# News: August 21, 2026

> 12 curated links on AI agents, Python, security, Google Cloud, and engineering, selected by Médéric Hurier.

- Author: [Médéric Hurier (Fmind)](https://www.fmind.dev/)
- Published: 2026-08-21
- URL: https://www.fmind.dev/news/2026-08-21/

- [Ornith-1.5: From Self-Scaffolding to Self-Improvement](<https://ornith.ai/ornith_1_5.html>): Ornith releases Ornith-1.5, an open-source model family spanning 9B Dense to 397B MoE architectures that enables autonomous self-improvement by generating novel tasks, dynamic scaffolds, and reinforcement learning rollouts.
- [DeepSeek is back... and Silicon Valley is terrified](<https://www.youtube.com/watch?v=xBByvFrqmWU>): Fireship analyzes the rapid evolution of DeepSeek's open-weights frontier models and their disruptive competitive impact across Silicon Valley and closed-source AI labs.
- [Expanding Google Antigravity for enterprise customers](<https://cloud.google.com/blog/products/ai-machine-learning/expanding-google-antigravity-for-enterprise-customers>): Google announces the expansion of Google Antigravity to Gemini Enterprise subscriptions, introducing pooled token quotas, project-level spend thresholds, centralized audit logging, and IDE extensions for development teams.
- [Enterprise | Google Antigravity Docs](<https://antigravity.google/docs/enterprise>): Google publishes the enterprise administration documentation for Google Antigravity, detailing administrative spend controls, pooled quota management, audit logging, and workspace sandboxing policies.
- [Make your website agent ready with WebMCP](<https://www.youtube.com/watch?v=FARxSG_EY98>): Google Cloud Tech demonstrates how developers can implement WebMCP to expose typed tools directly within web pages so autonomous AI agents can interact with websites reliably and efficiently.
- [How to modernize legacy codebases using AI coding agents (Step by step guide)](<https://www.youtube.com/watch?v=RKvckuC6Bsk>): Google Cloud Tech delivers a step-by-step practical guide demonstrating how software engineers can safely refactor, test, and modernize legacy enterprise codebases using autonomous AI coding agents.
- [Slack is launching collaborative vibe-coding channels](<https://www.theverge.com/tech/982628/slack-code-vibe-coding-channels-launch>): The Verge reports on Slack's rollout of dedicated collaborative vibe-coding channels that enable software engineering teams and autonomous AI agents to build and iterate on code together in shared spaces.
- [Gemini Spark versus Hermes Agent versus OpenClaw: Who Wins and Why?](<https://hackernoon.com/gemini-spark-versus-hermes-agent-versus-openclaw-who-wins-and-why>): HackerNoon compares Gemini Spark, Hermes Agent, and OpenClaw across security architectures, pricing models, and execution capabilities to evaluate the optimal agent runtime for developers and enterprises.
- [AI 101: Hermes vs OpenClaw: Local AI Agents Compared](<https://www.turingpost.com/p/hermes>): Turing Post provides an in-depth architectural comparison of Hermes Agent and OpenClaw, examining memory management, self-improving skill execution, task scheduling, and local agent safety.
- [How to build AI agents with Gemini Enterprise & Workspace](<https://www.youtube.com/live/BFtkDiJYYHk>): Google Cloud Tech illustrates how enterprise administrators and developers can build and deploy autonomous AI agents by connecting Gemini Enterprise with Google Workspace APIs and data connectors.
- [Building And Structuring An AI Native Company](<https://www.youtube.com/watch?v=Z3JyAqh4ixg>): Y Combinator General Partner Tom Blomfield shares strategic research and architectural insights on structuring AI-native organizations around autonomous agent loops and continuous data feedback.
- [11 INSANE Use Cases for Grok Bot](<https://www.youtube.com/watch?v=5CSXUsljJ_E>): Matthew Berman highlights eleven practical use cases and workflow automations unlocked by xAI's Grok bot across web research, data parsing, and multi-step reasoning.