If you are paying or willing to pay for the AI code reviewer,
You should read this article, which is very well summarized by @swarogan
Dhaval retweeted
Day 26, 27 and 28:
OpenAI is officially giving us two banked resets, which means they want us sitting on a mountain of spare tokens for... GPT-6.1 ASTRA, OBVIOUSLY. Or something just as big.
Get ready, because next week is going to be on fire.
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 👀.
help.openai.com/en/articles/…
With this we are done for the 28 days improvement streak.
I am not gonna to judge them harshly but ngl they have done some significant improvements in the Codex and I can feel it.
Thought most of 28 days improvement was like the reset fest for the OpenAI users, but now it is coming to an end.
I hope and really wish by Tuesday we have new better model from OpenAI and they have already hinted us to store banked reset for it.
I am🤞. Let's goo @thsottiaux @OpenAI this was fun tbh!
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 👀.
help.openai.com/en/articles/…
AI agents won't replace software engineers.
They are turning every average developer into a Systems Architect overnight.
Writing syntax was always the easy part. The real engineering work has always been system design, edge cases, state management, and knowing what to build.
Now that LLMs can spit out 2,000 lines of code in 5 seconds, the bottleneck has shifted completely.
The developers who stand out are not the ones who write code fast. They are the ones who can look at 5,000 lines of AI-generated code and spot the exact architectural flaw before it hits production.
For anyone running agents in production right now: what is the single worst failure mode or edge case you have caught an agent making?
Anthropic seems to have manage user expections and usage limit much better than their competitors.
They have just removed the 5-hour limit from ultra plan and there is no Fable on pro plan. Clearly hinting that Fable is not made for lower end plans.
They have already merged the Co-work and chat into one and they both consume same usage limit.
They are keep doing the things that seem easy to follow, but somehow becoming hard for everyone.
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: claude.com/pricing
The most mind-bending thing about current AI is how fast our brains have normalized it.
We went from "holy crap, intelligence is artificial" to "this latency is annoying" in record time.
18 months ago, an AI agent autonomously debugging a script or executing multi-step browser actions would have felt like straight-up magic.
Today, if Claude or OpenAI takes more than 12 seconds to generate 500 lines of clean, functional code, people sigh and refresh the tab.
This is really great news for the Pro and business users, OpenAI have rolled back the cut in usage limit for pro and business users.
That is why I say most of the time. OpenAI is necessary for this AI race. We whole day roast them, criticise them, but they still fully listen to their user base and come out with things that benifits all in general, I still wish they could have included in Plus users and hole their time also comes soon as well.
I am convience and happy for this decision they have taken! They listen to people and implement things and course correct it.
@sama @thsottiaux this is really welcome step. 👏
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.
OpenAI isn't an AI company anymore.
It is an operating system for human cognition disguised as a corporation.
Everyone is obsessing over whether they reach AGI first. But they have already won a much bigger game.
They are not just giving us tools to answer questions. They are quietly embedding synthetic reasoning loops into how humanity makes decisions.
The real moat was never their model weights or training tricks.
The real moat is making human productivity completely dependent on intelligence as a utility.
Hard truth: We are not building with OpenAI anymore. We're outsourcing human execution to them.
Agree or change my mind?
Google dropped Gemini 4 Carbon, and god, this is a damn good, impressive frontier model.
It ended Opus 5.5's reign, which had been running high for a month. It beats Opus 5.5 by a good margin, and no, this isn't bench-maxing, as we saw in the case of Argon!
Every other benchmark, the parameters I checked for Gemini 4 Carbon, I swear, I am not even hyping it all; they really, really are damn good, with 3D rendering, coding, or any other parameter.
So whoever has leaked the news of RSI at Google wasn't lying; first Argon, though it was not that good, but Carbon is a really, really impressive model, and I am loving it.
Now, in the next few days, you will hear about Carbon's impressiveness all over your feed, people making things out of it, and the real goodness of it.
Dhaval retweeted
Everyone knows Tibo from the Codex resets.
Meet Thariq - the guy who did something similar for Claude Code.
Thariq Shihipar (@trq212)
> From Toronto, based in SF
> Studied Engineering Science at the University of Toronto + Master’s at MIT Media Lab
> Co-founded Chime (acquired by HubSpot) and scholarly platform PubPub
> Founded YC-backed gaming company One More Multiverse (raised ~$17M)
> Joined Anthropic in 2025 and spent over a year on Claude Code
> Worked on skills, context engineering, and dynamic agent workflows
> Helped cut the system prompt by ~80% as models got stronger
> Became one of Claude Code’s most visible faces on X by actually talking to users
> Left Anthropic in late October 2026 for a break before his next thing
He went from building startups and games to helping shape one of the biggest coding agents, while staying unusually close to the community.
The person many developers associated with Claude Code just stepped away right before Anthropic’s IPO.
