AI Agents are great... but wouldn't it be nice if we had a workspace where agents got better the more we used them, instead of burning through limits & losing quality over time? The solution is simpler than you think: if we pair the core elements of agentic AI & with automations, we get the beginning of this self-improving system. I have been exploring Codex by OpenAI (https://openai.com/codex/ ) & today I'll show you how to set up your own self-improving agentic AI workspace β¨ In today's video, we'll go through: 1) The Core Architecture aka The Building Blocks 2) Automations and how to make them self-evolving 3) Expanding your workflows (local, worktrees, and cloud operations) Automations are only useful if we build them on a solid foundation. In agentic systems, that's the architecture, composed of 4 core building blocks: 1) Agents.md - The Constitution 2) Skills - Standard Operating Procedures 3) Plugins + MCP - Extending The Reach 4) Memories - The Adaptive Layer But these are just the foundations of agentic AI systems. Automations help turn Codex from a simple coding assistant into a self-evolving agent workspace. By pairing automations with the core features, we're able to provide an autonomous, self-evolving system that gets better the more you use it. Standalone & thread (heartbeat) automations help you automate tedious tasks, but also introduce self-evolution by running regular feedback loops like auditing your agents.md & skill.md files for improvement opportunities. Once you've got your workspace set up, you can then take it on the go with you, using cloud execution. The cloud system containerizes your space, a virtual computer, that enables you to start a chat at your desktop before taking it to your phone on the go. Together, these features are the beginning of a truly self-evolving agentic workspace, one I'm excited to explore further. In particular, I've really been enjoying ChatGPT 5.5, it's done a great job analyzing my LLM Wiki & helping me continuing building apps more effectively. I hope you enjoy! β¨ Thanks to OpenAI for sponsoring today's video. There is a new Pro plan for Codex (and ChatGPT) that gives you 5x the Plus usage (quota). For a limited time, until May 31, that usage is doubled, giving you 10x the Codex usage. Now is a great time to get your system set up and test it out with extended usage. https://openai.com/codex/ P.S. I greatly appreciate any feedback, please let me know what you think π π Sign up for my [free newsletter: Recalibrating](https://paragraph.xyz/@wanderloots.eth?referrer=wanderloots.eth) π‘ Wander my Digital Garden https://wanderloots.xyz Timestamps: 00:00 Intro To Agentic AI Workspaces 00:48 Today's Outline & Goals 01:09 Core Agentic AI Architecture 01:37 Agents.md - The Constitution 02:17 Skills - Standard Operating Procedures 02:55 Plugins + MCP - Extending The Reach 03:19 Memories - The Adaptive Layer 04:00 Using The Core Features (Manually) 04:30 Skills Demo 05:52 Local vs Global Skills 06:15 Automations 06:56 Automations Demo 08:19 Heartbeat Automations 10:50 Self-Improving Automation Examples 11:50 Parallel Workflows & Subagents: Theory 12:15 Parallel Workflows & Subagents: Practice 13:17 Subagents (Within Threads) 14:38 Cloud Usage: Coding On The Go 16:47 Global Workspace Concept 17:29 Recap & Concluding Thoughts WATCH NEXT π± Obsidian & Digital Gardens https://www.youtube.com/playlist?list=PLWhMzDKA7vJ7p50vW-oeZgKR2aDReZFW6 π§ Recalibrating (Main Storyline) https://www.youtube.com/playlist?list=PLWhMzDKA7vJ4SLS2e6KiPmu9mku81B__Y LINKS (MY WORLD) π§ [Recalibrating Newsletter Home](https://paragraph.xyz/@wanderloots.eth?referrer=wanderloots.eth) ποΈTo start reading from the beginning: [Recalibrating Newsletter Entry 1: What Recalibrating Means To Me](https://wanderloots.substack.com/p/1-what-recalibrating-means-to-me) π My [Website](https://wanderloots.com/) πΈ My [Print Shop](https://wanderloots.darkroom.com/) β¨ SOCIALS π£ [Farcaster](https://warpcast.com/wanderloots.eth) πΈ [Instagram](https://www.instagram.com/_wanderloots/) π° [Flipboard](https://flipboard.com/@_wanderloots) π [Pinterest](https://www.pinterest.ca/wanderloots/) π¦ [X (Twitter)](https://twitter.com/_wanderloots) π§΅ [Threads](https://www.threads.net/@_wanderloots) π€ [Reddit](https://www.reddit.com/user/_wanderloots) MY FAVOURITE TOOLS π΄ π€― The [Waking Up App](https://dynamic.wakingup.com/guestpass/SC4914439) (use this link for a 30 day free trial) π [Obsidian](https://obsidian.md/) (decentralized note-taking) πΉ [Adobe Suite](https://prf.hn/l/lQ9DwpA) (general creativity) β€οΈ [Welltory](https://app.welltory.com/payments/plans/main/?coupon=wanderloots) (Health Tracker) EQUIPMENT USED 6. Camera [Sony A7iii](https://amzn.to/3seSHv6) 7. Lens [Sony F2 28 mm](https://amzn.to/3TiWCT2) 8. Tripod [K&F Concept](https://amzn.to/3soCKCP) 9. Main Lighting Neewer 660 PRO RGB: https://amzn.to/3CEcU2V

Obsidian Smart Plugin Workflow π New Smart Connections + Context AI
7.6K views

How To Build LLM Wiki In Obsidian? π§ A Memory Layer For Any Agentic AI
33.5K views

Why LLM Wiki? π§ Future Of Knowledge For Agentic AI & Humans
82.7K views

Academic Search Engine + Agentic AI π Consensus & Zotero Deep Research
2.2K views

How To Stay Ahead In An AI World? π‘ Curating Knowledge With Recall Update
4.3K views

Vibe Design First THEN Build Apps π¨ Full Google Stitch Tutorial + MCP Agentic AI
29.6K views