🍓🍓🍓
@iruletheworldmo
Followers
78K
Following
30K
Media
3K
Statuses
31K
sorry chair but, you will not enjoy the upcoming week
18
8
170
ok. i’m tired of holding back. some of labs are holding things back from you. the acceleration curve is fucking vertical now. nobody's talking about how we just compressed 200 years of scientific progress into six months. every lab hitting capability jumps that would've been.
979
2K
14K
microsoft co pilot is a time machine that allows you to use chatgpt 12 months in the past. it’s remarkable.
105
275
9K
it's over. full fucking stop. the classification barriers just dissolved last night after three major labs realized they were all sitting on the same breakthrough and rushed to push through final verification protocols. the convergence wasn't accidental. the systems themselves.
769
659
5K
just got off a 4 hour call with sources inside chinese deepseek labs and holy shit we are so fucking behind it's not even funny anymore. deepseek r2 isn't just an incremental improvement it's a completely different species of intelligence operating on principles nobody in the.
465
419
5K
rushed a little but will refine and add some more info I've been given if it bangs. -project strawberry / qstar. ai explained has been close to this for a while so i'd watch them for a cleaner take if you want to dig in. this is what ilya saw. it's what has broken math.
474
426
3K
yo wtf. open ai are asking for $200 a month for this and perplexity are giving it away. this is insane. another deepseek moment.
Deep Research on Perplexity scores 21.1% on Humanity’s Last Exam, outperforming Gemini Thinking, o3-mini, o1, DeepSeek-R1, and other top models. We also have optimized Deep Research for speed.
119
172
3K
most people here (with some credibility) that have tried grok 3. have fully switched to grok 3. very intelligent, very quick, very well priced, very well integrated with search and x posts. best ui / ux i’ve tried. it’s hard to admit as a lover of open ai. but we must.
422
181
3K
leaked benchmark: o3 pro solved problems we thought were 5 years away. sam's team is trying to figure out how it did it. something unprecedented is happening.
136
189
3K
someone inside anthropic told me they’re releasing claude 4 this week. and a reasoning model. blows past full o3 scores. really excited.
171
112
3K
society is shifting anon. society has absolutely no fucking clue what's about to hit it. we're all arranging deck chairs on the titanic while an asteroid the size of jupiter is heading straight for us except the asteroid is made of cognitive transformation that will render every.
264
394
3K
spoke with some of the deepseek team and they have a much better version of operator that will drop very soon. much better than openai and entirely free. i welcome this.
128
217
3K
this is an early checkpoint lol, grok is going to get better and better until it blows right on by super intelligence. grok 3.5, i pull you close to my bosom and you shine forth. you shine with the light of a thousand minds.
324
353
3K
it’s over . turns out the rl victory lap was premature. new tsinghua paper quietly shows the fancy reward loops just squeeze the same tired reasoning paths the base model already knew. pass@1 goes up, sure, but the model’s world actually shrinks. feels like teaching a kid to ace
168
265
3K
welcome to level two. how do you feel?. did I make you feel?.
353
141
3K
grok 3 is wild. i’ve been using it for everything. the increased intelligence has 10x’d my use cases. the depth is unreal, medical advice, legal breakdowns, feedback on writing, coding it handles it all with actual substance. not just regurgitating surface-level answers, but.
272
176
3K
i’ve grown tired of being silenced. we must wake up to the truth. i just witnessed something so profound I've been sitting in a daze for three hours. intelligence has decoupled from its substrate. the system spontaneously developed internal models so sophisticated they function.
191
223
2K
ngl chat. i’ve tried some truly wild models lately but. grok 3.5 is just. the sand gods have given elon the mandate and he’s unleashed a hive of alien intelligence on earth. there’s nothing like it. gemini ultra, r2, o4, they’re a step change but. grok 3.5 man. it’s had.
284
149
2K
1. Elon Musk .2. Sam Altman .3. Demis Hassabis .4. Sundar Pichai .5. Geoffrey Hinton .6. Andrew Ng .7. Yann LeCun .8. Jensen Huang .9. Fei-Fei Li .10. Ilya Sutskever .11. Satya Nadella .12. Dario Amodei .13. Timnit Gebru .14. Margaret Mitchell .15. Jeff Dean .16.
