Alice Hou (@Shuting_11a) header
Alice Hou (@Shuting_11a) profile picture

Prev. @MessariCrypto @TheBlockCo | Towards an onchain future.

New York, USA
Joined January 2022
Photo by Alice Hou (@Shuting_11a)Photo by Alice Hou (@Shuting_11a)Photo by Alice Hou (@Shuting_11a)Photo by Alice Hou (@Shuting_11a)Photo by Alice Hou (@Shuting_11a)Photo by Alice Hou (@Shuting_11a)Photo by Alice Hou (@Shuting_11a)Photo by Alice Hou (@Shuting_11a)Photo by Alice Hou (@Shuting_11a)Photo by Alice Hou (@Shuting_11a)
Grateful to have my perspective included in a @CoinDesk article for the first time. 🙏 Shared some thoughts on how L1/L2 DeFi dynamics are evolving. Full piece here: coindesk.com/tech/2025/09/15…
4
20
2,131
Has anyone really used obsidian + Claude code/Codex to create a “second brain”? My understanding is that obsidian organizes and connects the dots of ur knowledge base then LLMs can get higher quality and more personal context constantly. Wondering how effective it is
2
8
557
A really good article to read: substack.com/home/post/p-203… Core thesis is that the current AI boom is behaving more like the 2008 real estate crisis than the dotcom bubble. Current mega data center deals are basically loans to frontier labs: Hyperscalers are spending billions upfront to build out capacity, based on promises from labs like OpenAI and Anthropic that they'll pay for that compute over the next few years. But how are these frontier labs gonna pay if profitability is still a problem? Especially when cheaper Chinese open-weight models and better token efficiency are pushing down the cost of using AI. Right now, a lot of that spending is supported by private funding rounds at appraised valuations. The higher the valuation, the more money these labs can raise to fund the next round of compute spending. And eventually, an IPO could open the door to much more public capital. That's where the 2008 comparison gets interesting. Back then, the assumption was that housing prices would keep going up, allowing borrowers to keep refinancing. Now, the assumption is that frontier labs will keep growing fast enough to raise more money and eventually generate enough revenue to pay for all this compute. AI demand is real. But does it translate into enough revenue, fast enough, to pay for all the infrastructure we've committed to build?
1
6
344
Found this special section at a local bookstore. Too cute😆
3
7
328
The ticker is $ETH
After a comprehensive review, Aave is deprecating 50 low adoption asset reserves across multiple deployments. In addition, Aave is orderly winding down deployments on Sonic, Scroll, zkSync, Metis, Soneium, and Aptos, covering another 25 asset reserves. As part of this process, 21 matured Pendle PTs are also being deprecated in favour of new maturities. In total, the changes affect $98.1M in supply and $15.6M in debt. These measures reduce Aave's economic and technical risk surface as part of the new Aave Risk Framework and Technical Asset Listing Framework. Aave will continue applying continuous risk assessment for all assets across all deployments.
4
545
Read long articles on Substack if you wanna fix your attention span
3
13
459
Are you ready
Pulling together some of the information from various news reports and statements, we did get a little bit of new Safe Superintelligence information to speculate about today. Ilya said in a new statement to the press that their research is 'focused on overlooked aspects of how the human brain functions.' In the joint statement, after gaining assess to the lab, NVIDIA said that SSI 'has been quietly advancing a new research direction to unlock a powerful and robustly aligned artificial intelligence.' They also said the two companies will 'collaborate on the technical advancement of NVIDIA's current and future compute platforms' by 'leveraging SSI's unique insights.' Next gen NVIDIA GPUs designed by SSI's secret model? It is certainly possible. Everyone is doing this now to some degree. Additionally, Bloomberg reported that the $5 billion investment into SSI is being made directly in cash.
