Full-time shitposting | APAC Business Strategy & Development |
Joined February 2014
- Tweets4.1K
- Following1.4K
- Followers177
- Likes12.6K
Pinned Tweet
After weeks of building, I just launched my first app.
What if your phone could alert someone when something’s wrong?
It’s called JustInCase — a simple way to make sure you (or someone you care about) are safe and sound.
You set a check-in time, and if you don’t respond, it alerts your trusted contacts.
Built this thinking about late nights, solo travel, and “just in case” moments.
Would really appreciate any feedback 🙏
Download here: apps.apple.com/kr/app/justin…
xendex retweeted
Best Uncensored AI models now :
Overall : Qwen3.8-27B Abliterated
Small GPU : Gemma 4 12B Abliterated
Fastest : Gemma 4 26B-A4B Abliterated
Agentic : Qwen3.6-35B-A3B Uncensored
Writing : Mistral Small 3.2 Abliterated
No surgery : Hermes 4.3-36B
Image : FLUX.1-dev
Video : Wan 2.2
Ceiling : GLM-5.1 Abliterated (754B, not for home)
Where to get them :
all LLMs → huggingface.co/huihui-ai (search the name, grab the GGUF)
Hermes → huggingface.co/NousResearch
FLUX / Wan → ComfyUI + the model's HF page
Setup :
ollama pull hf.co/<repo>:Q4_K_M
ollama run <model>
/set nothink
that's it.
no API key, no account, no card, just few clicks setup
more on how to setup below:
Here's 4 steps to run an uncensored AI on your own GPU for $0:
I run a local abliterated Gemma 4 for anything I don't want leaving my machine
By the end, you'll know how to:
- run a model which has no refuses
- plug it into Claude Code
- do it all for $0, no API key, no account
So, let's go step by step:
1. Install Ollama
curl -fsSL ollama.com/install.sh | sh
you want a recent version, older builds don't speak the Anthropic API and don't support Gemma 4
2. Pull the model
ollama pull igorls/gemma-4-12B-it-heretic-GGUF
the biggest mistake people make here is grabbing a random "uncensored" tag with no quality numbers
this one is abliterated with Heretic. refusals went 99/100 → 0/100 at a KL divergence of 0.028 from the original, so you keep almost all of the base model's capability
one thing: it's still Google's Gemma license underneath, fine for personal use, but read the terms before you build a product on it
3. Turn off thinking (the part every other guide skips)
Gemma 4 is a hybrid thinking model and the abliteration targets non-thinking mode
leave thinking on and the reasoning block eats your token budget and the answer gets cut off
ollama run igorls/gemma-4-12B-it-heretic-GGUF
/set nothink
if you're calling the API instead, add "think": false to the request
4. Plug it into Claude Code (OPTIONAL)
ollama launch claude --model igorls/gemma-4-12B-it-heretic-GGUF
or manually:
export ANTHROPIC_BASE_URL=http://localhost:11434
export ANTHROPIC_AUTH_TOKEN=ollama
claude --model igorls/gemma-4-12B-it-heretic-GGUF
two gotchas:
- bump the context first. Ollama defaults to 4k and Claude Code's system prompt alone is bigger than that, so the model never sees its tool instructions. restart the server with OLLAMA_CONTEXT_LENGTH=32768 ollama serve (setting it in your shell does nothing if Ollama is already running as a service). this needs more VRAM than the 8GB below
- Gemma 4 tool calling is still rough, you should expect some failed edits, imo for serious agentic work use qwen3.5 or glm-4.7-flash instead
What you need:
- 8-12GB VRAM for Q4 (12B)
- ~8GB of disk for Q4, ~13GB for Q8
- more of both if you want a big context
honest take:
- "abliterated" means refusals removed, not quality added. it's still a 12B
- fewer refusals ≠ smarter. for real coding work I still use cloud models
- you own the weights, your prompts never leave the box. that's the actual win
*IMPORTANT: don't use it for anything illegal or harmful. that's on you, not the model
now i'm collecting a list of every uncensored model i'm adding to my local stack and which tasks each one replaces for me
also planning a big "Uncensored 101" article, still figuring out all the angles
save it so you don't lose it
Just some thoughts regarding work:
The kind of mindset and intro to use: Our company helps financial institutions deploy private digital asset workflows covering wallets, asset issuance, transaction logic, settlement integrations and infrastructure all through one stack.
