I was having issues with @grok 4.6 on my @openclaw agent taking too long to respond. It turns out, it was thinking too much! Probably because of the built-in agent loops in the harness so when I turned thinking low, it actually produces better results and 57% faster!

Cancelled all weekend plans to build with @OpenAI 's new Astra 6 Model. I'll give full review of @ChatGPT 's latest and greatest (and got 2 resets to burn!) #ChatGPT #GPT6Astra

Me vs @claudeai - should I do one for @ChatGPT and @bot ? 🤐🤔🤪

Lol 🤣 true #ai

Don't even use eleven labs anymore because of this 🤖 @NVIDIAAI @FishAudio

🚨 @grok just joined the usage reset race. I have 𝟭 banked SuperGrok Heavy reset sitting in my account right now. One button. Back to 𝟭𝟬𝟬%. That matters more than people think. OpenAI has been absolutely crushing the marketing side of AI subscriptions over the past month. Not with ads. With usage resets. They trained builders to feel like the subscription keeps getting more valuable after the payment clears. Hit the cap. Get reset. Keep building. That is how you make power users feel spoiled. I have been begging @claudeai to do the same. And I say that as somebody who made $𝟮𝟴𝟮,𝟬𝟬𝟬 with Claude in 2023. Claude was early. Claude was powerful. Claude owned the long-form content lane before most people understood why context mattered. But @AnthropicAI has gotten very comfortable. Claude still does not give me the same direct harness experience that @OpenAI and Grok give me. Meanwhile OpenAI and Grok are turning usage limits into a 𝗺𝗮𝗿𝗸𝗲𝘁𝗶𝗻𝗴 𝘄𝗲𝗮𝗽𝗼𝗻. That is the part people are missing. A reset is a behavior trigger. It tells builders to go harder. Burn the tokens. Run the agents. Test the workflow. Break the prompt. Build the thing. Then come back and click the button. Grok giving me a reset that expires on September 13 means I am going to use the hell out of Grok before September 13. That is exactly the behavior they want. OpenAI figured this out first. Now Grok is copying the right move. Claude should be paying attention. Because model quality matters. Access matters too. And builders do not build where the model is theoretically good. They build where the friction is lowest and the value keeps stacking. The competition is getting fierce. OpenAI spoiled us. Grok just raised its hand. Claude can either respond, or watch builders move their daily workflows somewhere else. The reset button is small. The message behind it is not... ⭐⭐⭐⭐⭐ 5 Star Savage Marketer Rating

We are seeing an evolution of OpenSource

Yo @OpenAI @thsottiaux - What ever happened to this? I've put the project on Pause since submission since it was supposed to be announced August 12th? Really excited to go back in to the repo and actually finish what I started 100% with Codex & 5.6 Sol Ultra! Thanks again for the opportunity @devpost devpost.com/software/emberbo… #ChatGPT #devpost #Codex #BuildWeek

Lol .. truth. 🤐

Yet another @OpenAI ChatGPT usage reset. Thanks @thsottiaux & @sama 💯

Anthropic's Fable 5 rollout has been a 𝗰𝗼𝗺𝗽𝗹𝗲𝘁𝗲 mess. ❌ Terrible communication ❌ Gatekeeping their own flagship model ❌ Random refusals ❌ Abrupt plan changes with no warning And now? They announce 50% usage is "𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱." Translation: 20x users now get 𝗵𝗮𝗹𝗳 the previous limit. Pro and Team plans? Pushed to paid credits. And the replies are... thanking them? "Thank you for the clarity!" "Appreciate the update!" This is Stockholm syndrome in real time. 😂 They made a mess. Then announced they're 𝘬𝘦𝘦𝘱𝘪𝘯𝘨 the mess. And people are cheering. Before anyone says I'm hating... I have 2x max 20x plans. I use Anthropic models 𝗱𝗮𝗶𝗹𝘆 for various workflows. I run the BEST Claude Community in the world with 1.4+ Million members and growing by 30,000+ every week. I'm not a hater.... This is 𝗳𝗲𝗲𝗱𝗯𝗮𝗰𝗸. You don't thank a restaurant for announcing your meal is now permanently half-sized. You don't applaud an airline for making delays "standard policy." Anthropic makes world-class models. But their rollouts, communication, and customer experience? A mess. And praising them for it doesn't help anyone. Not them. Not us.

Most people are using Claude like it’s just another chatbot. That’s why they keep getting average results. The real advantage isn’t collecting more random prompts. It’s knowing: ☑ When to use Claude ☑ When Cowork is the better choice ☑ When a project belongs in Claude Code ☑ How to give Claude enough context to produce usable work ☑ How to turn one good result into a repeatable workflow So I created The Claude Power User Kit. It includes: ☑ A simple Claude/Cowork/Claude Code decision guide ☑ 19 copy-and-use prompts ☑ Practical workflows for real business tasks ☑ Quality checks to catch weak or unreliable outputs ☑ A step-by-step action plan ☑ A guide to turning your Claude skills into useful offers and income I’m giving it away free. Comment CLAUDE below 👇 ... and I’ll send you the Claude Power User Kit.

