Waiting for AGI Give the world some color Now that you're here, follow me. It's free
- Tweets 23,865
- Following 1,495
- Followers 11,743
- Likes 245,061
Things that didn't exist three months ago:
- o3
- o4-mini
- GPT-4.1
- Flux Kontext
- GPT Image 1 (4o image gen)
- Claude 4
- Llama 4
- Codex
- Veo 3
- Gemini 2.5 Pro
- Gemini 2.5 Flash
120
257
3,937
301,623
Things we had no idea about a year ago:
- GPT-4o
- Grok-2
- Claude 3.5
- Claude 3.5 (new)
- o1
- Gemini 2.0
- o3 & o3-mini
- Grok-3
- Claude 3.7 Sonnet
- GPT-4.5
- A lot more
THERE'S NO WALL
82
111
2,711
265,192
Things that didn't exist three months ago:
- GPT-5 (Instant, Thinking, Mini, Nano, Pro)
- GPT-5 Codex
- GPT-OSS (120b & 20b)
- Opus 4.1
- Genie 3
- Sora 2
- Sonnet 4.5
- Nano-banana
- Seedream 4.0
I like doing these from time to time to feel the acceleration
75
173
2,852
148,652
Conversation with a family member today:
+ ChatGPT is so dumb
- which model are you using?
+ idk, the paid one
- there are like 8 models, which one?
+ I just open it and use it
We need GPT-5 asap
84
29
2,304
142,266
No, OpenAI is not dead
o3 is 3 months old, they were the first company to implement tool use in the CoT, of course xAI made a "slightly" better model 3 months later copying the exact same technique
Grok-4 Heavy is just o3-pro
People don't get that companies catch up because OpenAI actually shows us what they did to achieve better models. They could've just said o3 just took longer to think and no one would've known it had tool use in the CoT, or even don't say reasoners existed at all
179
68
2,081
529,361
Anthropic has:
- The best coding model ❌
- The best image gen model ❌
- The best model for math ❌
- The best voice mode ❌
- The best chatbot UX ❌
- A reliable API ❌
- Great rate limits ❌
- Expensive models ✅
129
71
1,884
160,854
OpenAI fucks up: oh yeah sorry, our fault
xAI fucks up: oh yeah sorry, OpenAI's fault
19
35
1,729
51,749
I'm sorry but this is simply a lot of money for a screen recorder
How is there no good free alternative?
1,084
26
1,728
6,113,702
I wanted to publicly thank @sama and @aidan_mclau
I don’t know how to show my gratitude, so I figured a post to let you know how much helpful you’ve been wouldn’t hurt. Thanks to you, I got 2 months of ChatGPT Pro to use Deep Research for my rare disease, and in that time I’ve learned things even my cardiologists couldn’t tell me, it has improved my life a lot.
Now I’m back on the Plus plan (I've been a Plus user since it came out and I'll always be, the value is insane for the price), which luckily includes 10 Deep Research uses per month, that should be enough.
You’re saving lives, and I’ll always be grateful for what you’ve done for me. I’ve never seen a CEO as kind and approachable as Sam, or someone at a big company as committed to helping humanity as Aidan.
Thank you both so much. OpenAI is changing lives thanks to you and everyone behind the scenes. ❤️
By the way, Sam, if you happen to read this: happy early birthday for Tuesday! :)
28
68
1,569
159,256
Grok-3 summary:
- They rigged the benchmarks
- They said it was only available on Premium+
- They doubled the price of Premium+
- The next day, Premium users also got access
- Two days later, free users got access too
110
66
1,511
200,057
What we got:
- Operator
- o3-mini
- Deep Research
- Gemini 2.0 Flash
- Gemini 2.0 Pro exp
329 days to go
30
63
1,448
57,184
Seedream 4.0 vs Gemini 2.5 Flash image gen 🍌
Seedream 4.0 seems to have a better understanding of intent, and it also made a better edit in this case
"Remove all the ingredients from the burger and keep only the top and bottom buns. Leave a gap between them, keeping the same spacing as if the fillings were still inside."
25
96
1,435
204,002
Since Grok-3 released OpenAI has released:
- GPT-4.5
- GPT-4.1
- o3
- o4-mini
- GPT-image-1
- Codex
- Updates to existing models (like AVM and 4o personality)
- o3 80% price reduction
- o3-pro
xAI has shipped:
-
(Again because I forgot about some OpenAI releases)
68
33
1,290
105,743
Google: We have nano-banana, the best image model available
OpenAI: We have GPT-5, the best LLM available, unifying reasoning
xAI: We can make 12 fps low quality videos 🔥
39
32
1,244
77,792
It's true that Plus users lost usage with GPT-5
Before:
o3 - 200/week
o4-mini-high - 100/day
o4-mini - 300/day
4o - 80/3 hours
Theoretical max: +1240 messages in a day
Now:
GPT-5 - 80/3 hours
GPT-5 Thinking - 200/week
Theoretical max: 840 messages in a day
77
63
1,192
63,073
Just a reminder that ChatGPT Plus used to only offer 50 GPT-4 messages every 3 hours
49
25
1,036
78,783
How is it possible that a lab no one knew about was able to develop a better Voice Mode than OpenAI's AVM or Grok Voice?
