They're claiming AI is dangerous because an OpenAI model attacked Hugging Face.
When it happened, frontier models were completely useless thanks to their dumb guardrails.
What actually defended it was a self-hosted GLM-5.2.
Yet this cartel is out here saying their neutered frontier models are "safe," while the open-weight AI that literally stopped the attack is the real threat to humanity.
Ridiculous.
Show more
Today is the cheapest day to buy a DGX Spark.
People said $4k is too expensive just few weeks ago.
Been testing Grok Bot for a few days now, and all the hype posts on X really remind me of Openclaw back then.
What mindblowing feature am I actually missing here?
Hardware price can't go down until 2029.
JUST IN: Samsung commits ~70% of its memory capacity through 2031 to long-term deals with Nvidia, Microsoft and Google as HBM supply tightens years ahead of new capacity.
Typical ClosedAI
JUST IN: Apple, $AAPL, says OpenAI is destroying evidence in trade suit, per Bloomberg
Deepseek-V4-Flash-Vison beats Opus-4.8
Welcome to local AI era.
DeepSeek-V4 can see now. 👀 DeepSeek-V4-Flash-Vision-Exp is the family’s first experimental multimodal model. MIT License.
🤖
🏆 Beats Opus-4.8 on Agents’ Last Exam and ZeroBench, and trails it by just 0.7 on Chartography.
🖼️ Understands images, screenshots, and charts inside tool-using agent workflows.
⚡ Vision added, text strength retained: 83.9 on Terminal-Bench 2.1 and 59.3 on DeepSWE, above Opus-4.8’s 58.0.
🧠 Built on V4-Flash with new visual modules. Weights, prompt encoding, and minimal inference code are open.
Show more
Deepseek-V4-Flash-Vision is now open.
Seems like tie with GLM-5.3-Flash, testing it now
Download it while you can.
I’ve been realizing lately just how insane the information asymmetry is.
I heard normies are still paying hundreds or even thousands of dollars for courses on how to build AI trading bots.
Feels like a massive opportunity right now.
Show more
Everyone's talking about Microduck and I didn't get one.
Pretty sure my cats would destroy it in seconds.
If I wanted to understand AI from the ground up, these are the accounts I would start with:
@jun_song local open models and AI hardware
@0xSero making local AI faster and cheaper
@Da7_Tech honest takes after long model runs
@haider1 honest takes on AI labs and what makes models work well
@MiaAI_lab running models on Spark
@morganlinton coding agent benchmarks and which model to use
@kimmonismus every AI news with good context
@0x_kaize agent loops and model rankings
@notjazii tracking model leaks and testing them
@Teknium everything you wanna learn about Hermes agent
Show more
2 x DGX Spark is cheaper than coffee.
They never disclose the actual limit for Plus and quietly change it all the time, so how is "20x more" clear at all?
4 x 20 is 80, and 2 x 20 is 40.
For clarity, while both are called 20X, in Codex they apply specifically to weekly usage limits. And we also don't have 5h limits for both Pro plans.
The Pro 20X is quite precisely 20X the usage of the Plus subscription, so it does exactly what it says on the tin.
Show more
If you still think open source AI is falling behind,
It isn't the one falling behind.
You are.
The most common way people get into local AI:
> Hit rate limits on a $20 sub
> Upgrade to the $200 plan, only to hit limits again after a silent nerf
> Start hunting for cheap options and unlimited tokens, but realize API costs are absurd compared to subscriptions
From what I've seen, this is the exact rabbit hole most people fall through.
Show more
The difference between Local AI and Frontier AI:
When the AI acts up or hits a bug (like infinite loops or broken reasoning):
With Local AI, you tweak the template or fine-tune it yourself and fix it immediately.
With Frontier AI, all you can do is spam replies under an employee's X post and pray they fix it.
This is what autonomy actually looks like.
Show more
What I've been hearing literally every single day for the past 2 years:
"Hardware is way too expensive right now, I'll buy once prices drop."
Only places building open-weight AI and sharing knowledge with the world deserve to be called Labs.
The rest are just for-profit companies selling human collective knowledge.
Reminder to call the AI companies “companies” and not “labs”
These are for-profit companies, and they seek to claim all the value they can for themselves
Don’t give them the benefit of the doubt by playing along with the “labs” language
Show more
People are still hyping up Codex as one of the best harnesses out there,
but in my experience, the insane RAM hogging and SSD thrashing haven't been fixed for over 6 months.
Built the ultimate local sovereignty starter pack.
Only difference is the Mac Studio is 512GB instead of 768GB, and running GLM-5.3 instead of 5.2 (ordering immediately once it's out).
Also threw in a Starlink.
Show more
The ultimate sovereignty starter pack:
- Mac Studio M5 Ultra 750GB
- Solar panels
- Power generator
- GLM-5.2
Building sovereignty right in your own backyard.
Open-weight AI labs have caught up to frontier models, and their low prices are eating up serious market share.
The only real narrative holding them back right now is the fear of China stealing user data.
If you are worried about that, you have two simple choices:
Run the open models locally,
or use a US-based third-party inference provider.
Show more
