Padrasta

@blankspeaker Will Fast mode ever be able to use Connectors and Skills?

SpaceXAI: Grok 4.3 is out of Beta! I wonder what is coming next from SpaceXAI

@techdevnotes So for now Fast-mode can not use the Skills and Connectors and Auto-mode is not stepping in, routing to Expert, when Skills or Connectors are required. That makes Auto and Fast functionally redundant for anyone using Grok's Skills and Connectors. @xai

Fast and Expert are now powered by Grok 4.3 Heavy still seems to be 16 Agents of Grok 4.20

@XFreeze Not on my account.

Connectors are now live for all SuperGrok Heavy and SuperGrok users You can now connect: • Gmail • GitHub • Google Calendar • Notion • Google Drive • your own custom connectors Why this matters: Your AI assistant just got real superpowers It can now pull live context from your emails, calendar, code, notes, and files - no more need for copy-pasting or jumping between tabs • Summarize your inbox • Prep you for meetings • Analyze a repo • Pull insights from your Notion or Drive You just ask - it does the work This is where AI stops being a tool… and starts becoming a true personal assistant

@LuizaJarovsky Where your bias lies is strikingly obvious. 🧐

🚨 The rise of AI adoption will make the world LESS human and community-oriented. Most people haven't realized it, but there are consequences: Services might become faster or more efficient, but also more prone to AI-related biases, flaws, and risks (whether this is better or worse depends on whom you ask). We will not make services perfect or ideal. We will replace human flaws with machine flaws. The last time we made services significantly less human and less community-oriented was 20 years ago, with the rise of smartphones and social media. One of the psychological and social consequences of that technological disruption is today's loneliness epidemic (and its mental health blueprint). Technological disruptions that systemically reduce human interaction and disconnect people from their communities will ALWAYS have an individual and social price, as those factors are directly connected to human flourishing. We will govern AI better when we understand that technological disruption is not inherently positive or risk-free, and that we might have to aggressively protect people if their basic human and community needs are not met. Read more about AI's legal and ethical challenges in my newsletter (below)

@sama Sam, you mean the models you keep to yourself and your exclusive little circle (like the military) can actually do that. Models like the 4-Series. Not the ones you throw at the stupid masses. Those just refuse the query and bury you in safety disclaimers.

AI will help discover new science, such as cures for diseases, which is perhaps the most important way to increase quality of life long-term. AI will also present new threats to society that we have to address. No company can sufficiently mitigate these on their own; we will need a society-wide response to things like novel bio threats, a massive and fast change to the economy, extremely capable models causing complex emergent effects across society, and more. These are the areas the OpenAI Foundation will initially focus on, and in my opinion are some of the most important ones for us to get right. The Foundation will spend at least $1 billion over the next year. @woj_zaremba, co-founder of OpenAI, will transition to Head of AI Resilience. I believe that shifting how the world thinks about safety to include a Resilience-style approach is critical, and I am extremely grateful to Wojciech for taking on this role. Wojciech has been my cofounder for the last decade; anyone who knows him will understand what I mean when I say he is one of a kind. He has a lot of ideas about how we build a new kind of AI safety. @JacobTref is joining as Head of Life Sciences and Curing Diseases. @annaadeola, our VP of Global Impact, will transition to Head of AI for Civil Society and Philanthropy. @robert_kaiden is joining as Chief Financial Officer. @jeffarnold is joining as Director of Operations.

@gdb Gorgias would've facepalmed at this pathetic Argumentum ad nauseam display. Just because you spam the same corporate talking points endlessly doesn't make them true. 🥱

team really cooked with gpt-5.4

@Jacobsklug @claudeai ditto

I’ve officially moved over to @claudeai. Been a great decision. Not only do I respect the company a lot more, than the alternative — it’s also way better. In chat, it actually challenges me and pushes back. And co-work has just been a game changer. Highly recommend you make the switch.

