Matt (iotapi322) ❖ retweeted
Now I also confirmed that 2080ti unlock is successful with similar approach as the 30-series' method
it reach nearly same speed as 350w 3090(locked) with only 250w in mmapeak and mamf-finder
github.com/KohakuBlueleaf/Ko…
Confirmed that 3090 unlock is successful
with 360watt power wall, I got 75T -> 135T (measured in Pytorch)
Which is insane
I used github.com/Zagitta/gpu-unloc… with commit a9b0c75
Matt (iotapi322) ❖ retweeted
the $noid dev funds parked 1m noid on oct 16. 4 minutes apart, exactly 500k each. find a big redflags🚩
tl;dr oct 16 both dev funds moved exactly 500,000 noid to fresh addresses, 4 min apart. 1m noid, ~9% of supply
the setup, from parano1d's own economics page:
- 10% of every block goes to two funds (O(1) Network Fund + Parano1d Lab)
- thats 2,304 noid each per day, paid every 2,880 blocks
- disclosed, nothing hidden there
what the explorer shows:
- oct 16 13:59 utc: fund A sends 500,000 to o1rv9y…
- oct 16 14:03 utc: fund B sends 500,000 to o1m0p7…
- both still sit there, zero outbound
- oct 20: each fund sends exactly 25,000 through one fresh hop into safetrade's hot wallet. 50k noid
and the big miner:
o189…sv6p59 still won 1,611 of the last 2,901 blocks (55.5%), down from 62% on oct 20. it sweeps to safetrade too
xitter.cf/Checker1crypto/status/…
Conclusion:
🚩As soon as I brought up the topic that no one talks about, the developer immediately blocked me, which speaks volumes. 1m Noida's strange transfers account for 9% of the total supply.
🚩One address mines 55-60% of the entire mining pool; the person in my comments received nothing at all: xitter.cf/Solid_Primal/status/21…
🚩And mining itself, according to whattomine, is unprofitable:
xitter.cf/Checker1crypto/status/…
Now when you hear "noid new btc" think twice before buying and reread my posts. nfa.dyor.
addresses and blocks in the reply. check it before you trust it
That's all for me.
the $noid dev blocked me
tl;dr every number i posted on parano1d came from his own docs and his own explorer. they all stay up, and you can check each one without him
for context, the last week here:
- v2.1 launchpad vs 46k new noid a day (his economics page)
- empty block proves in 16.6s, full one in 17.6s (his performance page)
- asked for signed wallet hashes so fake wallets are a 10 second check
no leaks, no fud, no price calls. just his tables read out loud
and to be fair to him: devs block for all kinds of reasons, and parano1d's docs are some of the most detailed in this whole niche. thats exactly why i could write those posts
so nothing changes here. if you want to check $noid yourself, 3 pages, no dev needed:
- performance: prove and verify times per block type
- explorer stats: live hashrate, supply, block time
- parameters: 30s blocks, small vs large limits
links in the reply. dont trust me, dont trust him, read the tables
which number do you want me to pull next?
to put things into perspective and to show how many miners have left mining... Before PoS Etherum was 21.2M a day in emissions and Pearl is 2.4M today.
We all think we are doing great...
DeepSeek V4.1 Flash 1017 on 4× CMP 170HX — open recipe 🛠️⚡
1 user: 296 tok/s structured / 150 prose
8 users: 796 tok/s aggregate
1M context · KV 1.57M tokens
Prefill 1,759 tok/s
189 GiB Engram tables streamed off NVMe
180 W per card
github.com/GNmac-lab/deepsee…
Best thing about AI.... @MiningRabid who couldn't even write a bash script a year ago, now has a bad ass website, his own mining OS and a mining pool...
Super proud of this guy!
Matt (iotapi322) ❖ retweeted
We will not host any Blake2B miners for the following reasons:
1. You are mining a fake chain and will probably quit, not pay your bill after 3 months.
2. The entire XBT community is delusional and difficult to work with.
3. I fully understand what XBT is and do not need to be “educated”. I just think you are all retarded.
Matt (iotapi322) ❖ retweeted
Replying to @ashxhart
@ashxhart has said he plans to support ampere on Tensorfold and you can be sure I’ll be giving it a go as soon as it’s ready. It has some properties that make it good for AI enabled optimisation, which vllm lacks.
For now, I will continue working on the 170hx vllm fork. I’ve spent the past few days retesting existing optimisations to confirm they all earn their keep after p2p was switched on and the additional SMs were unlocked (hint: no!). I’ve also been moving as much of the code as possible behind vllm plugins so that rebasing to the newest vllm is not such an ordeal (you can guess what my last few days have been spent on).
