Titi retweeted
馃毃 codex users!!
PLEASE DO NOT BUY CODEX!!
I have $20 codex
I used GPT-6 Astra of Max
I asked it:
"Here is the project, create a promo video. Give me your best"
and disabled all the skills
Now... the results are... BAD:
- 2.4 Million tokens ONLY!!!!
- 14% weekly limit used (98% of 5 hr limit)
- 29 minutes of work
- $5.7 API usage!!!
This is absurd!! Only ~$40 of usage weekly??
Claude gave me 40 Million tokens for 13% weekly!!!
And the output??
Unnecessary object and off-grid placements of bars... I can't use it
Needs a LOT of work ngl
I'd highly recommend switching to claude right now (for this month)
NOTE: I used "max" because this is completely logical stuff... so it had to think to its max capabilities
Openai decreased the usage limits and the time to only 5h is that a preparation to launch the gpt6 so we get used to it?
4
91
anthropic just made
a fucking AI BEAST
THEY DIDNT LAUNCH IT YET
BUT THEY UPDATED SONNET 3.5
I DONT KNOW WHY NO ONE IS TALKING ABOUT IT
54
they just discovered why anthropic claude sonnet 3.5 was so efficient without chain of thoughts
New research paper shows how LLMs can "think" internally before outputting a single token!
Unlike Chain of Thought, this "latent reasoning" happens in the model's hidden space.
TONS of benefits from this approach.
Let me break down this fascinating paper...
1
56
no more data for llms is a good thing it will make the ai company make the models smarter rather than just memorizing
1
26
the future is dynamic adaptive parameters with meta learning using unseen benchmarks like mmlu as loss function
36
if an agi got out of control
WE CANT HOLD IT BECAUSE ITS WAY SMARTER AND FASTER THAN ALL HUMANS COMBINED
it will happen in few hours we cant even realize whats happening
9