solar

the world you grew up in no longer exists and you will never be who you used to be

[solar @solarnius]

@CookerFlips i mean capital flight from risky assets as the AI trade unwinds.. pretty straightforward to sell off in the 70s and 80s if youre worried about the risk environment

[solar @solarnius]

Hey guys does anyone have an update on the value of my central african republic land plots that I purchased for $CAR tokens in 2025?

Could really use some good news right now!

[solar @solarnius]

oh so this is what the US government was frontrunning with the affair news, sick

[will depue @willdepue]

I have a major result to share.

I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China.

This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model.

Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves.

I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems.

I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used.

It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20.

These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs.

I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at https://t.co/YOtzND0Jqk.

There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.

[solar @solarnius]

update gave em a high enough number that i had to close my laptop after sending the email let's see

[solar @solarnius]

considering a j*b for the first time in 5 years........... pray for me.

[solar @solarnius]

@BowTiedBroke also FYI your friend is breaking the law by bringing this back to TN... probably fine but idk if you want that out there on Twitter

[solar @solarnius]

@BowTiedBroke You should get your friend in rehab ASAP; it's very possible that concentrated 7-OH is scheduled by the DEA as soon as 2 days from now (seriously, November 13: https://t.co/kwjoTlkkQk) at which point it would be as illegal to buy/possess/take as heroin.

[solar @solarnius]

@Jason______A idk if u saw the update on this but it got repoed

the $3.1 number was coming from him putting 400k down and 45k for 60mo... add it all up and u get $3.1M lol

[Rhone @Rhoneus]

i may have wasted 10 years of my life in crypto but at least i alienated my friends and went insane as well

[solar @solarnius]

not really related to trading performance tho i had a decent year i'm just pretty tired of trading crypto and want to be on the frontier with smart people

[solar @solarnius]

considering a j*b for the first time in 5 years........... pray for me.

[solar @solarnius]

i mean the correct path here is clearly to migrate the token using some vehicle, it’s not particularly hard and could be vibecoded in like an hour

it’s the same deal as if a founding team needs to raise capital but all the shares are vested… you issue more shares and dilute everyone

since mint is revoked on pump, need a new token with a higher supply to do this but this is trivial to do

sure some holders will be sad they got diluted but you can keep them happy by raising at a premium to market eg if you are trying to raise $200k you could sell 5% at 4M with the token trading at 3M so any dilution is offset with an increased valuation (but buying 5% at market would cost more than $200k due to price impact so everyone is happy)

it’s just lessons that have been learned in tradfi startup investing already that also are clearly best practices in crypto

maybe i’m missing something but feels obvious to me cc @AgentChud

[solar @solarnius]

@zinceth am i wrong tho?

[solar @solarnius]

bro imagine how good it must feel to be a chinese spy getting cracked by jensen

[solar @solarnius]

@lex_node @VictorTaelin this, the business is a slam dunk for 10 on 50 from VCs it's just too rich for 40 on 120 and trying to raise 8 on 120 from genpop is ridiculously misguided

[solar @solarnius]

i mean the business model slide is where i imagine you lost every fund, along with the ask being for a lot of $

bend 3 being closed source doesnt make any sense

probably if you establish a "formal verification as a service" business and hit a hockey stick revenue curve you could raise on these terms but the business is too ephemeral to invest in at this valuation otherwise

[solar @solarnius]

@gonkrahs you played well, two unlucky games in a row unfortunate but gg great run

[Evanss6 @Evan_ss6]

Extremely dumb and reflects poorly on @MetaDAOProject by association knowing they worked with the team to reach this decision

The project is obviously dead on arrival now. You are better off just not wasting your time building it at this point.

[solar @solarnius]

@MiLB_LUL @NicoleTipple in typical riftbound fashion tho it's never enforced sadge

[solar @solarnius]

@MomSkyRB If you make top8 I will rename my riftbound profile to Ron Jiftbound in your honor