People praising the starship landing, are the same people hating that the Optimus was tele operated.
They love you when you’re winning. They hate you when you’re down.
Respectfully go fuck yourself. Don’t you see you will soon meet your own demise when they become autonomous?
Just to keep you guys excited, here’s all we have to look forward to at the end of the year
Gpt 5
Grok 3
Gpt o1 full
Midjourney v7
Claude 3.5 Opus
Gpt 4o voice mode
Optimist Gen 3
Robo taxi
Remember when we predicted AGI would arrive in 3-4 years? It's time to start preparing because the journey leading up to it won't be dull.
Dev day is in 10 days.
Interesting find on Gpt 4o Voice Mode
**DO NOT LEAVE IT SITTING OUT LIKE ALL THEY SAY IN THE DEMOS**
A lot of open ai employees have been saying “oh i like to leave it out and talk to it when i need” but for us if we do that, we will just run out of rate limit…
As one of the people who’s “been shown” these glasses theyre pretty cool 😭
Don’t count meta out, they might not win in LLMS but they have countless hardware projects they are working on
Just got off the phone with my meta leaker
juicy leaks
-llama 3 just the beginning
- they will announce another ai system that beats gpt before end of fiscal year
- may be closed source
-outside of the llama family
-meta ray bands glasses 2 late this year
There are many people I worked with at Anthropic whom I love and who still work there. I believe they are doing good work. However, the HR hearsay and the amount of bashing I’ve seen about OpenAI is ridiculous. No company is perfect, and Anthropic is no exception. Be Kind
just called my meta leaker he DID NOT want to release much but here’s what I got
llama 3 is 1000% confirmed for the summer finished red teaming and all that jazz
meta ray bans glasses 2 is in full swing
llama 3 will be really exciting
llama 3 will beat gpt 4 benchmarks
Not so juicy leaks but still interesting
- meta has internal predictions for other major ai companies and when they’ll drop
- they have a spreadsheet for a release schedule for all ai products that only the researchers and devs have access too
-zucc knows all release dates
The CEO of my previous company said that AGI could arrive in two years (2027) AND IT COULD WORK 10-100X FASTER THAN HUMANS AND EVERYONES ACTING LIKE THIS IS NORMAL
@adcock_brett
You have no product? That is even remotely at minimal manufacturing, and I still can’t get a flight on any of archers products. How dare you compare figure to Anthropic and open ai and say you look “cheap” 😭 ship one robot to a consumer then say that
God this is playing out exactly like a sci-fi movie before everything goes horribly wrong
"[OpenAI CTO] Murati, too, had been concerned about safety.
Unlike Sutskever, Murati decided to stay to try slow down Altman's accelerationist efforts from within."
Warning after
@QuantumBreaks_
@DrLoupis
How are they not special forces they snuck into a hospital undetected with kitted out m4s and no one even heard the gun shots and they left without anyone noticing for 2 hours
i remember waking up every morning last year excitedly checking
@flowersslop
and
@apples_jimmy
leaks from open ai &hyping me up about the future, thank you guys for making me look forward to something everyday lol
@dylanseychell
more than that it is a valid point but it doesn’t make sense, why would he build his entire legacy in life off a lie it doesn’t seem like a very safe move for him to gain investments in the future if this didn’t work out considering how young he is
it’s actually really smooth the only notes I would get for ChatGPT 4o is that I don’t like when I hear it click and then it goes to like slow latency mode like how the old voice mode was, and tends to have a lot of glitches toward the end of your limit
This weekend, the
@xAI
team brought our Colossus 100k H100 training cluster online. From start to finish, it was done in 122 days.
Colossus is the most powerful AI training system in the world. Moreover, it will double in size to 200k (50k H200s) in a few months.
Excellent
We’re testing SearchGPT, a temporary prototype of new AI search features that give you fast and timely answers with clear and relevant sources.
We’re launching with a small group of users for feedback and plan to integrate the experience into ChatGPT.
With context solved, we now focus on unbounded inference-time compute as the next (and potentially last) breakthrough we believe is needed to build reliable AGI.
Imagine if you could spend $100 and 10 minutes on one task and reliably get a great pull request for an entire
yoooooo grok 2, which was trained on a very small amount of gpus compared to what they have now already beat the first version of gpt 4o mini THE EMPIRE STRIKES BACK BABY GROK 3 by the end of the year
i’m so excited for agents because once we can see where they are at and what they will be able to remember and how much they’ll be able to cross reference: then we can finally see how far we are from AGI
@adcock_brett
i’m not going to post my resume and the frontier models I worked on, however I am pointing out $2.6 billion does not look cheap next to companies that ship useful products to millions of people. you’re using open ais intelligence which is good. Now it’s time ship within >4 years
is there any better graph to explain what’s happening at openai? imagine orion (not meta, lol) being on top of it—would you still be able to keep your job?
How do we have people who think artificial intelligence doesn’t exist? Artificial intelligence absolutely exists by definition, as it refers to machines or software that can perform tasks typically requiring human intelligence we have this in many domains
@adcock_brett
no im pointing out that you say a 3B eval looks cheap, yet you have a couple robots doing basic work in a bmw factory. its incomparable to anthropic or open ai who have shipped products to millions
also, remember who owns you.
LTM-2-Mini is our first model with a 100 million token context window. That’s 10 million lines of code, or 750 novels.
Full blog:
Evals, efficiency, and more ↓
Sometimes I wish I could post more about stuff but sometimes the leaks are so deep you get genuinely scared lol. You could easily say anyone who says this is larping. But it holds true for all real leakers too
Exclusive: As OpenAI looks to raise more capital, it's trying to launch AI that can reason through tough problems and help it develop a new AI model, 'Orion.'
From
@erinkwoo
,
@steph_palazzolo
and
@amir
@masfiq018
They planned on releasing 4.5 but from what we’ve been told, they changed their mind recently. They instead are focusing entirely on “GPT-5” to beat Opus as it has already been red teamed for months. They focused on improving 4o as a substitute to 4.5. We still want 4.5 though!
You failed to deliver you said AGI is here. You gave a plethora of things that you thought were going to come out tomorrow and none of them became true.
SAME ALTMAN JUST SAID “ITS A PRIVILEGE TO BUILD AGI” YET AGAIN DOUBLING DOWN THAT HE WILL RENDER ALL COGNITIVE LABOR REDUNDANT! Nobody’s losing their mind at this???
We tested GPT-4o's voice capabilities with 100+ external red teamers across 45 languages. To protect people's privacy, we've trained the model to only speak in the four preset voices, and we built systems to block outputs that differ from those voices. We've also implemented