I’m back on optimising kernels, but think there’s probably little juice left to squeeze out of glm 5.3 flash without targeting some other bottlenecks. The pcie links are a significant bottleneck and I’ll be looking to see what can be done to work around that.
Exl3 kernels are coming along well and I’ll be releasing a first attempt at an optimised qwen 3.8 next flash recipe next week. This will be a 5.05bpw exl3 checkpoint that runs on 2 cards with 1 mil kv capacity.
Yet another reason not to fly @Delta
You're not going to believe how nasty Delta is.
Delta CEO Ed Bastian got caught trashing Elon Musk at internal meetings to explain why they rejected Starlink. But don't buy the corporate moral high ground for a second.
Delta tested Starlink, but Elon refused to let them trap passengers behind ad-tracking portals to farm and sell user data. SpaceX wanted instant connection just like home, with zero data harvesting.
So what did Delta do? They signed with Amazon’s completely untested satellite network which won't even launch fleetwide until 2028. Strictly to monetize passenger information.
United flyers get 300 Mbps; Delta flyers get dial-up and a data harvest.
My god
BREAKING: Anthropic’s IPO prospectus reveals a staggering $42,000,000,000.00 net loss in 2025.
Matt (iotapi322) ❖ retweeted
Replying to @0xSero
Matt (iotapi322) ❖ retweeted
Apparently Lockheed hired a Chinese front company to handle its logistics of the most advanced and classified airplane modules to Australia that ended up rerouted to China instead!
F-35 stealth parts, including a classified radar-absorbing canopy, were rerouted from Australia to Hong Kong this summer. The Pentagon says some components are missing.
Politico reports the shipment was traveling from Australia to the US for repair when an intermediary handling logistics for Lockheed Martin diverted it across the Pacific to Hong Kong instead. Nobody has publicly explained why. Nobody knows who currently has the parts.
The canopy matters specifically. It's coated with a classified radar-absorbent material that makes the F-35 invisible to radar. That coating is available only to the US and its partners. A physical canopy lets an adversary study and potentially reverse-engineer the material, something stolen digital blueprints alone can't fully replicate.
Congress has been receiving briefings since June. The Pentagon confirmed the incident and said it's working to retrieve the parts and add safeguards. Lockheed said it follows all regulations but won't discuss specific shipments. The Chinese Embassy said it wasn't familiar with the matter.
For context: a 2023 GAO report found over 1 million F-35 spare parts had gone missing over five years, and the Pentagon couldn't audit the losses because Lockheed holds the data.
The most advanced stealth fighter ever built. A classified coating that makes it invisible. Last seen in Hong Kong.
New gpu chassis, from 2018
$300 bucks. Should run 8 170hx’s
Matt (iotapi322) ❖ retweeted
There you go... This entire "AI is going to kill us" was an orchestrated psyop.
Dario Amodei was on @FaceNationCBS this morning saying "The government and the public needs to have a stake" in AI and that he wants the entire AI industry regulated.
Translation: We need a taxpayer funded bailout but we'll call it an "investment."
They want government "regulation" because it means protection from the open source Chinese AI models.
As I said, this entire Jacob Coxon media frenzy was an orchestrated psyop to manipulate public opinion into providing a bailout.
When companies go public they must file paperwork disclosing their financials.
Doing so would expose the massive cash burn OpenAI is using. Therefore, Sam Altman can't go public but he also has run out of money.
Enter "whistle blower" Jacob Coxon.
It made no sense how an obscure nobody got 165 million views on his first X post saying AI could kill us all. Then he was on every news station within 48 hours, scaring the public.
This was a psyop to "regulate" the AI companies... aka give them government funding to "protect the public from the dangers of AI."
Many have forgotten about the Orchard exploit for $ZEC (which was never accidental to begin with, it was intentional)
I have it on good authority that an equivalent amount of the supply has been secretly minted
And at the right time, Zec is a luna 2.0 waiting to happen.
You have been warned.
Matt (iotapi322) ❖ retweeted
🚗 Volkswagen's board approves cutting 50,000 more jobs, in addition to the 50,000 already approved.
Also closing 4 plants
Don't know if true but, no 50 super series
en.gamegpu.com/news/zhelezo/…
Matt (iotapi322) ❖ retweeted
A hacker just drained $75 MILLION from a lending protocol so they shut down the entire blockchain to stop him
The attacker pumped a token's price 100x in 20 minutes, then borrowed against the fake value to drain everything
Cronos, the chain built by Crypto com, froze all activity mid attack, which trapped $60 MILLION of the stolen money before it could escape
Only $6 MILLION made it out
Now the validators have to decide, restart the chain and let him keep it or roll it back




























