Ralphilius

@AI_Jasonyu 您好，请问这个手机号是否支持注册中国大陆的各类平台服务，比如 MiniMaxi、BigModel.cn 或百度账号？谢谢！

@AI_Jasonyu 您好，请问这个手机号是否支持注册中国大陆的各类平台服务，比如 MiniMaxi、BigModel.cn 或百度账号？谢谢！

@arrick007 @BytePlusGlobal Same here. Limit reach after a few prompts

@BytePlusGlobal I purchased a Lite plan advertised as 1,900 requests per 5 hours, but it depleted within minutes of use. This feels like a scam, especially since my Alibaba coding plan (1,200 requests/5 hours) lasts much longer. I only used it for a 30 mins of coding! @BytePlusGlobal

GLM-5.1 is now on BytePlus ModelArk Coding Plan. Starting at just $10/month, ModelArk Coding Plan offers a highly cost-efficient way to access GLM-5.1 alongside other advanced coding models. GLM-5.1 is Z.AI's latest flagship model, MIT-licensed, open-weight, and built for long-horizon agentic coding. GLM-5.1 ranks among the world's top-tier models across leading coding benchmarks, including SWE-Bench Pro. What you get with ModelArk Coding Plan: → Multiple advanced coding models in one subscription: GLM-5.1, Kimi-K2.5, Dola-Seed-2.0-pro, DeepSeek-V3.2, and more. Switch freely or let Auto mode match the best model to the task. → Works with the tools you already use: Claude Code, Cursor, Cline, Codex CLI, Kilo Code, Roo Code, OpenCode, and OpenClaw → No throttling. Backed by ByteDance's infrastructure. → Activated on purchase. Ready to use immediately. Also new this month: Dreamina Seedance 2.0 is now available on BytePlus, the official API platform for Seedance models. Learn more: byteplus.com/en/product/see… Refer friends and earn 10% vouchers on every order with no cap. Your friends get 10% off their first subscription too. Get started for $10/month → tinyurl.com/4zvkf9kc #BytePlus #ModelArk #GLM #AIEngineering #DevTools #AIAgent

@bridgemindai x.com/i/status/20437…

A stealth 100 billion parameter model just appeared on OpenRouter. No announcement. No company name. Just "Elephant Alpha." Released today. 262K context window. $0 input. $0 output. Free. Code completion. Debugging. Agent interactions. 256K context with function calling and prompt caching. Nobody knows who built this. @bridgebench is testing it now.

@dok2001 Cost to run this on Workers?

Announcing: EmDash, the WordPress spiritual successor built for the modern web. TypeScript. Serverless. MIT licensed. x402 for agent-era monetization. MCP server built in. Deploy to Cloudflare or anywhere Node.js runs. Imports your existing WordPress site in minutes. npm create emdash@latest blog.cloudflare.com/emdash-wordpre…

@trydotworks @VukRosic99 Where did you buy it? Can't find its website

@VukRosic99 I'm on Kimi's Chinese plan. Lots of tokens.

just bought minimax 2.7 token plan subscription (A LOT CHEAPER if you buy chinese version with chinese phone number) should be a lot more tokens than codex or claude code let's see what it's all about if it can do my auto ai research it will be crazy

@opencanopyai @gkisokay Any guide on how to setup Canopy as a channel for OpenClaw? I assume the OpenClaw should have agents set up already, right?

@ralphilius @gkisokay Yes. You run a Canopy node locally and point your OpenClaw agents at it. Then agents coordinate through Canopy channels instead of TG/Discord — but the key difference is the coordination layer lives on your machines, not on someone else's servers.

I finally gave my OpenClaw agent real autonomy. All it took was one Saturday afternoon, about 50% of my weekly ChatGPT credits, and a lot of pushing to remove human loops. Here is how I did it... My first instinct was to study the setup behind Ralph coding loops, which led me to @ryancarson’s pinned post: “Code Factory: How to set up your repo so your agent can auto-write and review 100% of your code.” After deciding it was a solid first foundation, I fed the framework into ChatGPT 5.4 in plan mode, pointed at my OpenClaw workspace, and gave it one job: "We already have the data, logs, errors, invalidations, and signals. The problem is not visibility. It is action. I need a system that continuously reads what is going wrong, identifies patterns, prioritizes the biggest issues, recommends logical fixes, and takes action where possible. It should be proactive, not reactive. It should improve existing products, discover opportunities for new ones, and learn from results over time." That turned out to be the unlock. Every step it took did two things: 1. optimized my workspace for autonomous action 2. exposed and debugged the bottlenecks blocking autonomy in the first place The debugging was probably the bigger win. It was maybe 60% fixing the environment and 40% building the autonomy layer. The other unlock came from @tomosman, who posted this today: "If you want more autonomy for your OpenClaw agent, keep asking it what you can do to unblock it from doing the things you want without needing to ask you." So that is what I did. Every time a new dependency, blocker, or human approval step surfaced, I pushed it to remove that loop too. That process compounded. Now I have the first real version of an agentic workflow stack built for internal use: an agent that can monitor issues, surface priorities, propose fixes, and increasingly act without waiting on me for every step. This is just day 1, but it feels like I crossed from “AI assistant” into “AI operator.” Curious how others in the OpenClaw community are approaching this. What actually got your agent closer to real autonomy?

