GPT-5.4 Pro is an absolute beast.
07.03.2026 11:46
👍 0
🔁 0
💬 0
📌 0
GPT-5.4 Pro is an absolute beast.
5.4 is rolling out fast globally.
Finally. Well done 👏
GPT-5.4 is finally live. Time to take it for a spin. 🔥
These are what I've been waiting for:
"Context left: 10%."
Reality: 0%.
Nothing kills the vibe of a coding session quite like Claude Code hitting a wall it claimed was miles away.
The worst part is that it gave hope that /compact would actually save me.
It did not.
> 2023: "It just makes stuff up"
> 2024: "Nothing but hype"
> 2025: "AI won't replace your job"
> 2026:
It seems Anthropic secured the perfect domain for Claude: clau.de
A dialogue box prompts users to import memories to an AI named Claude, detailing the process for exporting conversation data.
This is very smart.
The biggest friction to switching AI models for now was the context gap. We stay where the AI "knows" us.
Anthropic just nuked that memory moat.
If the only thing keeping you on ChatGPT was your history, Dario just opened the door.
If you’re currently monitoring the situation,
step up your geopolitical doomscrolling setup: worldmonitor.app/
Anthropic quietly built out a free full CS curriculum for the AI era.
Claude Code, MCP Mastery, API design. The depth here is startling compared to the random YouTube tutorials I often binge.
This is how you grow an ecosystem: you teach people to build.
anthropic.skilljar.com/
The Board of Peace got bored of peace *really* quick.
We don't ship bad code into production.
But we let bad ethics slide right into our monthly subscriptions.
Time to fix that:
quitgpt.org
Why didn’t Dario simply tell the DoW about:
--dangerously-skip-permissions?
Sam Altman, after Trump banned Claude AI and OpenAI raised $110B
It seems more and more likely OpenAI will just skip 5.3 altogether and go straight to GPT-5.4 👀
PewDiePie enthusiastically speaks about training his own AI model while colorful confetti falls around him.
OpenAI burns billions on compute.
Meanwhile, PewDiePie causally dropped an LLM from his room in Japan that just washes GPT-4o.
It apparently outperforms DeepSeek v2.5 and LLAMA-4 too.
I don't think anyone has "Youtuber solves AGI" on their 2026 bingo card.
It seems we are getting Nano Banana 2 today 🍌🍌
Is it easier to believe this is AI than reality?
This Qixing Mountain sky ladder footage is breaking my brain. The aesthetic, the lighting, the vibe... it looks generated.
Except it is real. 😵💫
We're in for a very confusing decade.
People overestimate what they can get done in a week, and underestimate what Claude Code or Codex can ship in 10 minutes.
One year of Claude Code!
If you told me last year I'd be managing ten different AI agents in worktrees instead of writing unit tests, I would have laughed.
We'll never write code by hand anymore.
We guide it. 🚢
x.com/AnthropicAI...
POV: That moment when GPT-5.3-Codex-Spark starts generating code and your brain simply goes into screensaver mode. 🚀
GitHub repository interface showing the OpenClaw project with sections for Code, Issues, and Pull requests.
POV: You're Peter Steinberger. You write a cool library over a couple of weeks. You think "maybe someone finds this useful", at least you do.
Two months later: you're staring at 4.4k issues and 4.2k pending PRs.
Archaeologists will be digging through these commits in 2050. 🦕
Scorecard for the week.
Predictions vs Reality:
Gemini 3.1 Pro ✅
Grok 4.20 ✅
Sonnet 4.6 ✅
Qwen 3.5 ✅
GPT 5.3 ☑️
Deepseek v4 ☑️
Your move, Sam. 🔥
Forget the datacenter for a second.
Put a Taalas chip in a laptop. Suddenly you have a local LLM that runs faster than you can read.
No wifi needed.
No subscription fee.
Your AI would think with you in real-time. This is how we’ll get true personal computing back.
OpenAI: "We're not really « open »"
Anthropic: "Hold my beer."
This commit in OpenCode says it all.
They really are the most ClosedAI of them all.
github.com/anomalyco/o...
Okay, Gemini 3.1 is showing off now.
It's understanding of SVG animations is out of this world 🤯
Prompt: "Create an animated SVG of the Artemis II moon mission rocket launch. Make it look as realistic as possible."
Gemini 3.1 rolling out to the app worldwide now.
The latest Gemini just dropped. What a wild week in AI.
Almost head-to-head with Opus 4.6 on SWE-Bench Verified. Cannot wait to put this to the test.
The current war between OpenAI and Anthropic.
Summarized in a single image. 👇
