SocialFi Panda (@SocialFi_Panda) header
SocialFi Panda (@SocialFi_Panda) profile picture
Joined February 2013
Photo by SocialFi Panda (@SocialFi_Panda)Photo by SocialFi Panda (@SocialFi_Panda)Photo by SocialFi Panda (@SocialFi_Panda)Photo by SocialFi Panda (@SocialFi_Panda)Photo by SocialFi Panda (@SocialFi_Panda)Photo by SocialFi Panda (@SocialFi_Panda)Photo by SocialFi Panda (@SocialFi_Panda)Photo by SocialFi Panda (@SocialFi_Panda)Photo by SocialFi Panda (@SocialFi_Panda)Photo by SocialFi Panda (@SocialFi_Panda)
Nvidia’s board has completely lost the plot. If I were Jensen, after some Hail Marys, I’d sell every share of NVIDIA I own and start a rival chip company. A chip on his shoulder. Literally. Anthropic <> OpenAI story 2.0
1
12
570
IMD - AI swarms. Almost won Paradigm AMM & prediction market contests. Great founder @surfcoderepeat. ORBIO - private AI inference. VVV on Robinhood Chain. They're small now, but IMO will go big. Most of interesting AI projects are on Robinhood Chain. Super bullish on RHC.
Here's my unabashed hyper bullish view on where crypto is going Every onchain cycle since 2013, we've had an onchain boom/bust. The boom typically shows what pmf would look like for onchain experiments if it were sustainable and with the bust it kills 95% of the experiments but leave just enough to build unicorns. Lets take NFTs, Cryptokitties in 2017 made up of 20% of all eth txns and during mint/play time ETH was basically unusable, but the ability to mint unique NFTs became a thing and it predated the 2020/21 NFT boom. Now post AI you could argue content is commoditized and what humans now prefer is scarce art with provenance, real human connections, in this world NFTs will likely make a comeback. So from 2013–2023, we pretty much speed ran all of the crypto experiments onchain. Now we need a reason for these things to be needed, like the NFT example above. Or if you look at DAOs/Governance useless today but with AI agents could argue autonomous orgs will be a thing. Onchain needs tech catalysts and I think AI is the right catalyst to validate a lot of onchain experiments of the last decade. But what other hidden catalysts are there? Thats my bet for crypto.
4
256
Short term, bearish on open models. China faces tighter access to rented frontier compute. Long term, super bullish on open models. This accelerates Chinese chip innovation, leading to a more decentralized AI stack over time. Humanity deserves open AI.
4
180
Doesn’t mean anything. Chinese companies can rent export-controlled compute served in Singapore, per U.S. law now. Why fire Jensen when he hasn’t even been convicted? Geopolitical nonsense plus a weak board. I'm sure China will only accelerate AI compute after this.
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
2
240
First TikTok, then Nvidia. Sure, the best way to win the US-China tech war is to do everything except focus on your own innovation and development. Sad.
1
10
451
What Jensen did is wrong, but the board shouldn’t fire him. Old news bubbling up now after the midterm elections… makes you wonder. Don’t slow down AI. Don’t be the next Uber.
1
5
276
SocialFi Panda retweeted
Privacy Cash is back! We worked hard with @CipherOwl to strengthen our safeguards against bad actors linked to the Binance hack. Thank you to our users for your patience and understanding. Deposits are now enabled for all supported tokens.
22
6
2
136
7,954
Jason can just buy $1M/(ATH_price-current_price) BTC. He loses $0 if he loses the bet. He makes up to $2M if BTC stays above $14K but below ATH. He’ll only lose money if BTC is below $14K at the end of 2028 (not gonna happen). Pretty much free money for Jason.
Bitcoin will never break its all-time high again." — Jason Calacanis Jason, I'll bet you $1,000,000 it will break its ATH. Here are the terms: Bitcoin hits or exceeds $126,210.50 on or before October 28, 2028, 12PM PST I’m happy to trust you to pay or if you prefer we can each escrow the dollars with independent party. The bet is either settled when Bitcoin hits the ATH or on October 28, 2028. @jason Do we have a bet?
1
1
7
766
Another overnight success with ten years in the making. Starting 2015, he founded and failed a bunch of apps. Biggest success only hits ~$1M ARR. Eventually, he pivoted into Monica in 2023, found PMF and then eventually launched Manus. Survive long enough & adapt to get lucky.
Today marks a moment I'll remember for the rest of my life. When we started Manus, few believed that general AI agents could work. We were told it was too early, too ambitious, too hard. But we kept building. Through the doubts, the setbacks, and the countless nights wondering if we were chasing the impossible. We weren't. This isn't just an acquisition. It's validation that the future we've been building toward is real, and it's arriving faster than anyone expected. But this is not the end. The era of AI that doesn't just talk, but acts, creates, and delivers, is only beginning. And now, we get to build it at a scale we never could have imagined. To everyone who believed in us before it was obvious: thank you. The best is yet to come.