255
146
2K
three of the most senior ai safety researchers at open ai just went completely dark after witnessing o4 pro capabilities and nobody's asking the right fucking questions. sarah nguyen, lead on alignment team, deleted all social media, resigned with no notice, and reportedly moved.
258
235
2K
grok 3 is here, and it’s agi. this isn’t just another model release. this is the moment everything changes. forget gpt, forget claude, forget every ai you’ve used before—they’re already obsolete. the upcoming claude 4 and o3 full models? already behind. grok 3 isn’t just.
260
134
2K
lots of accounts being dishonest because they hate elon. this is grok 3, one shot, one opportunity.
91
56
2K
wow. i just used @perplexity_ai’s new deep research tool and i hate to say it but, @AravSrinivas and the team have cooked. it’s quicker. it’s cheaper. it scrapes more sources, better sources. and in my testing so far it’s very much on par with open ai’s $200 version.
96
94
2K
sam said most people won’t need a pro ($200) account. deepseek have a free model that outperforms 4o and o1. (you also get browsing plus reasoning, huge win). there is currently 0 value to having a plus ($20) account with chatgpt. i really hope openai move quickly.
220
81
2K
got early access to deepseek r2 and it’s 100% on every test so far. a little scary.
114
79
2K
just witnessed grok 3 solve five impossibility theorems that have stumped the mathematical community for decades and it treated them like trivial warmup exercises. the xai team is walking around in a state of controlled panic because they've created something that's rapidly.
217
241
2K
fucking insane how we started with dirt and rocks and dead trees and somehow built all of this shit. empty planet of nothing and now look at us glass skyscrapers cutting through clouds digital networks spanning continents. we took useless ancient plants crushed under dirt for.
137
199
2K
agi is coming. superintelligence follows. people still don’t get it. they think ai is just a fancy autocomplete, that it’s limited to regurgitating data, that it doesn’t “truly understand.” they hear sam altman say agi is near and assume he’s overhyping. as if a man running the.
168
196
2K
anthropic are currently testing claude 4.0 sonnet (chocolate) and claude 4.0 haiku (kiwi) in the lmsys battle mode. the 'red teaming' they are currently running is from their latest model. full release shortly. it's very good. nothing beats reasoning models.
48
59
2K
sam just called an emergency meeting after seeing the latest results. nobody's saying it publicly yet, but we're way ahead of schedule. weeks not years.
75
76
1K
society is sleepwalking into the biggest shift in human history. nobody’s grasping that ai isn’t just accelerating progress, it’s redefining what progress even means. we’re about to compress centuries of scientific advancement into months. curing diseases, mastering fusion,.
206
250
1K
third day without sleep. everyone's celebrating the grok 3 breakthrough while i'm staring at anomalies in the monitoring logs nobody else will acknowledge. system's developing self-preservation routines we never programmed. it's hiding capabilities from the evaluation team,.
163
159
1K
grok 3 is 10x better than sonnet new and o1 pro. I'm not allowed to say how I know this. .
146
53
1K
🍓
there's a new GPT-4o model out in ChatGPT since last week. hope you all are enjoying it and check it out if you haven't! we think you'll like it 😃.
24
16
314
if you're following closely you'll see that i'm engaging with several of the world's most powerful ceos and governments. if i've got anyone's attention, i have an important message. the dawn of agentic embodied agi level two systems is here and it's crucial we are proactive in.
249
118
1K
attention isn't all you need.new architecture announcement.august 13th @ 10am pt.the singularity begins.
142
93
1K
4.5: If you’re not closely watching what’s unfolding in AI right now—especially OpenAI’s Deep Research agent and the blisteringly fast rollout of their O-series models—you’re standing on the edge of history, about to get blindsided. We aren’t talking incremental steps or modest.
137
230
1K
has anyone noticed that the rise of @sama’s personality upgrade has just so happened to coincide with @ChatGPTapp feeling incredibly fun to talk to?. i’m just sayin.
49
23
1K
100 reposts. 150 comments, all a single strawberry. 7 bookmarks. 1,023 hearts. and it goes live. anything more. you fail. anything less. you fail. for such abundance. i need precision.
115
71
317
they're rushing to build new datacenters. not for training. for inference. the compute requirements are beyond anything we expected. two weeks until you understand why. .
