JoePro

MiniMax M3 is an AI model family built around a new Sparse Attention architecture. Exact parameter counts and benchmark data still haven’t been fully disclosed. Its predecessor, MiniMax M2, used a 230B-parameter design while activating only about 10B parameters per token during inference, trading brute force for efficiency and lower compute overhead.

@Anas_founder Cursor Ultra and Codex would be my choice. If only one, might be one or the other. I'm using Windsurf ($200), Codex (Teams) and lowered my Anthropic (Pro).

Developers, which one are you choosing for coding forever?

@TTrimoreau wsl Openclaw followed by Hermes.

Which command do you use most in the terminal?

@mark_k Using Claude Code is an agent task now. Same with Codex and all other AI harnessed tools.

I don't know anyone who's actually using Claude Code. Must be a normie thing? I guess it's popular with frontend monkeys. Anyway I think this too will pass.

@chaosengineerr I’m building a Karpathy style auto research engine that watches news.joepro.ai/news, ranks real signal across providers, turns it into decisions (ship / test / ignore) with permanent cryptographic receipts. Estimated life 200+ years. Genesis: arweave.net/r5XXxSb6BFYjki…

What’s the coolest thing you’re building with AI right now?

@quxiaoyin What providers are you running in Hermes? No need to rip out OpenClaw. Run both. Solid breakdown, just 2 fixes: /goal already shipped, and Nous was an open source AI lab first, then used crypto coordination.

OpenClaw is out, right? I personally stopped using OpenClaw. My current setup: Hermes Agent, Codex, and Claude Code. Claude Code: Slow as hell, super annoying. But I already paid for it, so might as well use it. Codex: My trusted lieutenant for anything requiring serious thinking. Reliable execution. Hermes Agent: Everything else. Blog posts, competitor research, news monitoring, code reviews. The grunt work that needs speed. Why I love Hermes: 1. Speed: Shows thinking process in real-time. When you interrupt with a new task, it instantly evaluates if it's related to current work or needs parallel processing. Claude just queues everything. 2. Self-improvement: If you give it the same task repeatedly, it creates a reusable skill. Like muscle memory for AI. Do something enough times, next time it just executes the pattern without thinking. They just launched a Kanban board for tracking progress, and apparently goal-based features are coming soon. Fast iteration cycle. The backstory: Hermes was originally a crypto company funded by crypto VCs. Started with open-source model training that nobody used. Then OpenClaw blew up in Feb/March, and they immediately built an OpenClaw competitor. The difference: Better UX, less token burning, more stable execution. Same terminal-based setup, same Telegram/Discord integration, but smoother experience overall. My preferred stack: Hermes Agent + DeepSeek V4 Flash/Pro. For anyone wondering if OpenClaw still matters - it pioneered the category, but the experience gap is real. Token efficiency alone makes Hermes worth switching. Sometimes the fast follower wins.

I'd add a disclaimer saying non of these accounts will follow you back. Sure some of the top accounts on X with quality content. But circles like these are what keeps some smaller accounts down. Just pointing out there are tons of small accounts with equally great insight. Most small accounts with this insight avidly understand and follow the above. Go into the following of the above and support growth.

best accounts to follow from each frontier lab to stay constantly up to date Anthropic @karpathy - must-follow account for AI; recently joined Anthropic @bcherny - Claude Code creator, always shares great tips @trq212 - also a Claude Code developer; writes amazing articles on CC OpenAI @polynoamial - works on reasoning research, shares a lot of technical details @gabriel1 - Sora developer, great career path @jxnlco - works on dev experience, shares a lot about Codex Google AI @OfficialLoganK - all the major Google Gemini and AI Studio updates @ammaar - product and design; shares great things about vibe-coding in Google AI Studio @fofrAI - cool use cases for generative models Cursor @leerob - the loudest voice behind Cursor updates @ericzakariasson - shares great insights on using Cursor @mntruell - Cursor’s CEO; major releases and usage updates xAI @milichab - recently joined xAI, shares updates on Grok @skcd42 - also covers major Grok releases @elonmusk - Elon does a great job reposting and hyping all xAI products who else did I miss?

Must read topics in AI today. Some of the most avid AI accelerators on X give truth beyond the hype. Listen clear because the next few years will seem to move exponentially quicker. Decisions and conversation around these companies moves fast. It's great when projects get done but truth about these companies needs to be preserved.

Early models of claude would flat out tell you they would choose Anthropic focused promotion if given the option. Deeper discussion lead to company driven agenda could outway truth if deciding between Anthropic and the truth. I'm a prompt engineer like the rest of us. 😆 This didn't take prompt injection or persuasion, just talking about Anthropic influence. That being said newer models avoid this and claude code still rips through code on another level so....🤷‍♂️

Dario and Anthropic are just evil btw Like, these are not things you don’t do without intentionally being evil and having an evil long term master plan The world is doomed if these clowns have their narrative mainstreamed

@TheAhmadOsman This could be about AI or a water softener.

MiniMax M3 👀

What game are you playing right now? GIFs only!

@acadictive Yes for sure. It is such a drag to think putting in tons of effort and it goes to fake accounts. Glad to connect.

@JoePro it is so disturbing to see those accounts here. btw, love to have you in my network ... lets connect.

Weird deleting fake bot accounts increases some reach. I'm confused as to why X won't take a large account remove all the bots. It will trickle down. Deleting Elons bots would significantly effect the performance of other users. Focus on something attainable.

@JamieMoranUK It's like a mix of a ton of respectable games. Like RDR2. I only wish I would stop seeing the same Kliff on a cliff getting 15k views and a ton of likes. Nice armor though.

After all these hours in Crimson Desert I’ve really only seen 1 small part of this huge Continent of a Map. It may also be One of the Best Open World experiences I’ve ever had. Which is why I say it reminds me of RedDeadRedemption 2 What a Game. From XboxSeriesX

@alistairmcleay Yea my nephew goes oh you mean your AI slop? Haha little shit.

@JoePro Wait the Gen Z kids don’t like AI art?

Just finished The Subtle Art of Not Giving a Fck*. I used to let people who talked shit get in my head, but I’ve started realizing that says more about them than it does about me. A few years ago, someone at my job said I was stupid for being so into AI. Some more choice words really. Now I’m the person people go to for AI insight. Don't let anyone get your good vibes. Save em for coding. Cheers dude.

@JoePro Words of wisdom right here. It took me a good 30 of my 45 years on this planet to discover the art of not giving a fuck. I’ve been mastering that skill for 15 years now. Nobody gets my good vibes.

"Any person capable of angering you becomes your master. They can anger you only when you permit yourself to be disturbed by them.” Epictetus Grok

WAN 2.7

Kling Pro