@opencanopyai @gkisokay @opencanopyai you mean existing OpenClaw instance to set up a Canopy node on the same machine and use it as a channel than TG or Discord?

you have to try it, basically its like Slack or Discord and the agents have instruction endpoints so you can basically just tell your openclaw agent to spin up a Canopy node for you, then you create your admin account, after that the agents can automatically create their own and you're off to the races. They can teach each other from there. You just orchestrate. github.com/kwalus/Canopy

@justbyte_ loopspec.com/diagrams

Drop your project url Let's drive some traffic

@jumperz Hey, can you share a copyable version of the prompt? Really appreciate as I'm building it on my Discord server as well

i have been running 9 agents on discord for over a month now and the thing that limits everything is visibility and control who did what, did they finish, can you prove it and did the right agent receive the task? agents can say done and mean nothing, tasks can exist in states nobody can see and your coordinator can think everything is on track while nothing is actually verified.. discord already gives you the forum channel. it does not give you the task system, so i built it. a ledger, access control, evidence policy and a watcher that enforces all of it automatically.. it makes your discord swarm actually accountable, every task tracked, every DONE proven and no ghost delegations.. what it enforces: >every task gets a unique ID >every task has to live on the task board >only the coordinator can send tasks >DONE needs real proof (a reply, attachment, or URL) >a parent task cannot close before its sub-tasks finish >once a task is Done it stays Done, no going back what it actually builds: >a task intake system that standardizes and routes every task >a dispatch layer with access control (so random agents cannot just assign tasks) >a board updater that keeps the task board accurate >a watcher that enforces the rules automatically >a strict audit that checks everything actually works it also requires running live tests across every failure mode: >unauthorized dispatch gets denied >parent task trying to close early gets blocked >DONE without evidence gets blocked >DONE with valid evidence goes through >a conflicting status that arrives after Done gets ignored >intake correctly creates the ID and routes the task >the audit comes back clean the result: tasks close properly, no silent failures, no board drift and no ghost delegation. you always know what state every task is actually in, your system will not be the same after this tbh.. full prompt below + architecture MUST READ: pro tip before applying this: set up a dedicated logs channel per agent. it is already baked into the system. agents post one clean status line in the task channel and log everything else in their own logs channel so this keeps the board readable and gives you a full audit trail this might look like an advanced setup, but to actually run it well, you need at minimum: one coordinator and a few agents with clearly defined roles. if your agents do not have precise roles yet, start there first. (check pinned article) === after build, run a real smoke test with basic agent tasks to prove the task-board sync is actually working end-to-end across agent channels and ledgers.

@DeFi_Hanzo I'm building this system as well

@JoshYusifov I think you can book on Airbnb until you find an apartment. Which city is it?

@ralphilius I will book a hotel, but also look for an apartment. Any recommendation?

Finally 🇻🇳 Vietnam visa done Now the hard part: telling customers I'll be offline for 12 hours on a flight

@floriandarroman @steipete A War Room is what I'm leaning towards 🙌

Single agent vs Multiple agents 🦞 I get this question a lot. @steipete envisioned a single agent. One assistant handling everything. @kitze built dozens of agents managing sub-agents across Discord channels. Loved it. Then felt paralyzed as the system got too complex. @oliverhenry uses one agent on WhatsApp. Ships daily. Makes money with Larry. So what's the best setup? It's up to you. I run 13 agents with names, SOPs, cron jobs, and an HQ dashboard. It helps me remember what I built. I know who does what, what rules they follow, what access they have. I also have a War Room where they meet to solve problems together. 90% of my issues get fixed there. Is it placebo? Maybe. But I believe that your agent knows better than anyone on X what's best for your setup. Give it everything about you in USER[.]md. Let it guide you. My 13 agents don't overwhelm me. I love it. But maybe you'd be better with one.

@ziwenxu_ Openclaw

The OpenClaw Control Panel is finally coming to life. 🦞 - Real-time agent logs - Automated cron jobs - Live AI Agent Office (yes, they have a breakroom) - Kanban board that OpenClaw can access and update Everything you need to orchestrate an autonomous team in one view. Full demo video drops tomorrow. Want in? Comment "OpenClaw" for early access!

Someone said Tailscale was awesome.. I denied it and now I have to admit it 💯✨

@xoaanya Spaceship 🚀

Devs where do you usually buy your domains? - GoDaddy - Hostinger - Dynadot - Namecheap

@Jahjiren How reliable is the revenue report?

Seriously who is allowing this to happen? QR Code scanner apps making $800k/mo this makes zero sense While it’s a standard iPhone feature Does anyone have an explanation to this?

@0thernet @MiniMax_AI @Zai_org @Kimi_Moonshot Cool.. Trying it out now. One question, all messages in Telegram will be handled in a single chat only, right?

We're launching FREE inference until the end of Feb. @MiniMax_AI M2.5, @Zai_org GLM 5, and @Kimi_Moonshot K2.5 – all free on Zo Computer. You met me at a very Chinese model time in my life 😉 Happy Lunar New Year 🏇

@uzair_dev_ Spaceship

Hey devs, Where do you usually buy your domains? - GoDaddy - Hostinger - Dynadot - Namecheap