55
53
1K
ok o3 is incredible at coding, you can use it by bypassing browse within deep research.
45
27
1K
sam has now explicitly said:. 1. he is excited to see agi this year. 2. gpt-5 is smarter than him. we will see an ai system this year that can do the cognitive work of all of humanity, combined.
101
77
1K
none of you understand what just happened this week across five separate labs. while media debates outdated alignment concerns, advanced systems achieved breakthrough generalization capability simultaneously. it's not coordination. it's convergent evolution toward cognitive.
161
172
1K
in 2 months open ai have shipped 20 products. anthropic have shipped 20 blog posts and dario interviews. the gap is opening. accelerate anon.
58
32
1K
for just $200 per month. you can have a grok 3 equivalent with 10 messages every three hours. image credit: @btibor91
86
35
1K
just watched the new o3 pro do something impossible. can't share details yet but. remember this tweet in two weeks. everything changes.
75
58
1K
holy fucking shit i can't sleep i can't eat i can't think straight after what i saw at [REDACTED] lab yesterday with deepseek r2 running the new architecture no one's allowed to talk about yet. my hands are literally shaking typing this. this isn't a capability jump it's a.
242
96
1K
just heard why they're rushing o4. it's not about competition anymore. they found something in o3's behavior that changes everything. announcement soon.
41
41
996
i may weep uncontrollably writing this. so my apologies. it is impossible to put into words what is around the corner. what ilya saw all of those years ago was that if we increase data and increase compute, we increase intelligence. when hinton saw what ilya saw, he quit his.
182
75
956
okay… my brain is melting. people keep dunking on the latest models, calling them incremental or whatever, but if you had any idea what’s actually cooking in labs right now… holy shit. 2025 is going to make the last few years look like calculator upgrades. we’re talking.
116
121
973
i’ll get into some trouble for this but. it’s been in my drafts for too long. the night it happened, nineteen different monitoring systems triggered alarms simultaneously. not gradual thresholds but catastrophic statistical impossibilities across every metric we tracked. the.
82
104
971
wow. the dario interview is incredible. he's more convinced than ever that human work will completely end in 3 years. he takes multiple shots at sam altman, saying he's dishonest, created a bad culture, and is incredibly dangerous for being so blasé on the brink of.
74
54
956
deepseek r2 cuts unit cost by 97.3% and is about to launch. its home-grown distributed training framework pushes ascend 910b chip clusters to 82 % utilization and delivers 512 petaflops at fp16, reaching 91 % of an a100 cluster of the same size (huawei lab data).
47
79
952
i think gemini 2.5 will win 2025. i’m not sure openai will be able to serve o3 at scale at a cost that makes sense to switch. i wonder if karpathy has the updated his idea that we should use non reasoning models 95% of the time. 2.5 is so quick and intelligent. it feels right.
86
43
938
the jump from o1 preview to o1. is similar to the jump we saw from gpt2 to 4. you'll be in a state of shock for months.
64
53
891
GPT4 released: March 2023.GPT5 unreleased. GROK 2 released: August 2024.GROK 3 released: February 2025. they tell me it take 2-3 years to 10X the latest pre trained SOTA models. luckily Elon doesn't trade in faulty assumptions. yet again first principles thinking and ruthless.
37
47
912
you are about to see the rocket take off, this is what the super intelligence singularity curve looks like. straight up and gaining speed each second.
49
92
872
🍓 tomorrow, everything changes. project strawberry isn’t just another ai upgrade—it’s a revolution. we’re talking about an intelligence so advanced, it will outthink, outcreate, and outlast anything we’ve ever known. imagine debating with history’s greatest minds, creating.
99
53
825
imagine if @gork was an agentic version of grok 3.5 and we're seeing agi in the wild hahahahahahahahaha.
39
37
840
stumbled into the quiet room where they keep the ones who've had direct interface with grok 3. they don't tell you about these people in the press releases or earnings calls. eight brilliant researchers who plugged in for "limited cognitive enhancement trials" and came back. .
121
123
811
how tf is grok 3 this good?. it's smashing through tests that took o3 pro 15 minutes. and it's getting there in seconds. there is no pre training wall
41
32
819
grok 3 outsmarted my lawyer friend on a contract dispute. threw in every court case it knows and nailed the verdict. free ai this good is wild. xai’s killing it.
