Michael Sealander (@msealander) profile picture

Machine Learning Engineer, Data Scientist, and Bon Vivant

Puerto Rico, Vietnam
Joined October 2008
Photo by Michael Sealander (@msealander)Photo by Michael Sealander (@msealander)Photo by Michael Sealander (@msealander)Photo by Michael Sealander (@msealander)Photo by Michael Sealander (@msealander)Photo by Michael Sealander (@msealander)Photo by Michael Sealander (@msealander)Photo by Michael Sealander (@msealander)Photo by Michael Sealander (@msealander)Photo by Michael Sealander (@msealander)
Michael Sealander retweeted
NEW and BIG Nvidia booked $104bn of revenue through its Singapore subsidiary in 2025, up 72% in a year and almost half its global sales. That was the same year it stopped breaking Singapore out in its results. By @OwenWalker and @AntoineGara ft.com/content/dbde7a30-548c…
59
470
138
1,803
425,296
Michael Sealander retweeted
Puerto Rico has now received 4x as much 🇺🇸 propane in the past 8 months as it received from the US mainland in the previous 22 years combined. The waiver has brought to life a domestic supply chain that the Jones Act had previously crushed. – at Washington, DC
10
121
22
929
65,019
Michael Sealander retweeted
Yes. And the postmortem must include more than protecting against corruption. We need a national movement dedicated to reducing the power of the executive branch.
Democrat Senator Elissa Slotkin says the quiet part out loud: “We need to make sure that we are thinking very hard about how to Trump-proof the government going forward so that no leader can do this again.”
22
68
2
496
9,938
Michael Sealander retweeted
It's been almost two weeks since the election, and the GOP has obviously learned nothing. That might be good for Democrats in the short term, but it's terrible for the country in the long term. We need two healthy parties. In the absence of a conservative renaissance, "populist kleptocrats and brutal bigots will be the only alternative for voters who may become disappointed with Democrats. Our military alliances will always be one election away from peril, and the integrity of the Department of Justice will hang on the outcome of every election." nytimes.com/2026/11/15/opini…
65
144
15
826
64,292
Abandoning imperialism is the best way for Russia to address its security concerns.
43
94
9
410
17,970
Michael Sealander retweeted
Mechazilla catches Starship! Congratulations to the entire SpaceX team on this incredible achievement!
6,798
27,311
9,086
140,951
17,163,449
Michael Sealander retweeted
Lutnick gets some heat from conservative media: “.. I am sure it is just a coincidence that Cantor Fitzgerald .. has a $362.3 million stake in $NVDA, which makes up 7% of Cantor’s entire portfolio.” washingtonexaminer.com/opini…
56
535
28
1,444
65,142
Michael Sealander retweeted
Jet-powered Russian drones with Nvidia chips are now breaking into Ukrainian airspace and killing people because Starlink restrictions hinder Ukrainian air defense.
1,374
8,070
301
20,858
459,691
„Finlandia dostarcza Ukrainie broń nie tylko do obrony, ale także po to, by pokonać Rosję w wojnie.“ „Jedynym prawdziwym rozwiązaniem, by powstrzymać Moskwę, jest „zmilitaryzowanie Ukrainy po same zęby”.” - Aleksander Stubb
1,011
5,654
172
25,103
163,188
Michael Sealander retweeted
Putin gained a lot of experience bombing hospitals in Syria. Now applying it in Ukraine. Bastards
We were on the scene minutes after a Russian drone slammed into a hospital last night in central Kyiv. I found a burning building, dazed patients & firefighters. The attack came after Ukraine reported that @SpaceX had again throttled the Starshield system theguardian.com/world/2026/n…
46
1,141
9
3,614
76,721
Michael Sealander retweeted
NATO nations withdrawing from *their own* borders is a preposterous demand by Russia. Putin has proved once again that he’s not serious about a negotiated peace. Further talks are a farce at this point. Putin will only agree to stop the war when he knows he can’t win. We need to help Ukraine prove that point. nytimes.com/2026/11/13/us/po…
2,622
3,032
374
15,527
1,183,673
Michael Sealander retweeted
Russian Olympic jackets, just in case you still weren’t sure which team they are on.
Meet America’s “negotiators”, ladies and gentlemen 🤦‍♀️
549
6,138
156
29,854
457,543
Michael Sealander retweeted
Kennedy: Apparently, the Department of Commerce greenlit the sale of $4 billion in advanced chips to the spy’s company. I'd like to hear from the Secretary of the Treasury. And more to the point, the Secretary of Commerce. I mean, why were they selling the chips? This whole thing is an intergalactic freak show.
106
649
50
2,510
175,490
Michael Sealander retweeted
. 👀👀
Replying to @TianaTheFirst
I'm sure the Lutnick sons' Liberation Day Treasury shorts were a coincidence. I'm sure their AMD stake was a coincidence, and I'm sure their plan to purchase tariff refund rights (betting AGAINST the legality of the tariffs) was a coincidence. washingtonexaminer.com/opini…
5
58
338
10,289
Michael Sealander retweeted
90% of voters said corruption was a motivating factor in voting. The Trump Administration response is 🖕 to the voters.
Todd Blanche suggests that perhaps Alice Huang isn't really a spy after all and says he has no concerns about Lutnick's alleged involvement in the Nvidia scandal (they are circling the wagons!)
81
466
12
2,377
108,552
Michael Sealander retweeted
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.
1,820
6,413
2,194
41,179
6,662,557
Michael Sealander retweeted
Incentives matter. And these incentives are terrible.
The questions surrounding Nvidia/Huang, China, the Commerce Department (Lutnick), & Trump are one of many reasons why the US government shouldn't have a financial stake - a cut of overseas sales, equity stakes, etc - in the same firms it regulates. Today it has DOZENS.
2
18
80
4,345