When viruses came, a whole antivirus business was born to stop them. Now AI is helping hackers find bugs and break into online services so the next big business could be AI that protects us from AI. Just like before, every new kind of attack creates a new industry to stop it.

Claude Duet hype is almost dead because all eyes are on the recent security breaches.

Tibo might know when Fable 5.5 is coming out before Claude does. Claude seems to be waiting for Tibo to hint at the next ChatGPT model, then they get Fable 5.5 ready and drop it right before. So if you want to know when Fable 5.5 is coming, don't ask Claude. Ask Tibo.

I used to think AI was the one writing bad and unsafe code, like a new worker who makes all the mistakes. But this month showed that humans have been writing unsafe code for years. Poor AI took all the blame.

I asked every major AI about the PHP RCE exploit and they all refused to share exploit code or attack steps. Then I tried DeepSeek and it handed me everything with full steps, payloads and a complete guide So could open source AI be helping fuel some of these attacks?

WTF is happening with security this October > PHP - two critical bugs that can take over most PHP sites > GitLab - 10/10 critical bug going back to 2014 > Windows - zero-day already being used in attacks > Cisco Expressway - 9.8 bug giving attackers root > Railway - customer secrets and API keys leaked > SafePay - ransomware hit 600+ critical infrastructure targets > eMudhra - certificate keys so weak browsers blocked them > ConnectWise - backups used to attack 2,250+ Microsoft 365 accounts > Chrome - 353 bugs fixed, 6 of them critical > Node.js - bug that allows remote code execution > Java - emergency fix for a 9.1 critical flaw > BeyondTrust - zero-day used by hackers since August > Ivanti - VPN login bypass used by attackers > Cisco Email Gateway - attacked through email attachments > Palo Alto - root level bug in firewalls > Onamae - 27,000 servers shut down after malware > LiteLLM - release shipped with a hidden password stealer > OVHcloud - CLI downloads swapped with malware > GitHub - 94 repos hit with a hidden password stealer > UpdraftPlus - stolen keys used to backdoor WordPress sites > Plex - bugs chained together in real attacks > SiteGuard - root level bug under attack > AhsayCBS - bug chain giving full system control > Siemens - Software Center used in active attacks > Zeabur - breached again and now shutting down > OpenCode - AI API keys exposed and users charged > TiDB Cloud - internal systems accessed and data copied > SourceNext - 22.5 million member records stolen > Burger King Japan - 11 million customer records exposed > Nikkei - breached for the third time in a month > SBI - customer data found on the dark web > Rank Math - malware in an official PRO download > SubQuery and Tensorlake - npm packages backdoored And the month isn’t even over yet.

First every lab raced to build terminal coding agents. Then those coding agents got their own desktop apps. Then computer use, AI controlling your desktop. Now every lab is launching a personal assistant that runs in the background. What will they race to build next?

This might be a DUMB question. o1 Pro was once the most expensive model from OpenAI at $150 for input and $600 for output. Now their top model GPT 6 Astra is way smarter, yet it costs only $10 for input and $50 for output. So a much better model is 12 times cheaper than o1 Pro, which makes me wonder what the price is really based on. How do AI companies actually decide the price of a model?

Fable 5.5 might come with a lower price, and it kind of has to. Right now Fable costs 2.5x more than Opus 5.5, and Opus 5.5 is already so good that most users just stick with it. So for Fable 5.5 to make sense, one of two things needs to happen. It has to be way better than Opus 5.5 to justify the price. Or bring the price down to 2x or 1.5x Opus and let users use their full weekly limit on it instead of only 50%.

Many users expected Fable 5.5 to drop on Tuesday, but it didn't. Claude seems to be waiting for the right moment. As soon as ChatGPT announces its next model, Fable 5.5 will drop before it and take all the attention. The Fable 5.5 release date seems to depend on ChatGPT.

When Astra came out so many users were calling it AGI and the hype was everywhere. Now almost nobody talks about Astra being AGI. What happened?

Even if Fable 5.5 drops and isn't as good as everyone expects, Claude will be fine because Opus 5.5 is still ahead of everything else out there.

How are Gemini models still behind Claude and ChatGPT when they have all the data and infrastructure?

A few users are saying Opus 5.5 got nerfed today. I spend half my day on it every day and honestly it still feels exactly the same to me. Are you guys actually noticing a difference or am I missing something?

I still haven't used my Claude banked reset from the Opus 5.5 launch. It's about to expire, but honestly the normal usage has been more than enough so I never needed it. But on Codex, I need at least 2 resets just to make it through one week of work. Never thought a free reset would expire just because the usage never ran out.

Claude loves launching its big models on Tuesdays. Fable 5 came out on a Tuesday. Fable 5.1 came out on a Tuesday. Opus 5.5 came out on a Tuesday too. So if history repeats, Fable 5.5 could be landing tomorrow.

GPT 6.1 Terra might be coming soon. But users don't need another middle model when what they really want is something better than Opus 5.5. Even GPT 5.6 Terra barely got used because people picked Sol for hard tasks and Luna for easy ones. So another Terra feels like one more model nobody asked for.

Anthropic is giving around 20% of its shares to the public good. Elon says he'll give a SpaceX share to every kid in America next year. OpenAI is giving us another reset.

So claude is proving you don't need constant resets to keep users, just decent usage and a better model?

Asked gemini 4 argon what 2 + 2 is and it ran code to find out, tried it multiple times and it ran that code every single time. Every other ai I tried just says 4 right away. Why can’t argon do the same? The smartest gemini yet needs to write code to add two numbers.