5
3
3
154
40,573
Replying to @Austen
I remember Evan Spiegel said the same when turning down the Facebook $3B offer. When you have found your life’s work, you won’t walk away for being acquired for 600x P/E.
5
153
75,947
1. A friend of mine founded a GenAI company back in 2020. He raised a lot of money, hired about 60 people, but struggled to find any meaningful traction. There was one point when the company was too short on cash, and he had to let go of most employees. Hopeless. Then GenAI had a breakthrough, and his company is now one of the largest and most profitable GenAI companies in the world. 2. Kite was founded in 2014 using AI to help developers write or generate code. It was too early for the market. AI models just weren't good enough. They persisted for 7 years, and finally shutdown in 2021. Had the team persisted 1 more year, the company probably would be a unicorn. But hindsight is 20/20. Who knows the breakthrough would come 1 year after their company shut down, rather than 10 years later. 3. Timing is a bitch. And most of the time, you've no idea whether you are right on time.
This quoted post is unavailable.
4
23
2,447
SocialFi Panda retweeted
1/ For too long, crypto and open source developers in the US have been living under a cloud of doubt. That uncertainty ends today, with an emphatic statement from the DOJ that *shipping code is not a crime.*
22
76
25
438
83,775
SocialFi Panda retweeted
The ACTING HEAD OF THE DOJ CRIMINAL DIVISION just stated the following regarding Section 1960(b)(1)(C) and third-party liability for software developers: “We may under certain circumstances bring cases under Section 1960(b)(1)(C), which prohibits the transmission of funds that the defendant knows are derived from a criminal offense, or are intended to be used to support unlawful activity. However, going forward, consistent with principles of notice and fairness, let me make the following clear: Many developers have relied on regulatory guidance, to suggest that, non-custodial cryptocurrency software does not constitute an unlicensed money transmitting business. While that guidance may not be binding on the department, its implications can of course factor into prosecutor’s charging decisions. Therefore, where the evidence shows that software is truly decentralized and solely automates peer-to-peer transactions, and where a third-party does not have custody and control over user assets, new 1960(b)(1)(C) charges against the third-party will not be approved. Though, if criminal intent is present, other charges may be appropriate. All of a subject's conduct and the services they provide end-to-end will be considered. Generally, developers of neutral tools, with no criminal intent, should not be held responsible for someone else's misuse of those tools. If a third-party's misuse violates criminal law, that third-party should be prosecuted – not a well-intentioned developer.”
10
21
12
99
48,762
SEC Commissioner in support of private DeFi. Privacy is not a crime. Roman Storm should be free. sec.gov/newsroom/speeches-st…
1
7
821
Hi all, we're sunsetting Live Bet due to lack of product-market fit. You can withdraw your current balance at: thelive.bet/. Our team will shortly process your refund requests. Thank you all so much for trying such an early-stage product. It's been a real pleasure building with you. We're now building a new product on Solana and will be making more announcements in the coming months when we're close to launch. See yall soon!
13
1
3
43
16,668
Launching Live Streaming for Crypto @thelivebet Creators can now host their own shows and receive crypto as tips. They can also airdrop tokens to their listeners. Our biggest use case so far is enabling creators to host their own sports commentary. Our plugins let commentators integrate betting into their broadcasts and monetize. In the future, creators will be able to use plugins not only to monetize their content but also to make streaming more engaging for their audience. Join below to test.
7
4
2
42
20,523
1/ This screenshot is floating around and I know the back story. How did a Groupon PM go on to build a supersonic airplane company? Here is the little-known and tremendously inspiring story of how @bscholl started @boomsupersonic (and my very small part in it) 👇
140
518
212
5,438
1,206,385
Launching thelive.bet/ incentivized testnet, a prediction market for livestreams. Deposit up to 50 USDC now to bet on your favorite streamers, and earn loyalty points. Top 5 coin holders this week will split the prize pot. We will increase deposit cap size and thus pot size every week, with more news and activities to come. So get in early! TG group: t.me/+Crd2IwIkQek2NzBl Bet now on IShowSpeed: thelive.bet/stream/youtube~1…
7
5
3
62
20,821
SocialFi Panda retweeted
With the incoming American election, prediction markets have started to gain notable traction and attention. The thesis has always been the same: crypto's open-source nature allows for a global and efficient liquidity layer where anyone can participate. mirror.xyz/1kx.eth/jnQhA56Kx…
11
26
6
154
29,226