I can officially retire from X now. Sam Altman just quoted my tweet. 🔥

Something feels different after the latest codex reset. Same Sol 6.1 High. Same kind of work. Before the reset it took most of the day to burn a decent chunk say 25%. Now I’m already down to 41% weekly after just a few hours. What exactly changed?

Good news for Google Antigravity Users Claude Opus 5.5 and Sonnet 5.5 are in Google Antigravity now, on Pro and Ultra plan.

Tibo just reset Check your ChatGPT Codex

Grok Bot just reset usage for everyone. Meanwhile ChatGPT users are still waiting when Tibo’s reset is coming.

Gemini 4 Argon status check: Still locked to Fairwind cyber defenders. Still not publicly available through the API. Still no public release date. But Google is already using it internally for massive code migrations, including an 800K+ line section of Fuchsia’s Zircon kernel. And Argon can generate up to 1M tokens in a single trajectory.

Gemini 4 Argon is breathing fire The rest are just watching.

I’ve stopped jumping between models so much. Sometimes you just need to make it think harder. What I’ve noticed: • GPT-6.1 Sol → raise the reasoning first • Opus 5.5 → Medium already handles most serious work • Astra → Max starts mattering on hard agentic tasks • Fable 5.1 → High/Max actually changes how it behaves Model-hopping is expensive. Effort-hopping is often the better first move.

Gemini 4 Argon just dropped a quiet bombshell. Hallucination rate on AA-Omniscience: Argon → 15% Astra → 51% Sol → 54% And it used ~62K output tokens per task. Longer thinking. Way fewer hallucinations. This is the first time Google looks genuinely dangerous on reliability.

Tibo dropped a global reset for everyone tomorrow at 10am PST. Meanwhile I’m sitting here with 70% of my weekly limit still left… and it was supposed to last until Oct 4. So what happens now? Does the free reset just wipe the remaining 70% and restart the clock? Or do I actually get to keep what I already had? You can’t bank it. You can’t choose. It just hits. This is the most OpenAI way possible to “help” people.

Opus 5.5 already feels nerfed. A week ago it was sharp. Now it’s making stupid mistakes, forgetting context, and slipping back into the same old Claude-isms we all hated. Same model name. Completely different feel. This is the exact same pattern every single time. They ship it hot so everyone raves about it… then quietly tone it down once the hype is locked in. Anyone else noticing this, or is it just me?

Claude Fable 5.5 is coming? Fable 5.1 is already routing to something that looks like Fable 5.5. Anthropic is testing the next version on live users without saying a word.

Gemini 4 Argon already announced… but the actual app is still stuck on 3.1 Pro and 3.6 Flash for free users? When does the app actually catch up? @GoogleDeepMind @GeminiApp

Claude shows you the model. Gemini shows you the model. ChatGPT just says “Ask ChatGPT.” All three free version. Why do two companies let you know exactly what you’re talking to… and the other one hides it?

🚨 Codex global reset is landing TOMORROW. Tibo says the reset hits 10:00 AM PST for all paid ChatGPT accounts. Here’s when that is around the world: 1️⃣ 🇺🇸 San Francisco: 10:00 AM 2️⃣ 🇺🇸 New York: 1:00 PM 3️⃣ 🇬🇧 London: 6:00 PM 4️⃣ 🇫🇷 Paris: 7:00 PM 5️⃣ 🇦🇪 Dubai: 9:00 PM 6️⃣ 🇮🇳 Mumbai: 10:30 PM 7️⃣ 🇸🇬 Singapore: 1:00 AM, Oct 3 8️⃣ 🇯🇵 Tokyo: 2:00 AM, Oct 3 9️⃣ 🇦🇺 Sydney: 3:00 AM, Oct 3 🔟 🇨🇦 Toronto: 1:00 PM And the reason? GPT-6.1 Sol had a massive load spike during its first two days. Tibo says it’s now back to expected speeds. Codex users: your tokens are coming back. 🔥

Google: “Gemini 4 Argon is rolling out soon.” Only trusted cyber defenders can use it. No date for everyone else. OpenAI and Anthropic have done this before with their high-risk cyber models (Mythos, GPT-5.4-Cyber, etc.). They drop the strongest model they’ve ever built… then lock it behind a closed program with no release date. This is becoming a pattern.

Gemini 4 Argon is out in name only. Right now: Google internals + vetted cyber defenders in the Fairwind Program. Next: AI Ultra + paid API. Everyone else: later. Announced yesterday. Still can’t try it.