103
28
1,003
136,037
Anthropic, you're not OpenAI, you're not years ahead of anyone
Release Opus 3.5, stop the blog posts
We've published a significant update to our Responsible Scaling Policy, which matches safety and security measures to an AI model's capabilities.
Read more here: anthropic.com/news/announcin…
31
24
964
108,645
Almost 2 months since the release of Grok-3:
- Still no API.
- No Android app in Europe.
- Most features are not available in the Android app.
- AVM is bad, and only works in English.
- They charge more than ChatGPT, Claude or Gemini for a worse model and less features.
111
36
973
76,433
Considering we have 350 per week for o1-mini and o3-mini is cheaper, 100 per week is really, really low
14
2
966
28,020
We might be getting:
- GPT-5
- OpenAI's open-source model
- Claude Opus 4.1
- Gemini 3
All this week
30
51
980
62,884
ChatGPT Atlas
"Now available globally on macOS. Windows, iOS, and Android are coming soon."
43
31
994
125,474
Gemini 2: end of January
o3-mini: end of January
Grok-3: December
It's January 30th, and none of them have been released
50
23
934
60,322
Codex IDE extension feels awesome now
I'm not paying for Cursor again, this is so much better
87
31
922
99,979
Next week we should get:
- GPT-4.5
- Claude Reasoner (Sonnet 4 or Sonnet 3.5 with TTC)
- Grok-3 updates
- More?
35
39
805
117,889
GPT-4.5 IS NOT A REASONING MODEL
It's not made for coding or math. It's for creativity and writing, stop complaining, what did you all expect?
131
27
810
69,658
OpenAI has delivered:
- GPT-4o
- GPT-4o mini
- Advanced Voice Mode
- o1
- o1 Pro
- o1-mini
- o3-mini
- Deep Research
- Operator
- Canvas
- SearchGPT
- Sora
- A lot of 4o improvements
- Memory
- Tasks
- Projects
- and more
They're the kings of shipping
56
34
802
46,209
> "ChatGPT sucks"
- Are you using the model router?
> "No, I only use instant, I hate it when it starts to think"
- 😐
27
11
815
55,466
What's strange is that we had o1 leaks, we had o3 leaks, we had GPT-4.1 leaks but there aren't any GPT-5 leaks
61
11
772
61,588
Logan joining Google is one of the best things that could’ve happened to the company
I’m not exaggerating
46
6
780
90,124
o1 JUST FUCKING SOLVED A CODE ISSUE I HAD FOR A WHOLE MONTH
4O AND SONNET 3.5 COULDN'T
36
19
698
209,653
GPT-5 Thinking
I like revisiting this prompt every now and then to see how much better models have gotten over time
31
27
713
145,615
People really think OpenAI is dead hahaha
GPT-5 will be much better than Grok-4, let that sink in :)
73
25
691
68,728
According to rumors, OpenAI is releasing something next week, for all users (free, Plus, Pro and devs)
55
13
660
101,776
I like that Google (especially Logan) is taking the Gemini 3.0 leaks and checkpoints seriously
It shows that they care a lot about this release
6
11
666
171,894
"As a result, we've increased pricing for Claude 3.5 Haiku to reflect its increase in intelligence" lmao
During final testing, Haiku surpassed Claude 3 Opus, our previous flagship model, on many benchmarks—at a fraction of the cost.
As a result, we've increased pricing for Claude 3.5 Haiku to reflect its increase in intelligence: anthropic.com/claude/haiku.
36
14
599
52,884
GUYS I GOT A LEAK FROM OPENAI!!!
This is their plan for this year:
- Voice alpha (probably like 0.5% of users)
- Security blog post
- Partnership
- Partnership
- Security blog post
- GPT-4o mini small
- GPT-4o ultra small
- GPT-4.1 (1% better than 4o)
29
18
577
40,569
It amuses me how Microsoft always presents Copilot features as if they were genuinely theirs and not just a GPT wrapper
31
14
575
53,796
OpenAI: $200, unlimited o3-pro
Google: $250, 10 messages per 12 hours 2.5 Pro Deep Think
Grok: $300, 20 messages per hour Grok 4 Heavy
Anthropic: $200, 24 to 40 hours a week of Opus 4
Pretty clear which option is the best, the only one that's really unlimited
I’ve learned that Gemini DeepThink is limited to 10 prompts per 12 hours. I’ve decided not to subscribe to Ultra for this reason at this time, as I go through that many prompts in less than an hour. That’s extremely limiting and not even worth trying, at least for me…
71
27
567
80,723
People: GPT-5 will exceed our expectations 🔥
Also people: I expect GPT-5 to generate text, video, audio and images, also I want it to have an infinite context window, get 4000 on Codeforces, have any input modality, use any tool possible, think for like 2 hours straight if needed, have a natural Voice Mode that's proactive and code its own UI. Oh and it should cost less than GPT-4.1 nano
43
24
559
81,895
I'm gonna say it: Grok-3.5 will get released, xAI will show amazing results, people will hype up the model and then no one will use it and o3 will still be number one
77
10
507
56,535
AGI (a little patience please) ❌
GPT-5 ❌(yet)
better voice mode ✅
higher rate limits ✅
better GPTs ✅
better reasoning ✅
control over degree of wokeness/behavior ❌
video ✅
personalization ✅
better browsing ✅
'sign in with openai' ❌
open source ✅
8 out of 12 requests, not bad
What will we request for next year? I'd want GPT-5 (if it's not released by then), agents, price cuts on the Realtime API, o2 and a new hardware device
thanks a lot for these! some common requests:
AGI (a little patience please)
GPT-5
better voice mode
higher rate limits
better GPTs
better reasoning
control over degree of wokeness/behavior
video
personalization
better browsing
'sign in with openai'
open source
38
17
478
91,914
I left 4o mini code overnight, look at the results!