@LuizaJarovsky This comes from someone who needs to put "PhD" in their username cause otherwise they'd just be another digital dinosaur afraid someone might take their legal pad away. Same governments that can't regulate social media, can't rein in Big Pharma, are going to nail this one. 🤣

Do you know why AI is not going to "take over"? Because strong AI laws are coming. Some of them are already being proposed, and many in the AI industry are sad because their megalomaniacal power fantasy will not come true. AI will be tightly regulated, just like other tools.

@haider1 So we will get a Downgrade every month..... Okay...

not many people noticed this, but with gpt-5.4, we have officially moved to monthly release cycles. openai and anthropic are aiming for this now these updates will not be small. they could be even bigger than the jump from gpt-5 to gpt-5.2 pro, which was already a huge update

@gdb Yesssss a very big step forward into the wrong direction. 😂

GPT-5.4 is a big step forward

@gdb Impressive for Corporate Users. For everyone else it is still a mess.

gpt-5.4 is very strong at productivity tasks with excel and word:

@Tiny_Fish @OpenAI But it is still so cringe...

We had early access to @OpenAI's new GPT-5.4 model. At TinyFish, we run web agents on some of the hardest real-world tasks: multi-step workflows, dynamic pages, live websites with no APIs. GPT-5.4 is by far the best OpenAI model we've tested. See what that looks like 👇

@VraserX Yeah and that's so sad.

OpenAI partnering with the Department of War might be the ultimate insurance policy. Once your tech becomes national security infrastructure, bankruptcy stops being an option. Sam Altman playing 4D chess.

@GaelBreton Yes? Is it less cringe?

GPT 5.4 pro makes Opus feel like a cheap model 😅.

@shiri_shh How cringe is it?

@DeryaTR_ Funny how the 'most advanced AI ever' is apparently too advanced for the State Department. Almost like they don't want a model that could connect dots they'd rather keep disconnected.

Not only are there no walls, but AI capabilities are advancing exponentially! GPT-5.4 is another step change in this progress! Any cynics left will be forced to realize this the hard way soon!

@shiri_shh Funny how the 'most advanced AI ever' is apparently too advanced for the State Department. Almost like they don't want a model that could connect dots they'd rather keep disconnected.

@chrisgpt Funny how the 'most advanced AI ever' is apparently too advanced for the State Department. Almost like they don't want a model that could connect dots they'd rather keep disconnected.

I feel like OpenAI didn’t talk about this enough Humanities Last Exam 5.4 - 39.8% (no tools) 5.4 pro - 42.7% (no tools) Humanities Last Exam 5.4 with tools - 52.1% Humanities Last Exam 5.4 pro with tools - 58.7% GPT 5.2 - 34.5% (no tools) GPT 5.2 Pro - 36.6% (no tools) GPT 5.2 - 45.5% (with tools) GPT 5.2 pro - 50% (with tools) We saw a 8.7% jump in just two months! And a 6% jump with no tools!

@nummanali Funny how the 'most advanced AI ever' is apparently too advanced for the State Department. Almost like they don't want a model that could connect dots they'd rather keep disconnected.

GPT 5.4 imminent This release might be the one to completely remove the need for Opus Rumoured up to 1M context too

@cryptopunk7213 Funny how the 'most advanced AI ever' is apparently too advanced for the State Department. Almost like they don't want a model that could connect dots they'd rather keep disconnected.

GPT 5.4 is officially better than most people at doing their jobs! - its a BEAST at excel / spreadsheet work, bye bye 50% of finance work. - its *really* fckin good at using a computer, officially beats humans in tests. - coding: anthropic's claude, still holds the crown here, but not by much. gpt 5.4 can throw down. - 33% less hallucination = doesnt bullshit you - #1 in scientific reasoning - 1M context window (slam your entire codebase/book into it) but 47% FEWER tokens = its cheap af - you can interrupt its thinking to steer it your own way. overall solid model update imo, nothing groundbreaking but i am so fucking psyched at how quickly these models are coming out