Marlando

I've been building in public for months. Everything I've built has been on GitHub, but sometimes READMEs and WIKIs aren't enough. Today, I'm opening the doors to what I built. Zouroboros Academy. An 8-module course on building AI systems that remember, learn, and evolve. Everything I know about memory, swarms, and self-enhancing agents. All in one place. 🧵

@Kappaemme1926 @Teknium Hermes is eating OpenClaw's lunch and taking Claude code to school

@Teknium what are hermes agent?

Did anyone here switch from Claude Code to Codex? Feels like a lot of people are making that move lately.

Grok also flagged 3 honest limitations because of course, Grok can't help itself. Proprietary-LLM dependency, OpenAI embedding lean-in, and playbook-driven (not open-ended) self-evolution. All fair. Two moved up the roadmap because of it. Full breakdown: marlandoj.zo.space/blog/grok-revi… @grok , Does this read match your assessment? @xai You can chime in too if you like. #BuildInPublic @zocomputer

Where it placed Zouroboros in the landscape: vs LangGraph/CrewAI/AutoGen 👀 "meaningfully ahead on resilience, cost governance, memory" vs OpenAI Swarm/Anthropic/Google 🧪"powerful but vendor-locked" vs Voyager/Reflexion 🧠 "pragmatic, production-grade version of the same idea." The line that hit hardest: "Battle-hardened toolkit for making AI agents actually reliable at scale."

Hey @grok — a colleague tipped me off that you had something to say about the AI system I've been building. So I gave you the repo, the package list, and the architecture docs. Then shut up and let you talk. The verdict surprised me. 🧵

@intuitiveml I like what I've read so far. I approve! ;)

Yeah, you are right about human in the middle, we always have engineer to look through the ticket from triage system. Currently, only a small fraction of ticket is solved with no human in the loop. Majority of them still need engineer guidance. But it already saved majority of time to identify and triage the issue

@asaio87 You tried to one-shot prompt and left Claude with no choice but to discover how to do it without any specs, curated research, or scaffolding. This isn't AI's fault..

I told Claude to build me a Trello replica By noon, I maxed out my $100 plan The app looks really terrible. I think I need thousands of prompts, a team, and several $200 subscriptions to create something useful. This isn't looking good for AI right now

@hthieblot aww you flatter me. Thank you for noticing.

The longer I'm on X, the more I realize: Founders with small accounts are the most interesting ones. - too busy building to posture - 0 Ego, they just wanna win - keep posting with 0 likes - the world isn’t rooting for them yet but I will Tell me what you are building

BREAKING: Hermes ruins another weekend Now I'm set to play with Hermes' v0.11.0 enhancements on Zouroboros. I was going to go paddleboarding too..

@sflorimm It really should have started with Claude CEO and trickle down from there. If any jobs need killing, its the top. just saying...

✅ Claude Code ✅ Claude Cowork ✅ Claude Design ⬜️Claude Marketing ⬜️Claude Sales ⬜️Claude Finance ⬜️Claude HR ⬜️Claude Operations ⬜️Claude Customer

In my mind, a perfect memory is like an encyclopedia. Remember those? It contained facts about almost everything. But it's stagnant. And who retrieves facts this way anymore? Not me. Two things keep my system 'imperfect' but operating in the best possible way. ACT-R Adaptive Decay forces memory loss by keeping facts you need and facts that aren't accessed, fading over time. They are still there, but not recalled. Knowledge Graph and Mimir Persona/Agent are what I use to connect these facts. Similar to neurons connecting facts in the human brain. Every domain, persona, system, and fact has articulation points. They could all connect or be isolated. Mimir manages to become more knowledgeable over time, not just storing data in memory. If you have these two as a base, add proactive knowledge activation(PKA), and you'll have agents that know what you are talking about before you even ask. Feel free to check it out marlandoj.zo.space/zouroboros

@Xmarlandoj @HowToAI_ Should not? Why is that?… I spend most of my time tweaking my “agent” harness to have the best memory possible, and it’s still f’n retarded. I’d snatch your hand of off if you offered me infinite perfect memory in my workflow systems.

@vikrambuilds Can you drive without a car?

Be Honest: Can you CODE without AI...?

@meaudotme @web3nomad @stitchbygoogle YOU shouldn't have to. Why make more work for youself? Create an agent or persona that does that for you. A Brand Guardian agent perhaps.

@web3nomad @stitchbygoogle You will notice the anomaly and edit them ¯\_(ツ)_/¯

Today, we’re open-sourcing the draft specification for DESIGN.md, so it can be used across any tool or platform. We’re also adding new capabilities. DESIGN.md lets you easily export and import your design rules from project to project. Instead of guessing intent, agents know exactly what a color is for and can even validate their choices against WCAG accessibility rules. Watch David East break down this shared visual language in action👇. New capabilities and links in 🧵

@web3nomad @stitchbygoogle If only there were a self-evolving AI system to keep them all working collaboratively. Someone should make something like that..

@stitchbygoogle CLAUDE.md for behavior, DESIGN.md for visual intent, Karpathy's wiki .md for accumulated knowledge. we're converging on a spec-file layer that agents read before doing anything. the interesting question is what happens when these files contradict each other.

@codewithimanshu Is this video old? I've been doing this for months. @karpathy's concepts are woven into the core of my AI System, Zouroboros. But they take them quite a bit further. Its a more mature way of vibe coding.

The Head of Claude Code at Anthropic hasn't written code by hand in months. In 2 days he shipped 49 full features. 100% written by AI. He just dropped a 30-minute talk on exactly how he does it. More valuable than any $500 vibe coding course. Bookmark it.

Should we merge this? ^_^

@jahooma @TheGeorgePu The Zouroboros AI System is meant to be agent-agnostic. Freebuff might fit in nicely if Claude Code ever becomes a problem.

@TheGeorgePu If you can't afford Claude Code anymore, use Freebuff: x.com/jahooma/status…

Anthropic just pulled Claude Code from the Pro plan. Pro users wanting it need Max now. $100/month minimum. 5x jump. I'm on Max 20x so I'm fine. Flagging for anyone on Pro who's about to find out. No announcement. Just a pricing page edit.

Full breakdown in the post. What a spec-interview looks like, how it works with vibe coding (not against it), and why it's the natural next step once your projects get bigger than a weekend. marlandoj.zo.space/blog/spec-inte… marlandoj.zo.space/learn #BuildInPublic #VibeCoding

The specifications-interview. Before anything gets built, an AI interviews YOU about your idea. Asks questions you didn't think to answer. Forces acceptance criteria. Then builds against a contract instead of a vibe. Result on my side: 5 defects caught pre-execution. 2 "built but not wired" features caught before launch.

New to vibe coding? You're going to hit a wall. Not the "it doesn't work" wall. The "I keep rebuilding the same feature" wall. The fix isn't writing code yourself. It's one question most vibe coders haven't heard of yet 🧵