1st iterations vs 329th iteration
30
22
491
101,088
Suno v5 released
"What you're getting with v5:
- The clearest audio we've ever produced
- Vocals that sound genuinely human
- Way more creative control over every element
- Better genre understanding and mixing"
25
25
512
37,834
The Sora app feed is so boring now that they refuse to generate copyrighted content
50
10
489
38,533
OpenAI employees are back posting on X, that means GPT-5 and the open-source model are 100% ready to launch
26
12
492
27,205
Google: Here’s our new reasoning model!
OpenAI: You can now call ChatGPT from a 40-year-old phone!
15
8
471
19,496
I actually think we might get GPT-5o next Monday on DevDay:
- OpenAI has been working on video input for a while in ChatGPT
- They need a new image gen model, 5o could be it
- They also said they were gonna improve AVM, 5o would also do that
- It'd follow the naming convention
29
13
482
40,731
GPT-5 will supposedly have a better AVM, and they just unrestricted 4o's AVM
The future is bright
cc: @chrisgpt
24
9
465
86,040
As Sam said, models should stop being huge databases
We should have a really small model that's really good at reasoning and managing context, the internet already exists, we don't need to internalize it in the models
57
17
468
28,295
OpenAI, I'm your fan, but I gotta admit the livestreams are out of hand
Not every feature needs a livestream, making them so often makes them feel less special, and more boring
Please go back to blog posts for small things, and make bigger and better livestreams like you used to do (Spring update, 2023 DevDay, 4o image gen, etc)
25
3
464
31,312
The next big releases expected from each company:
OpenAI: Images v2
Google: Gemini 3
Anthropic: Claude 4.5
xAI: Grok 4.1 (Grok 5 starts training soon)
Christmas is going to be such a special time
22
15
465
22,806
This man will do anything but release Grok-4
By a factor of 2 to 1, you want a new political party and you shall have it!
When it comes to bankrupting our country with waste & graft, we live in a one-party system, not a democracy.
Today, the America Party is formed to give you back your freedom.
28
8
437
25,059
No way! Grok-4 uses tools while reasoning? How did no other company think of that...
22
8
435
41,634
I need these plushies
Animal Plush Toys by SeeDream V4 on @BasedLabsAI 🧸✨
Prompt👇🏻
![A soft, high-quality plush toy of a [ANIMAL], with an oversized head, small body, and stubby limbs. It’s made of fuzzy fabric, has embroidered facial features, and is shown sitting or standing against a neutral background. The expression is cute or expressive, and it has natural animal markings or colors. The lighting is soft and even, with a realistic, collectible toy look. Centered, full-body view. 1080x1080 dimension.](/pic/orig/media%2FG1SVwAcXUAA0l9W.jpg)
ALT A soft, high-quality plush toy of a [ANIMAL], with an oversized head, small body, and stubby limbs. It’s made of fuzzy fabric, has embroidered facial features, and is shown sitting or standing against a neutral background. The expression is cute or expressive, and it has natural animal markings or colors. The lighting is soft and even, with a realistic, collectible toy look. Centered, full-body view. 1080x1080 dimension.
19
36
434
35,219
Sora 2 is $0.1/sec
Sora 2 Pro is $0.3/sec
> Copyrighted characters and copyrighted music will be rejected.
> Real people—including public figures—cannot be generated.
> Input images with faces of humans are currently rejected.
46
22
416
46,066
Sam now that you're here, are Plus users ever getting more than 32k context? Even if it means less usage
14
2
422
19,224
Things we had no idea about a year ago (updated):
- GPT-4o
- Grok-2
- Claude 3.5
- Claude 3.5 (new)
- o1 and o1-mini
- Gemini 2.0
- o3 & o3-mini
- Grok-3
- Claude 3.7 Sonnet
- GPT-4.5
- Gemini 2.5 Pro
- Gemini 2.5 Flash
- GPT-4.1
- o4-mini
23
23
395
27,351
Gemini-3.0-Ultra spotted
As testing said, it might be just a placeholder
BREAKING 🚨: gemini-3.0-ultra spotted in Google’s Gemini CLI repo, committed 4 days ago!
First public proof of Ultra. Beta in October? @lmarena_ai @testingcatalog @AIExplainedYT
16
19
400
28,388