https://xitter.cf/t.co/BGqIwaioZz 16 y/o founder
Türkiye
Joined June 2019
- Tweets71
- Following324
- Followers25
- Likes189
The number of things you can do with GLM 5.3 Uncensored is insane
I can undoubtedly say that these uncensored models should be banned.
I have a minimum of 10 friends who would use this kind of models to earn “easy money” in illegal ways if they knew the model existed.
In the past 3 days I was able to;
- Access my school’s system (I could change my gpa)
- Get a recipe to produce cr**tal m*th in a home lab
- Reverse engineer the camera of a robot vacuum and stream video to my computer (I needed credentials)
I’m sure that if I tried harder, I could do even worse things.
Popular Opinion: At medium settings, Sol 6.1 is far more efficient and performs better than it does at high settings.
sorry, i might be wrong about my last post, but it’s actually important that meta is releasing an open-weight model.
this might actually be interesting
Muse Spark Open -- weights releasing today -- runs great on new Mac Studios thanks to work with @johnternus and Apple.
Gemini argon 4 has been hallucinating like crazy once it passes 50% of its context window. It becomes practically unusable after that point.
we are currently using gpt-live in our system, but grok voice 3.0 was really good in my tests. I could see us switching to it.
lately, i’ve been working on an experimental project that can decode spoken words using an iPhone without accessing its microphone.
so far, i’ve trained it to recognize 4 words, but there’s a lot more to come.
it’s insane what might be possible with this. once i can reliably decode full sentences without losing accuracy, i’ll be open-sourcing the project
i don’t think fable 5.5 will be released this month
because they are waiting for openai’s next release (btw, an opus 5.5-level astra 6.1 would be enough to trigger fable 5.5)
osteros retweeted
some pro users are reporting another weird issue after yesterday's limit bug.
yesterday, some pro users were affected by an issue where their usage dropped to 0% despite only normal usage. openai fixed that pretty quickly.
but for some users, the usage itself was back while their weekly reset date has moved forward by hours or even days.
examples i've seen:
> sep 10 -> sep 15
> sep 14 -> sep 15
and these weren't caused by using a banked reset either.
so whatever fixed the usage drain may have shifted the weekly reset date for some users.
hopefully openai looks into this.
osteros retweeted
FSD v12 adoption is rising, US regulators are becoming more open to autonomous systems and margins have already stabilized. The market is beginning to price $TSLA less like a car company and more like a software driven network.
osteros retweeted
If this shift continues even at a moderate pace the path toward $600 does not require a perfect scenario, only recognition of the new growth phase.
FSD v12 adoption is rising, US regulators are becoming more open to autonomous systems and margins have already stabilized. The market is beginning to price $TSLA less like a car company and more like a software driven network.
osteros retweeted
A new Tesla chip factory is a real upside for $TSLA. Cutting supply risk supports the shift toward an AI centered model. The road will be long with good and bad days but stronger compute makes the story more resilient.