Building @0xSupergemma Running Datacenter hardware export business in Korea Open Source AI

Pinned Tweet
$500 a month is higher than the median income in 128 countries. That is 66% of all 193 UN member states. Outside the top 10 richest nations, almost nobody can afford that kind of subscription. This goes completely against what they promised when raising billions: "intelligence too cheap to meter." Intelligence built on our collective knowledge belongs to all of us. This is why I am all in on open source.
OpenAI is quietly cutting usage limits so they can launch a $500 plan. This is a pricing scam built on opaque subscription limits. Two months ago, burning through a $200 plan was hard. Last year, the $20 tier was more than enough for almost everyone. Now? A $200 tier fails after one day of real work. The $20 tier is just a free trial. We need real consumer protection for AI subscriptions.
64
52
686
38,729
Just got my payout. The new original content rewards feel pretty much the same as the old ad rev share, maybe slightly better. It is a bit lower than my previous $2k+ payouts, but I was insanely busy the past two weeks setting up a data center hardware business in Korea. I barely posted and my impressions dropped by half. Looking at it that way, the payout actually increased per view. Always grateful for everyone here.
6
36
2,055
More context : Microsoft is the biggest shareholder of OpenAI (27%)
JUST IN: Bill Gates warns AI is powerful enough to cause “a billion deaths.”
11
12
194
7,856
Everyone is underestimating Grok right now. Yes, they started late. But they are closing the gap faster than anyone else. The biggest difference? They do not rent their compute. They own it.
🚨 BREAKING: SpaceXAI will have 1 MILLION GPUs online NEXT WEEK and 1.44 MILLION by year-end >Colossus 1: 150k H100 + 50k H200 + 30k GB200 >Colossus 2: 110k GB200 + 440k GB300 >Another 220k GB300 next week >Another 220k in November >Another 220k in December 660,000 GB300s TURNING ON IN 90 DAYS
40
7
174
9,836
This is why I love the word "SUPER". Everyone loves it.
Trump says Xi liked the idea of calling AI “super intelligence.” Because that would be “SUPER!” Everyone agreed. I did not use superintelligence to write this post.
6
2
46
4,712
He is dealing with an obsessive fear about security risks. You can trace this pattern across several of earlier articles about him.
Anyone know why Dario wasn’t invited to the White House state dinner today?
2
33
3,618
Jun Song retweeted
Bernie Sanders wants to throw you in prison for 20 years if you BUILD open source AI. The DOJ wants to criminally charge you if you CRITICIZE AI. This is how tyranny works. You're a thought criminal either way. And they can selectively prosecute you any time they want.
NEW: DOJ now designating critics of data centers as potential “foreign agents,” warning that posting “anti-AI propaganda” may bring criminal charges.
32
235
708
16,432
Stupid scammers. People are dying of cancer all over the world right now, every single minute. It makes my blood boil.
JUST IN: Investigation reveals former Anthropic researcher Jacob Coxon reportedly worked with a “doomer PR firm” while launching his viral campaign warning of AI’s existential risks.
11
19
456
31,018
Compute and datasets are officially some of the most expensive things on earth now.
Fentanyl Grade Compute: Why a GB300 Rack Will Out-Price Blow by Weight in 2030 A GB300 NVL72 weighs roughly 1,580 kg fully populated and recent purchase orders put it at $5M per rack. That is $3,165/kg. Strip out the 1.5 tons of busbar, manifold and coolant and the GPU packages alone are well into gold territory, but we're pricing the rack, because that's what you actually take delivery of. Where that sits on the illicit commodity curve today ($/kg): $2,400 Cannabis flower $3,165 GB300 NVL72 $3,500 Fentanyl $28,000 Cocaine $65,000 Heroin $138,000 Gold So today NVIDIA ships a product that is denser in value than weed, roughly fentanyl-grade, and still an order of magnitude short of cocaine. For now... GB200 NVL72 was $4M, GB300 is ~$5M, and Vera Rubin NVL72 is already quoted at up to $8.8M at essentially the same rack mass. Rack weight is constrained by the datacenter floor loading; rack price is constrained by nothing. That's ~1.75x $/kg per generation with HBM, CoWoS and power all supply-limited through the decade. Cocaine: Colombia's new president has pledged a hard-line security crackdown with $1B in US aid behind it Every prior crackdown has consolidated the industry into fewer, better-capitalized operators with better logistics and higher yields per hectare. Consolidation is deflationary. We model US wholesale drifting from ~$28k/kg toward ~$12k/kg by 2030 as the supply chain professionalizes. **Crossover: 2030.** Our extrapolated NVL72-class rack hits ~$17k/kg while wholesale cocaine falls to ~$12k/kg. At that point the rational cartel pivots to smuggling racks, except a rack has a fixed 500+ kW power draw, a 3,300 lb forklift requirement and an export-control regime that is actually enforced. Cocaine has none of these problems, which is why it will remain the superior product for anyone without a substation. Key risks to the thesis: US retail coke is still ~$60–200/g, i.e. $60k–200k/kg so the crossover only holds at wholesale. Retail compute (H100 hour on a neocloud) is also marked up, so we consider this an apples-to-apples wholesale comparison. Full model available to Coke Research subscribers.
