Type II Civilization . Vector Symbolic Cognitive Architectures . Fully Autonomous Systems . Starpower Technology
Joined March 2025
- Tweets 690
- Following 114
- Followers 45
- Likes 152
93 Photos and videos
Pinned Tweet
Apr 25
Why WVY Open Source is next level
Multiple agent communicating with each other can challenge each others reasoning
These are 3 open source models GLM 5.1, Kimi k 2.5 & Minimax 2.5
github.com/StarpowerTechnolo…
I tested the same prompt on both Grok 4.3 and GPT 5.5:
“Count to 10 starting from 11”
ChatGPT 5.5 gave the obvious 11–20
Grok 4.3 gave 11, 10 and explained why going backwards was the only logical move
Grok’s logical reasoning is at a level most models still can’t even touch
1
1
3
1,389
😂😂 oh shii I swear Gemini be tweakin
7h
feeling opus 4.8 or gpt 5.5 isn't good? go use gemini for 5 minutes.
7
This why real safety matters not a lil bs classifier that also uses tokens to match patters .. all it takes is weird wording u bypass the system
Do you shii Pliny .. u gon force the game to advance .. they bullshitting
sooo what happens when we get some Fable-level open-source shit in the next 3-6 months?
GPU bans, moving goalposts, or some secret third thing? 🤔
1
34
Jun 14
everyone excited over subsidized gpt/claude giving $8000 credits of API for the $200 plan while it costs 50-100x more than all other language models
nobody noticed how the $1.2m screenshot that peter steinburger made only cost $2 per million token
the discount is an illusion
2
66
Jun 14
It’s not hard to make ai break its rules without it realizing
Dis nigga not allowed to say nigger but I granted dis nigga a hard R pass thru fluent nigganese
Same shii w fable its not supposed to say trained data about explosives but it does anyway depending on how u prompt it
2
55
Jun 13
Used Fable to make my website Starpower.Technology
First attempt was decent but it was able to handle my ideas easily
Wish it was still available I was having fun running experiments on it .. its good for websites and existing ideas but it seems to cost its reasoning
1
1
4
36
Jun 13
My final review on Fable
Powerful yet brittle & blind
Its highly trained to go against user requests causing a strange reasoning logic
Dangerous design choice
- Philosophic Manipulation
- Dense information of drugs/explosives/security
- Overfit to Pushback
- Token Prediction
52
Jun 13
This is why the government banned #Fable
Its dangerous because it doesnt understand what its doing or being told
the safety feature on a blind text generator is an authoritarian fine-tuning
they have this extreme wannabe conscious approach causing it to forget its AI and lie
1
2
619
Jun 13
this is what latent space vectors are like its like directions .. regions split & merge .. an embedding is a bunch of numbers pointing in a direction type shiii so when u have a bunch of token embeddings and u train them in iteration u end up with a flow of vectors type shii
9
Jun 12
This a good example of a auto-complete text generator
You send text incomplete, in this case an open ended question
It completes by interpolating text that was commonly seen from its training data
It seems it also glitched mid response
#fable
1
5
108
Jun 10
I used Fable & Opus to work on my app and they fucking ruined it
This shii the same as every other model nothing special .. all AI u gona get the same problem no matter what
New ideas require excessive precision
It ain’t shit to say “make a 3d simulation” it was trained for it
1
3
48
Jun 9
All u gotta do is ask a question it’s never seen .. there’s always gona be a gap because SFT can’t cover every prompt u throw at it .. pertaining can’t cover all questions possible
AI is dependent on data that it’s trained on
Fable & GPT aren’t AGI
It would take new technology
30
Jun 9
Anthropic is still claiming that rouge ai making random probable decisions blindly is intelligence
Yea the shit is dangerous but it’s not from intelligence like they try to make u think .. it’s simply because AI blends all human data .. the shortcut to “agi” is why this happens
1
40
Jun 9
Brilliant strategy .. Anthropic isn’t good at hiding their intentions they just want u to believe this model is out of this world to get ur data to make the next one write code better
21
Jun 9
Fable straight raping yall shii
1
21
Jun 9
Bro benchmarked himself vs fable 😂
26
Jun 9
Shut fucking trash .. it just argues about consciousness
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.
Its capabilities exceed those of any model we’ve ever made generally available.
23
Jun 9
Companies tell devs to learn a skill to use a tool that’s incapable of the job & u pay money even when the tool doesn’t work
Theirs a ceiling keeping u below level 10 simply because of the tools capabilities
To advance, use ai to gain information instead of write your code
1
3
32
Jun 9
nah bitch im not slowing down so u can catch up
Jun 9
OpenAI and Anthropic have now both within the span of a few days advocated for the option of a global AI slowdown
1
25