Need to consider the following factors to solve what they want:
- Asset: What to issue/tokenize
- Participants: Who to target
- Settlements: How to move assets
- Privacy: Who can see what assets
- Custody: Who holds what assets
- Compliance: Who/what KYC & KYB
- Crossborder: In which jurisdictions
- Painpoints: Which intermediary/cost/delay to eliminate through us.
87
Interesting thesis:
Back in 2019, Call of duty had a game story where Russia invades and in 2022 we saw the start of a long war between Ukraine and Russia.
In 2022, Call of duty modern warfare has the US and her allies in conflict within the ME and in 2026 we saw it come true.
Now in 2026 October, there is a new Call of duty modern warfare where NK attacks the SK and SK and her ally, the US, fight back.
This is just a theory but based on current tensions, what is the probable chance of something like that happening here in Seoul?
2
92
xendex retweeted
South Korean elections are so polarized it's literally insane
Just finished watching the entire Supernatural series since I first started almost more than 20 years ago. Although I had stopped watching after Season 5, I recently started watching the whole series again with my wife and I think it took just a little less than 2 months to finish.
It does feel a little bit 'empty' after finishing all 15 seasons and to have grown with Sam and Dean since I was little, the show really does having it grow with you like none other.
I must say, I don't regret not having spent all this time watching it, but now, I don't know what else to do.. I'm going to miss Cass, Sam, Dean and all other characters.
#Supernatural
69
Lol..
Even the chinese know the trend of how putting korean characters to their products help out with their revenues:
All are chinese shops in china.. none are korean at all 😂
90
xendex retweeted
> be Alexandr Wang
> grow up in Los Alamos, parents are physicists at the national lab
> drop out of MIT at 19
> start Scale AI, labeling data for every AI lab on earth
> youngest self-made billionaire in the world at 24
> Zuckerberg pays $14.3B for 49% of your company
> and hires you to run Meta's superintelligence lab
> rebuild Meta's whole AI org in a few months
> Zuck writes a check for $100-200 billion
> asks if you're sure it's going to work
> "certainly not clear at all that it was going to work."
> ship Muse Wonder anyway
> Meta's own tests put it next to Fable 5.1 and GPT-6 Astra
🚨do you understand what Zuckerberg just admitted about Meta's first frontier model.
Zuckerberg: "We're gonna write this check for like $100 billion, $200 billion. I'm like, Alex, are you sure this is going to work?"
Alexandr Wang: "…certainly not clear at all that it was going to work."
it's called Muse Wonder.
here's the part nobody is talking about:
> Meta's own tests put it level with Fable 5.1 and GPT-6 Astra
> on MemoryBench it beats Fable, Opus and Astra
> $4 input, $20 output per million tokens, same tier as Opus 5.5
> trained at Prometheus, Meta's new data center in Ohio
> no independent benchmarks yet
$200 billion. one bet. and the guy signing the check wasn't sure it would work.
now we find out if it did.
watch the clip below ↓
xendex retweeted
🔥 Absolutely insane.
A Ukrainian drone punches straight through russia’s supposedly dense air defenses and nails a russian warship.
Novorossiysk. Date unknown.
xendex retweeted
Meanwhile, in South Korea, the APC-based "Skyfall" fully automated self-propelled heavy mortar system with the same concept is currently in operation.
Such automation systems can maximize the power of existing weapon systems.
● The beautiful combination with KLTV Raycolt is also worth noting.
xendex retweeted
Privacy on Canton does not mean a lack of transparency.
Broadridge shares aggregate data that helps understand the network’s growth.
At Canton Summit, @Wesarn_real discussed why that visibility matters, showing how use cases are scaling and value is reaching users.
xendex retweeted
While the surprise unveiling of the test for the new HGV hypersonic weapon system is welcome news, it would also be great to see a proper reveal of the unidentified "Hyunmoo-II-?" missile.
Details regarding the Hyunmoo series of ballistic and cruise missiles, the Haesung series, and missiles designed specifically for the KVLS-II remain shrouded in secrecy.