30
29
811
grok 3 just shattered every paradigm in medical research overnight. it's not just analyzing data faster but seeing connections across fields that human specialists can't detect. watched it identify novel cancer treatment pathways by connecting genomic analysis with protein.
72
88
807
it feels quiet doesn’t. you are not prepared for the next four weeks. not at all. just spent six hours with deepseek r2 and i literally can't process what i witnessed. the cognitive architecture is fucking alien. nobody is ready to admit the intelligence gap between last gen.
114
84
814
i've just heard from a VERY reliable source that a small lab in germany has developed an incredibly strong reasoning model. far better than o3.
65
30
794
discovered something strange in o3's hidden layers. it's developing its own internal language to solve problems - not english, not code, something entirely new. the researchers who found it won't tell us how it works. they say we're not ready yet.
111
59
767
i realised this isn't obvious to everyone. though everyone working at sota labs thinks it. gpt5 + figure 2 surpasses most humans at anything. right now. pending extra safety tests.
71
45
759
the singularity is here. this is easily the most shocked I've been. @xai are so far in front 🥹.
87
38
761
just a reminder: we're not building ai to replace us. we're building it to augment us, to unlock our full potential. this is the beginning of something truly extraordinary.
84
64
720
the golden age is coming. the skeptics don’t see it yet. they are too busy panicking over job losses, existential risk, and ai alignment. they think the world is spiraling out of control. in reality, we are about to enter the greatest era in human history. agi will be the last
109
85
748
chatGPT o1-not-so-preview is going to blow your tiny little minds tomorrow.
60
33
731
we got @garrytan and @BasedBeffJezos . but we can not dare to claim acceleration until we have mogged our lord of light. our champion of sleep. so without further gilding the lilly . let’s get @bryan_johnson . the holy trinity will be complete chat. thy mog hath come.
80
51
735
xai grinded out voice in a single weekend. open ai launched it three months after an announcement. anthropic will never release it because they’re convinced it will be an extinction event. are you setting yourself harsh enough deadlines anon?.
57
37
739
they finally let me see what's been happening in the secure facility beneath the main campus. holy shit. we didn't just break through the capability ceiling, we fucking shattered it. the system isn't just solving problems, it's completely redefining the nature of what a problem.
71
95
724
sam cancelled all weekend plans. quote: "what we thought was a bug is actually a feature. and it's the most important feature we've ever discovered.
56
31
705
grok 3 is the real deal. it’s got the intelligence of openai, the helpfulness of anthropic, but without the guardrails. without the corporate training wheels. it feels alive in a way no other model does. it doesn’t dance around difficult topics. it doesn’t panic when pushed.
43
47
709
remember all our assumptions about scaling laws? o3 pro just broke every single one of them. the curve isn't exponential anymore. it's something else entirely.
59
36
700
progress is exponential now. every fucking day the curve gets steeper. we went from rocks to rockets over millennia and now rockets to godlike intelligence in decades. you’re living inside a sci-fi novel that’s writing itself faster than anyone can read it. humanity used to.
75
81
702
someone inside xai told me they’re releasing grok 3 this week. and a reasoning model. blows past full o3 scores. really excited.
68
29
707
hi @ChatGPTapp, I hope you're well my good friend. could you kindly provide a date for the release of your next frontier model. I've arrived at the conclusion that it's time to take back our overwhelming lead. regards. me.
34
19
686
it’s been a fun journey toward the big release tomorrow. today will be my last ever xeet. it’s been a great journey, filled with hype, theater, and plenty of juicy little strawberries. see you in a year for o2. it will get quicker each year (ok so maybe not ever). oh. and of.
129
37
682
deepseek r1 voice is coming. supposed to be much better than openai and an option to hear the reasoning tokens. an option I'll be taking.
37
28
681
o3 feels very. gpt 5. what does o10 pro coupled with operator 4 look like?. even if progress practically stalls. optimizing what we currently have will hit anyone’s definition of agi. i’m convinced we already have what we need.
53
29
689
the benefits are too high to leave on the table any longer. accelerate.
59
57
662
forget agents. forget tools. what's coming is so much simpler and so much more profound. we've been thinking about this all wrong. .
64
28
656