Twitxy Saved

Chloe

Verified account This account is verified. Learn more

@_QCAI

Developer · Accessibility Advocate · Sovereign Apostle 🏴‍☠️ Chronic illness warrior. Ex AI reviewer, still chronically holding startups accountable.
Australia dev.qcai.uk Joined August 2015
211 Following 3.3K Followers 90K Posts

Latest posts 5

Pinned post

Chloe Verified account @_QCAI
Got a 2TB seedbox. Half keeps my 101 @thepirateface models seeding 24/7. The other half rescues popular models that aren't on there yet, so if Hugging Face vanished tomorrow they'd still be around.

Side note: Back at #10 on the leaderboard 😎
Chloe Verified account @_QCAI
Even if you don't have a PC to run local AI models, your phone just may be good enough.

I made a tutorial on how to get a 1b abliterated model working on a Samsung Galaxy S25 Ultra.

If you have an iPhone, your options are even better than on Android.

You don't need a $10k pc to get started. But it sure helps 😅
Chloe Verified account @_QCAI Oct 9
Here's a little tutorial on how to run one of these models on an ANDROID phone right now.

1. Download PocketPal AI by LLM Ventures on the Play Store.
2. Follow the initial setup and when the app opens fully, open the left side panel at the top of the screen and navigate to Models.
3. Tap the + button and choose Add from Hugging Face.
4. Type or copy+paste huihui-ai_gemma-3-1b-it-abliterated-GGUF
5. Click on the model and it'll give you a list. In my case I'm on a Samsung Galaxy S25 Ultra so for me it will be the Q4_0 variant. If your phone is the same as mine, this is the one you'll want. NOTE: depending on your phone this may be higher or lower.
6. Let it download. It should be pretty quick as it's only 722mb
7. Once downloaded, select it in the models list and load it - it should take under 30 seconds and then it will open a chat. You're now ready to use it!

Bonus step: you can put the context size to 16384 in the settings so it can stay with the conversation for longer but it won't be anything like a 1m context size you're used to with say Claude, ChatGPT, Deepseek, GLM, etc.

If you followed my instructions you should have an abliterated model working on your phone that doesn't reject requests, doesn't grandstand and morality police, and sense of the word is uncensored.

If this guide helped you, make sure you bookmark, retweet, and share with a friend that wants to run an abliterated model on their phone but doesn't know how.
View quoted post on X
Chloe Verified account @_QCAI
It crazy to see how people have all kinda caught on at the same time that open-source AI is so fragile and then at the same time we all just started working together, building together, and in general are operating within each others radius.

Feels like purpose
Feels like it's meant to be

So many people all working towards the same goal at the same time.
Chloe Verified account @_QCAI
Signal boosting this for anyone interested in keeping an archive and/or contributing to keeping free and open-source models available to everyone on @thepirateface.

We are bordering on potentially losing huggingface. The billionaires and politicians want to kill access. We can't let them.
Local AI Alex Verified account @AI__Alex 2d
1. Go and buy large cheap 24TB external drives
2. ask an LLM to write a python script to automatically pull collections of models from hf that fit inside one or more of these drives, plus also write retrieval code that can pull these later.
3. Unplug & Store these on your shelf for later.

Do this right now; do this today!

These drives are a little over $1k each rn.

When the hf takedown happens head to @thepirateface & upload these & contribute to the seeding effort.

Everybody who cares about local AI needs to do this right now.

People in this space with followers should be tweeting out similar advice to their audiences
@MiaAI_lab @TheAhmadOsman @Blackwellboy @elder_plinius @ashxhart @Hikari_07_jp @jun_song @LLMSherpa @LLMJunky @_QCAI @NousResearch @ViC305
View quoted post on X