2
38
4,865
OpenAI is quietly cutting usage limits so they can launch a $500 plan. This is a pricing scam built on opaque subscription limits. Two months ago, burning through a $200 plan was hard. Last year, the $20 tier was more than enough for almost everyone. Now? A $200 tier fails after one day of real work. The $20 tier is just a free trial. We need real consumer protection for AI subscriptions.
OPENAI 🔥: The upcoming ChatGPT Pro Max plan will cost $500. So far, this will be one of the most expensive AI subscriptions available on the market. I hope "it will be worth the wait" 👀
186
133
1,828
157,655
Jun Song retweeted
Come check out my website. It includes a list of my latest & greatest recipes. The "Install Model" button gives you a ready-to-use prompt for Claude, Codex, or others. It adapts the recipe to your setup, starts the server and tests it. mia-ai.net/models
108
127
1,233
179,641
Some of tech people on X don't actually care about new technology. They just want dopamine. Every time something drops, everyone screams that it will change the world. A week later, it disappears like it never even existed. Remember when Codex Voice was supposed to kill typing forever? Nobody codes with their voice today. Same thing happens with every new model from Grok or Gemini. People cheer for 24 hours, and then everyone moves on. Live-streaming robots, brand new reasoning methods, the list never ends. Our attention span is shrinking just as fast as tech is moving.
36
5
222
8,371
A simple bet for all the AI doomers out there. I will bet $10M that AI does not end humanity. If humanity survives, you pay me $10M. If humanity is destroyed, I will pay you $10M.
67
10
390
14,063
Now that anyone can build almost anything with vibe coding, people get one big thing wrong. The moat for most apps was never the code. It is the users.
27
6
195
7,968
X Premium Plus is honestly the only subscription I need right now.
7
1
51
4,940
The AI race is not a race against each other. It is a relay where all of humanity runs together to reach a whole new level.
13
1
101
4,485
OpenAI and Anthropic are clearly running out of compute. Because of this, they are quietly serving heavily quantized and nerfed models to regular users. The full price API does not feel degraded like this. This only happens on the paid monthly subscriptions. Selling people a subscription while secretly downgrading the model is just a scam.
GPT-Sol is spitting out Chinese characters mid sentence right now, which shows it's heavily quantized. What they're giving users is straight garbage. Is this what we're paying for?
64
85
1,027
79,742
If you ever have the time, you should really visit China. It is very different from what many people assume, and there are so many fascinating cities to explore. Seeing all the advanced tech rolled out everywhere in daily life is super fun.
So yeah… this is what we were teasing last night. TenPayGo for the trip. And yes, WeChat supports international cards and e-wallets too. We’ve had that covered for a while — but good to know we can still surprise you. 👀 If the trip somehow turns into new friends, group chats and “when are you coming back?” That part’s still our department.
9
3
121
8,858
Jun Song retweeted
Introducing MTPLX V2.12: The fastest way to run local AI now even faster and with more models for lower memory macs and the new M5 Ultra. Upgrades: - 100 TPS Qwen Flash Next sustained on long coding (M5 max). - 2x Prefill speeds Qwen Flash Next: 872 --> 1695 TPS. - Requests with images now use the compiled kernels for max speeds. New Models: - Qwen Flash Next MTPLX Optimized Quality: 8 bit model built for the M5 Ultra and M3 Ultra. - Bonsai 2 27B MTPLX Optimized Speed: Works on 16gb and 24gb macs with compiled kernels and MTP running as fast as possible. 64 TPS. - MiMo V2.6 Qwen 3.5 9B MTPLX Optimized Speed: My personal favourite. Mimos magic distilled into a 9B model easily runnable at high speeds on a 16gb mac. 107 TPS. +40 bug fixes and improvements in real world coding.
44
17
355
20,201
A software update bug just broke thousands of Samsung fridges across the country. Your fridge stops cooling because of bad code. How is this real life in 2026?
삼성 냉장고 전국 단위 먹통 사태 터진 이유 삼성전자에서 테스트 중이던 소프트웨어를 실수로 전산에 배포해 버림 그 바람에 전국에 있는 삼성 스마트 냉장고들이 단체로 전원 뻗고 가동 중단됨 추석 연휴 앞두고 음식이며 비싼 약 다 쟁여놨는데 냉장고 꺼져서 난리 남 냉장고에 몇백만 원어치 마운자로 보관하던 사람들은 전부 폐기 처분할 판임 삼성에서 긴급 조치하겠다고 공지 올렸는데 피해 가구가 수천 세대라 감당이 안 됨 지금 AS 접수 넣어도 다음 달 중순 넘어서야 방문 수리 온다고 함 명절 앞두고 냉장고 음식 다 썩어 나가는데 수리는 한 달 뒤라 소비자들 단체 멘붕 옴 소프트웨어 실수 하나로 날린 보상금만 계산해도 감당 안 될 수준임
24
15
175
26,406