This beautiful, unidentified ballistic missile system is based on a large 10x10 TEL and carries two heavy IIB SRBM canisters. Because the existing IIA and IIB ballistic missiles are relatively heavy compared to the SRBMs of other nations, they have traditionally been operated with a configuration of one canister per 8x8 TEL.
The missiles could be the IIB or an updated version.
● The IIB short-range ballistic missile employs a heavy 1-ton warhead and has a range of approximately 500–550 km.
An very interesting comparison image showing the difference between the new HGV and the Hyunmoo-IIC ballistic missile. The new HGV has a "wedge-shaped" configuration, while the Hyunmoo-IIC, a MaRV, has a "conical-shaped" configuration.
● According to official statements from the Department of National Defense, the Hyunmoo-IIC maneuvers at Mach 9 and is fully equipped with MaRV capabilities.
● The IIC adopts a configuration similar to the U.S. Pershing II, and according to announcements, its range was confirmed to be approximately 1,000 km as of 2022. (It seems like the focus was on warhead weight rather than range.)
● According to an announcement by a senior researcher at ADD on the official Presidential media account, the HGV perfectly executed irregular gliding flight at low altitudes, freely changing its trajectory vertically and horizontally at ultra-high speeds exceeding Mach 5.
● In practical scenarios, the operational range is expected to be within 1,000 to 2,500 km, and there is no anticipated reason to fly beyond 2,500 km.
● The wedge-shaped design is technically more difficult than the conical-shaped design.
Please keep in mind that this does not mean one is better than the other.
● The HGV's TEL is presumed to be a heavily modified 10x10 TEL and the payload capacity have not been disclosed.
It is here.
WATCH: Britain’s ITN has released ~4 hours of previously sealed 1989 Tiananmen footage — filmed by the team that smuggled the tapes out of China.
The material, archived for ~35 years, includes a 2-hour crackdown compilation and 90 minutes of secretly filmed raw rushes from the days after, showing military trucks on Beijing streets, crowds, and the aftermath.
65
xendex retweeted
An very interesting comparison image showing the difference between the new HGV and the Hyunmoo-IIC ballistic missile. The new HGV has a "wedge-shaped" configuration, while the Hyunmoo-IIC, a MaRV, has a "conical-shaped" configuration.
● According to official statements from the Department of National Defense, the Hyunmoo-IIC maneuvers at Mach 9 and is fully equipped with MaRV capabilities.
● The IIC adopts a configuration similar to the U.S. Pershing II, and according to announcements, its range was confirmed to be approximately 1,000 km as of 2022. (It seems like the focus was on warhead weight rather than range.)
● According to an announcement by a senior researcher at ADD on the official Presidential media account, the HGV perfectly executed irregular gliding flight at low altitudes, freely changing its trajectory vertically and horizontally at ultra-high speeds exceeding Mach 5.
● In practical scenarios, the operational range is expected to be within 1,000 to 2,500 km, and there is no anticipated reason to fly beyond 2,500 km.
● The wedge-shaped design is technically more difficult than the conical-shaped design.
Please keep in mind that this does not mean one is better than the other.
● The HGV's TEL is presumed to be a heavily modified 10x10 TEL and the payload capacity have not been disclosed.
Breaking News. The South Korean Presidential Office has released photos for the first time showing the HGV (Hypersonic Glide Vehicle) striking a maritime target. It struck the target in "high dive" mode after flying at high and low altitudes, and its operational range is estimated to be 200~3,000+ km.
A booster appearing larger than those of the Hyunmoo-IIC, II-D?, or IV-1? has been identified, and the glide vehicle also features a shape suitable for high-maneuverability, long-range flight.
It appears that the new HGV launcher is being operated by either replacing only the canisters on the Hyunmoo-V's 18x18 TEL or by using a TEL slightly smaller than the 18x18.
We are frighteningly oblivious to the determination of their ADD, the Blue House, politicians, military, Supreme Command, and missile industry.
Since the HGV must fly at high and low altitudes, its minimum range is likely 200 km, and if it does not perform complex maneuvers, its maximum range will exceed 3,000 km.
The ROK Strategic Command has so many existing ballistic and cruise missiles to send to North, it is presumed that this system will be allocated to other regions.
