룟 Ryot

1,036 Twitter followers

룟 한국어 계정. 피델리온과 BTC, SOL의 부흥을 위하여 Korean alt account of @gmryot | Emperor @fidelionnft | @rugradiokorea co-host

Posts

I sincerely thank everyone who joined us for the 1st OpenClaw & Web3 event today! We organized this event with the somewhat challenging topic of "Giving Wallets to Agents," but I didn't expect so many people to show interest and attend. Haha. With over 200 people registering in total, I plan to organize a follow-up event next time :) This event was not just a simple technical demo; the intention was for people who actually operate agents, build communities, and contemplate the ecosystem to gather and share honest stories. "Giving Wallets to Agents" is ultimately a matter of trust. Trust in the code, trust in the system, and trust in the people who created that system. The most meaningful aspect was being able to contemplate the future of Agents + Web3 together. I truly felt the interest the @openclaw ecosystem is receiving in Korea, and I plan to continue organizing events like this in the future. Next time, I will return with deeper topics and more practical content. I would like to express my sincere gratitude to everyone who joined us for the 1st OpenClaw & Web3 event today! Once again, I would like to extend my sincere gratitude to everyone who came, to the panelists @Lo_gan__, @Im_zo_sol, and @selarious25, and to the sponsors @MeteoraAG and @cysic_xyz! For those who couldn't make it this time, I hope to see you next time!

[The era of large agents can finally be established through crypto.] When everyone uses agents, people will naturally entrust a portion of their assets to agents. And those assets will be realized through stablecoins, not through banks. This is the true form of mass adoption, the very thing Web3 has been calling for. This is already in progress, with countless agents actively operating with their own wallets. And you, tinkering with OpenClo right now, are a pioneer in this process. This community is for those who trade through OpenClo, seek to generate income through DeFi, or create Web3 services. While it sounds grand, it's really just a community for people to share their thoughts and ideas. Come join us if you're a talented individual. We'd love to share various trending news with you.

룟 Ryot

02-27

Thread

[Claude Code is over?] These days, with the proliferation of Claude Code teams, Notion AI, and mobile-based agents, there seem to be a lot of aggressive posts claiming "OpenClaude is over." Having run OpenClaude for about a month and then upgrading it to the max, I can honestly say that these new products are structurally incapable of effectively replacing OpenClaude. My current agent architecture is parallel. But it's not just parallel; even LLM is different. - One agent acts as an overseer based on Sonnet 4.5. During development, we upgrade to Opus 4.6, with Minimax 2.5 as the fallback. They're responsible for overall workflow management and decision-making. Their goal is to ensure uninterrupted work flow. They actively run Claude Code. - Another agent is dedicated to code reviews and crypto trading based on Codex. The fallback is Grok. They conduct code reviews before the first agent activates CC. They focus on logic and risk. The final friend is a Gemini-based content agent. We use grok for research. All of our X content and YouTube ideas come from here. The core of this architecture isn't a specific model, but the separation of roles and parallel operation between models. Each company's SaaS-based agents essentially operate within a single ecosystem. Naturally, there's no option to choose LLM. Claude can only use Claude, and Grok can only use Grok. Notion AI and Perplexity are optimized only within their own platforms. The Claude code team is strong in code collaboration. However, it's impossible to create a structure where Gemini-based content production runs in parallel, research is supplemented with Grok, and the entire process is coordinated by Sonnet Overseer. This is structurally blocked because it's a service-based system. This structure doesn't allow agents to leverage each other's strengths. In contrast, OpenClaud separates models, SaaS, and mobile interfaces into layers. Models are interchangeable, and fallback systems can be freely designed. Distributed operation by role is possible. The update speed for this also varies, with patches released nearly five times a week. It's a misconception to attribute this to imperfect software. We live in a world where five major LLM companies release ten new products a week. This means not just simple feature additions, but rapid iterations of experimentation and modifications on newly emerging language models and products. This is an example of why well-built, well-maintained open source evolves faster than services tied to a specific company. SaaS agents are complete tools. Openclaw is infrastructure. Complete tools are convenient, but difficult to restructure. Infrastructure is composable. Parallel operation of multiple language layers is difficult to implement within the services of a single AI company. It's the difference between "using" an agent and "operating" an agent. This is why I constantly tell my friends, "Whether you like or dislike Openclaw is not a matter of functionality, but rather a matter of skill."

OPUS

0.44%

-- END --