurity capabilities. A valuable resource for security teams and AI builders.

Anthropic Cybersecurity Skills — Full Tutorial

OpenClaw Agent Masterclass — Full Tutorial

A complete hands-on guide to installing, configuring, and extending OpenClaw. The tutorial explains gateways, skills, messaging integrations, deployment options, and automation workflows. It helps readers move from basic setup to building production-ready AI assistants. Perfect for anyone looking to self-host and customize their own AI agent ecosystem.

OpenClaw Agent Masterclass — Full Tutorial

Hermes Agent Masterclass — Full Tutorial

An end-to-end introduction to Hermes Agent and its unique approach to memory, skills, identity, and self-improvement. The tutorial covers installation, core architecture, the Curator system, memory management, and skill creation. Readers will understand how Hermes learns over time and evolves through experience. A must-read for developers interested in adaptive AI agents.

Hermes Agent Masterclass — Full Tutorial

TL;DR — TechLatest AI & Tech Weekly #20

✅ Google released DiffusionGemma and Gemini SQL2, pushing faster text generation and stronger text-to-SQL capabilities.

✅ Cohere open-sourced North Mini Code, a 30B MoE coding model optimized for agentic software development.

✅ Moonshot AI expanded the Kimi ecosystem with Kimi K2.7 Code and Kimi Work, bringing improved coding performance and large-scale multi-agent workflows.

✅ xAI launched the Grok Build Plugin Marketplace, enabling integrations with tools like MongoDB, Vercel, Cloudflare, Sentry, and Chrome DevTools.

✅ Databricks open-sourced OmniGenT, a framework for orchestrating and governing AI agents across multiple ecosystems.

✅ MCP adoption accelerated with new integrations from CircleCI, Descope, Circle, and other developer platforms.

✅ Open-source AI continued gaining momentum with releases from Google, Cohere, Hugging Face, and the broader developer community.

✅ AI infrastructure investments and government policy initiatives highlighted the growing importance of compute, hardware, and AI governance.

✅ This week’s biggest themes: Agentic AI, Coding Models, MCP Ecosystem Growth, Open-Source Innovation, and Enterprise AI Deployment.

The AI industry is rapidly moving from standalone models toward interconnected agents, tools, and production-ready AI systems.