Looking to connect people on @X if you’re into - vibe coding & video games - AI Marketing - Claude Code, Grok Build - Codex, Cursor - build in public say Hi or drop what you’re working on looking to follow active ones 👊👇

OpenClaw v2026.7.1 is out - and this one is a beast. I've personally worked to test this one (I'm a contributor). I've been using the Beta for 2 weeks now, and it's READY! This is a full-blown platform with a small army building it. The Control UI got a complete overhaul. Finding, organizing, pinning, renaming, and archiving your conversations is way easier now. There's a live Tasks view so you can watch background work happen or stop it mid-run. And you can inspect your usage, costs, quotas, and Gateway health right in the browser without digging through config files. If you actually use the Control UI, this is a night and day upgrade. Setup got way easier too. Onboarding now walks you toward a working first chat, checks your connections before saving, and doesn't lose your choices if you get interrupted halfway through. (If you ever installed my AI Persona OS skill for OpenClaw, you'll have already experienced similar onboarding) For anyone who's helped a newcomer set up OpenClaw, you know how much this matters. The official apps got major work. iOS, Android, and macOS all improved across setup, chat, voice, and recovery. Your recent chats stay readable during disconnects, and mobile messages wait in the right conversation even through an app restart. Still not how I'd run it day to day, but they're getting better fast. Now the part I care about most. Models and providers. GPT-5.6 compatibility across OpenAI and Codex routes. Of course Grok 4.5 (I'm rocking it now, and it's great - especially since it's one of the ONLY AI Subscriptions that give you use of the Image AND Video gen features in 3rd party harness). That means, your AI Agent can use Grok Imagine, without API. New Tencent Hy3 support. New Meta Muse Spark 1.1 support. Plus broader Claude, Ollama, and ClawRouter improvements. The model list just keeps growing. More options. More fallbacks. More ways to route the right task to the right brain. Codex and connected coding agents got stronger. Here's a slick one. openclaw attach gives Claude Code temporary, revocable access to a Gateway session. So you can pull Claude Code into a live session, let it work, then cut the access when it's done. Codex delegation returns tracked results more reliably, and long-running sessions are easier to resume. Every messaging channel got real attention. Telegram got albums, documents, topics, live progress, and better delivery. Slack keeps thread history more consistently. Discord improved replies, attachments, and voice sessions. Apple Messages got better routing and chat continuity. And here's the fix that made me smile. Gateway crash loops are dead. Repeated failures now leave a stable repair path instead of relaunching forever. If you ever watched your Gateway restart into infinity, you know exactly why that one matters. Plus scheduled work can now wake only when something actually changes, signed-in Chrome tabs can pair with a remote Gateway, and guarded workspace terminals work across the Control UI, iOS, and Android. Here's my honest take on all of this. Real infrastructure being build here. More reliability. More models. More control. Fewer ways to break. That's the stuff that lets you run this in a real business. And that's exactly why I keep betting on it. OpenClaw is absolutely THE best OpenSource Agent right now. And don't get me wrong, I love Hermes too, but it's just facts.

Ladies and gentlemen... I present to you... The BIGGEST upgrade to your AI Agents to date. And... Best part is... Included in your ChatGPT subscription. And... The subscription allows 3rd party harness... Meaning... OpenClaw gets an upgrade. Hermes gets an upgrade. ZeroClaw gets an upgrade. NemoClaw gets an upgrade. And... Your own custom built Agents get an upgrade.... Without paying extra for API usage... This Thursday... It's time. And OpenClaw already has THE BEST Codex Harness available and with Peter and Sam working together... This is just my prediction... But I think this will now be the BEST Agent you can build for the best price on in the Universe (I say that because Grok technically is not even on Earth, it's in Space soon and Anthropic is launching data center in space too with Space X) You think I'm right? Or am I just fanboying? 🤖🦞🤣 Time will tell.

Great opportunity for the right one👇

DO THIS before you lose access to Fable 5 tomorrow ! 🤖 After July 7 it's pay-per-token. The smartest model most people have ever touched, and they used it to write emails. So before it goes, I made it audit my ENTIRE chat history. It scored me 6 out of 10. Publishing it anyway. Here's what it caught: → I solved the same install problem 4 separate times instead of building a system once → I re-paste my brand guidelines by hand in almost every chat → I started a copywriter agent in May and abandoned it at phase 1 → Verdict: "elite output per session, almost nothing that compounds between sessions" Then it found my blind spot: "You never ask what to kill." Hundreds of chats asking how to build. Zero asking what to shut down. That one line was worth more than my subscription. The audit is 658 words. Two prompts. It digs through your history, ranks your highest-value moves by impact vs effort, scores your usage, and builds you a report card in your browser. You have about 24 hours with the best model to run it. Comment AUDIT and I'll send you the prompts 👇 PS. Post your results and hold yourself accountable See less