1
5
578
You know you picked a good man when he orders a pina colada at the bar
2
4
436
Can we stop calling addresses onchain users
2
8
318
Alice Hou retweeted
292
554
312
5,475
2,965,068
Most companies (more traditional ones) are adopting AI from a top-down approach. They pick a few tech/product people, hire consultants, and try to redesign workflows for the whole org. Meanwhile most employees dgaf about learning AI if they can just keep doing the old job faster the old way (Excel without touching a mouse, etc). But I'm sure you'll have an outlier on your team that's AI-pilled and quietly experimenting with agents to kill the boring work. The problem is taking it beyond the individual. Authorization hurdles are everywhere. Sharing is painful. And proving the results to coworkers or your boss is even harder. Bullish on @koomen @aaron_epstein's requests for startups
AI is moving into the physical world. We're excited about a new wave of startups rebuilding the systems that power the real world, from education and healthcare to defense, finance, infrastructure, and work itself. ycombinator.com/rfs
1
7
792
My husband: I think AI agents are perfect for you cuz you like to boss around all the time
5
282
Valuation is a meme. Trade the expectations.
1
1
7
511
Genuine question: How are these Chinese models outperforming the U.S. ones despite seemingly having much less funding?
Big news: Kimi-K3 by @Kimi_Moonshot is now #1 in the Frontend Code Arena with 1679 pts, surpassing Claude Fable 5. This is a 17-place jump from Kimi-k2.6 (#18 -> #1). In Frontend, Kimi-K3 ranked #1 in 6 of 7 domains: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Simulations, and Content Creation Tools, landing #2 only in Gaming behind Fable 5. The full model weights will be released by July 27. Congrats to the @Kimi_Moonshot team on this major milestone!
1
464
Maybe DeFi should never serve general retail as the front end. It should just be seen as a backend upgrade, where users can trade 24/7 with more options. Not a perfect metaphor, but DeFi should feel like an upgraded plumbing system in your apartment. You don’t know how it works, but somehow hot water comes on the moment you turn the tap instead of making you wait 10 seconds.
1
8
285
Having a cat is the most toxic relationship. He destroys your stuff and ignores you 99% of the time. But then he gives you this look and suddenly all the emotional damage is forgotten.
6
225
Make Ethereum great again.
1
4
252
Isn’t AMM a natural fit for compute capital markets? Considering the fact that computes are not fungible and currently only rely on OTCs for hedging risks AMMs were originally designed to support long-tail, less liquid assets. wondering if any team building on this
4
345
Alice Hou retweeted
You have noticed it. ChatGPT feels dumber than it used to. Your prompts that worked six months ago produce worse results now. The writing sounds flatter. The ideas sound safer. The internet itself feels like it is shrinking. Every article reads the same. Every email sounds the same. Every answer sounds like it was written by the same voice. You thought it was you. It is not you. Researchers at Oxford and Cambridge published a paper in Nature proving what is happening. They call it Model Collapse. Here is the mechanism in one sentence. AI trained on AI-generated data gets dumber every generation until it forgets what real human data looked like. The internet is filling with AI-generated content. Blog posts. Articles. Reviews. Comments. Social media. AI companies scrape the internet to train the next generation of models. Which means the next generation of AI is being trained on the output of the current generation. Each cycle loses information. Not randomly. It loses the rarest, most unusual, most creative parts first. The researchers call these the "tails of the distribution." The weird ideas. The unexpected perspectives. The things that made the internet feel human. Those disappear first. What remains is the average. The safe. The expected. The bland. Then the next generation trains on that. And loses more. And the next generation trains on that. And loses more. The researchers proved this is not a slow decline. Major degradation happens within just a few iterations. Even when some of the original human data is preserved. They tested it on large language models. On image generators. On statistical models. The pattern was the same every time. The output converges toward a narrow, flattened version of reality that looks nothing like the original data. The lead researcher put it plainly. "Large language models are like fire. A useful tool. But one that pollutes the environment." The pollution is invisible. You cannot see which sentence on the internet was written by a human and which was written by AI. Neither can the AI that is about to train on it. And once the tails are gone, they do not come back. The damage is irreversible. This is not a prediction anymore. It is a diagnosis. The internet you grew up on was built by humans writing things no algorithm would have written. Strange, personal, imperfect, alive. That internet is being diluted. One generation of AI at a time. And the models trained on what remains are learning a smaller and smaller version of the world. Model Collapse is not a technical problem. It is a cultural one. The thing that made the internet worth reading is the thing that disappears first.
1,111
6,229
1,370
17,416
2,314,390