←
David2,039 posts
David avatar✓

David ✓

@_dsencil

@BitcoinNews media product manager and writer

◷ Joined May 2022▧ 381 Photos and videos
Pedro Domingos avatar
⟳ David retweeted
Pedro Domingos@pmddomingos·Nov 17

A Ferrari is 10X faster than you on a racetrack, but you're infinitely faster than it crossing a jungle. The problem with the fast takeoff crowd in AI is that they think the world is a racetrack, when in fact it's more like a jungle.

David avatar
David@_dsencil·7h

A good reminder that while crypto-connected crime makes headlines, tradfi crime is on a whole other level. Yet seemingly, it doesn't command the same amount of outrage or attention.

David avatar
David@_dsencil·Nov 16

I hosted this week's episode solo. @2084Richardson please return next week to save me from myself.

21:09
David avatar
David@_dsencil·Nov 16

Bitcoin puked over the weekend. What's going to happen to the BTC/Nvidia chart once U.S. markets open Monday? I wrote and released this op-ed in the newsletter last Friday morning EDT. Subscribe for free to get it then. bitcoin.com/newsletters/

Post attachment
Venice avatar
⟳ David retweeted
Venice@AskVenice·Nov 15

Investigate in private. Publish when you’re ready. Journalists should break the story on their own terms.

0:57
David avatar
David@_dsencil·Nov 16

Crypto is working, but broadly speaking price action remains unexciting. We're likely in a repricing phase caused not by people selling (because most weak hands already have), but from a lack of new buyers specifically because there are now a handful of projects with real meaningful revenue which attracts a majority of net inflows.

David avatar
⟳ David retweeted
David@_dsencil·Nov 5

It's not always about...

0:17
Eric Balchunas avatar
⟳ David retweeted
Eric Balchunas@EricBalchunas·Nov 14

Coverage of gold ETFs in the media from 2014 to 2018 was absolutely brutal - as bad or worse than bitcoin lately - it was like a string of Morrissey song titles: 'new milestones in misery' 'massive disappointment' 'give up hope' 'a pet rock' and how it 'doesn't work' as safe haven or inflation hedge (tbf it did go down and stay down despite $5T in new money printed by gov). IMO notable for a few reasons. One, the disrespect came despite GLD 10x-ing SPX from 2004 to 2011 and was STILL beating it since 2004 ETF launch. Two, it proceeded to go up 250% over next 7yrs (easily beating bull mkt mode SPX) and Three, it had fallen to about $1200 an ounce, which would seem like a shockingly high number to anyone from distant past. Bigger takeaway here (and one that we point in the new book) is that assets without cash flows can be really hard to hold bc sentiment runs the show and they tend to do opposite of what you think/want and they get sht on the most from the media and high priests of investing despite how far they've come over long-term. It is a hard, hard road for investors. This was one of the reasons i was fascinated enough to write book on the topic.

Post attachment
Bitcoin.com News avatar
⟳ David retweeted
Bitcoin.com News@BitcoinNews·Nov 14

🇫🇷 Kidnappings. Physical attacks. Crypto holders targeted in real life. In a Bitcoin.com documentary special, @_dsencil investigates France’s wrench attack problem with @knafo_sarah and Marilyne Ordekian. If France is the warning, what comes next? Watch now.🔧

6:45
David avatar
⟳ David retweeted
David@_dsencil·Nov 13

BTC's breaking below $68K ahead of U.S. Friday morning opening. Something tells me U.S Friday market is going to be a pretty rough.

0:10
will depue avatar
⟳ David retweeted
will depue@willdepue·Nov 12

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 unemploymentlabs.com/checkpo…. 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.

Post attachment