Nexp

I track Al Models and Al News

[Nexp @Mr_nexp]

🚨 OpenAI needs to be careful with this upcoming usage change.

From December, ChatGPT and Codex will draw from the same allowance on Plus, Pro, and Business subscriptions.

Even if instant chat remains excluded, Plus users could still feel the squeeze when switching between everyday ChatGPT tasks and serious coding sessions.

We're already paying for a limited amount of usage. Combining both services without increasing that allowance could make things worse.

Anthropic is making its plans more attractive while OpenAI risks pushing users away.

Don't give Plus subscribers another reason to switch.

[Nexp @Mr_nexp]

@codexresets_ Let's hope for the best

[Nexp @Mr_nexp]

🚨 GLM 5.5 Update:

The model appears to have entered a serious post-training phase,

with https://t.co/RPk0ztLzbX seemingly working across internal 32B, 122B, and 256B models before merging them into its next-generation base model.

It's been a while since GLM 5.3 dropped, so hopefully we won't have to wait much longer for the next release.

https://t.co/RPk0ztLzbX might have something interesting coming.

[Nexp @Mr_nexp]

🚨 Days 26–28: OpenAI just gave us another reason to expect something BIG.

Two banked resets are officially coming, meaning users can save up extra usage instead of letting it go to waste.

Why make this move now? GPT-6.1 Astra is the obvious guess, but OpenAI could have something even bigger planned.

Whatever they're preparing, next week could get VERY interesting.

[Tibo @thsottiaux]

Day 26/ You can now easily switch between multiple accounts on mobile, and coming soon to desktop too. Next two days will be resets… banked… do with this what you may 👀.

https://t.co/iQxsqbuk0f

[Nexp @Mr_nexp]

🚨 Something BIG is happening across the frontier AI labs.

OpenAI is increasing usage ahead of Astra 6.1.

Anthropic is removing the 5-hour restriction as it prepares for Fable 5.5.

Google could also be making room for Gemini 4 Carbon with multiple prompts on Max Ultra.

This doesn't feel like a coincidence.

We're approaching a point where AI agents could handle increasingly complex development tasks with very little human input.

The next generation of models could redefine what's possible.

I don't think we're ready for what's coming.

[Nexp @Mr_nexp]

Anthropic just made the 20x plans a lot more appealing.

Removing the 5-hour limits is a huge upgrade for anyone using Claude heavily.

Now imagine pairing that with Fable 5.5.

Anthropic, you've handled the usage limits. Time to deliver the new model.

[Claude @claudeai]

We’re renaming our Max plans:

Max 5x → Max
Max 20x → Ultra

Max is still $100/month, and Ultra is still $200. If you’re a subscriber, there’s nothing you need to do.

We’re also removing the 5-hour limit for verified Ultra accounts, starting today with most existing subscribers. Newer accounts keep a temporary limit until they’re verified, which helps us prevent fraud and abuse. You'll get an email when it's removed.

Weekly limits remain unchanged. View our plans here: https://t.co/niRpyGF5Me

[Nexp @Mr_nexp]

@codexresets_ We deserve some love ♥️ too

[Nexp @Mr_nexp]

🚨 Anthropic is drawing a line between AI-inspired games and straight-up copies.

Claude Code now has instructions to avoid recreating existing games, even when users change the original title. It should focus on fresh concepts or build around broader game genres instead.

This could hit people using Claude to make unofficial Minecraft, GTA, Terraria, or Halo clones directly in the browser.

AI can help build games, but cloning existing ones is becoming a different story.

Looks like AI game development is getting some new boundaries.

[Nexp @Mr_nexp]

🚨Binance Hack So Far

«They've been hit by the biggest crypto hack in history, with around $6.3 billion affected.»

«Hackers gained access through a compromised laptop belonging to a Binance engineer.»

«Security researchers suspect North Korea's Lazarus Group, specifically its TraderTraitor hacking unit, was behind the attack.»

«Around $3 billion of the affected crypto has already been frozen with help from other companies.»

[Nexp @Mr_nexp]

@neerajjj6785 Couse i have got Claude

[Nexp @Mr_nexp]

@codexresets_ Agreed 💯

[Nexp @Mr_nexp]

🚨Codex Users!!!

All eyes on OpenAI today. The 28 Days of Shipping challenge is coming to an end.

If there's a major model waiting to drop, this is the perfect moment.

Astra 6.1 might be much closer than we think.

The Codex team has a real opportunity to challenge Claude's dominance and put OpenAI back at the top.

Anthropic has enjoyed the spotlight for long enough. Could today be the day Opus 5.5 finally gets dethroned?

[Nexp @Mr_nexp]

🚨 ChatGPT Plus users deserve doubled usage limits too.

OpenAI is finally giving heavy users more room with bigger allowances on its higher-priced plans. That's a welcome change, but Plus subscribers shouldn't be overlooked.

Not everyone has the budget for a $100–$500 monthly plan. Many Plus users depend on these models daily for coding, work, and other tasks.

The benefits shouldn't stop at the expensive tiers. It's time to give Plus users a meaningful usage boost too.

[Nexp @Mr_nexp]

🚨 Codex Users!!!

OpenAI just gave Codex users a reason to stay.

Usage is set to double across the $100, $200, and $500 tiers. That's a massive change for anyone burning through limits while coding.

I was ready to walk away from my subscription tomorrow. Now I'm reconsidering.

Credit to Tibo for hearing the complaints and pushing for a real improvement.

More usage is great. Now give us Sol 6.2 and Astra 6.1.

[Tibo @thsottiaux]

Day 25/

We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today.

We heard your kind feedback and won’t be reducing usage limits for Pro 200, so your Work and Codex usage will stay the same. We’re also going to continue offering extended limits for dots for now, and usage of your dot will still continue to be separate from your default usage.

And we'll of course also keep shipping better models and other efficiency improvements that increase your effective usage over time.

[Nexp @Mr_nexp]

Anyone else saving Opus 5.5 for the big tasks and doing most of their coding on Sonnet 5.5? 😭

Opus feels like a cheat code, but the usage limits on Claude Max 20x make it hard to justify using it for everything.

I'd rather have a slightly slower workflow that lasts all week than go all-in on Opus and hit the wall halfway through.

The real skill might be knowing when not to use your best model.

How are you balancing quality vs usage? Is Opus worth the extra cost and quota drain for your workflow?

[Nexp @Mr_nexp]

@codexresets_ Can't wait

[Nexp @Mr_nexp]

@hiarun02 Coaches don't play

[Nexp @Mr_nexp]

@Mr_Salio Big week ahead 😄

[Nexp @Mr_nexp]

@codexresets_ That's wild

[Nexp @Mr_nexp]

🚨 Is OpenAI slowly losing its edge?

Ever since the Hugging Face incident, it feels like they've been losing momentum.

Models are getting slower, less capable, and less efficient compared to both US competitors and Chinese labs.

The decisions keep getting worse, and we still haven't seen a genuinely impressive new model. And no, I'm not talking about Terra.

OpenAI went from setting the pace to constantly playing catch-up.

Imagine telling people a year ago that OpenAI would be struggling to keep up. The AI landscape is changing fast. If they don't turn things around, will anyone still consider them the clear leader by the end of 2027?