Original Title: Is AI & Robotics Stealing Crypto's Thunder? | EP 73
Original Source: Good Game Podcast
Original translation by Wu Says Blockchain
In this episode of (GoodGame), AllianceDAO co-founders Imran and Qiao delve into how the tech community's focus has quickly shifted to AI and robotics, overshadowing the current innovation cycle in cryptocurrency. They evaluate the stagnation narrative surrounding stablecoins and tokenization in crypto while highlighting the competitive landscape dominated by trading platforms and meme-driven token issuance.
Robinhood emerges as a strong competitor, bridging fintech and tokenized assets. Discussions cover the dynamics of emerging startups, AI's impact on coding and education, and the prospects of junior roles potentially facing extinction in law, medicine, and finance, ultimately painting a future reshaped by AI-native, founder-led micro-startups that redefine industry norms. This article is edited by Wu Says Blockchain.
The AI and robotics craze overshadows the current state of cryptocurrency.
Imran: Welcome to (GoodGame), where we provide no-nonsense insights for cryptocurrency founders. We've experienced ups and downs. I think there are many exciting new releases in the AI and robotics space that are overshadowing the dynamics in cryptocurrency.
Qiao: Indeed.
Imran: There will be some innovation periods that attract a lot of attention and mindshare. I think we are currently in a phase like AI and robotics, while cryptocurrency is undergoing a rebuilding process. If you're looking for the next startup idea in cryptocurrency, check out our startup request list and get inspired at align.xyz/ideas.
Qiao: Currently in the cryptocurrency space, whether builders or investors— even on crypto Twitter, there's almost only one narrative.
Imran: Stablecoins and tokenization are these two narratives.
Qiao: But it's the same story. It's like just putting traditional assets on-chain.
Imran: Now it seems this has reached a consensus. Yes. We see many institutions launching their own stablecoins. I think Fidelity just announced their own stablecoin recently. Who else? I mean, there are a bunch of companies trying to do exactly the same thing. From an innovation perspective, there doesn't seem to be anything truly new; they're just reinventing the wheel.
I'm planning to post this meme. I tried to get ChatGPT to generate a Ghibli-style meme. But essentially, what I want to express is that the competitive environment is actually very small. Everyone is fighting for each other's profits. If you look at Binance, they've launched two very large indirect statements. The first is, I don't know if you've seen the Hyperliquid incident, but the high-level summary is that there's a token called Jelly on Hyperliquid.
Intense Competition Among Trading Platforms Highlights Challenges for Hyperliquid
Imran: Hyperliquid's trading volume and liquidity have reached a manipulable level. Someone opened a short position to extract value from HLP. OKX and Binance publicly participated in this— posting about it on Twitter— many interpreted this as them wanting Hyperliquid to fail because it's seen as a competitive threat to their derivatives business.
Another example: Binance is listing tokens from FourMeme, a PumpFun clone running on the BNB chain. Tokens like Mubarak and Broccoli are being listed, sparking criticism of Binance's direction. But the logic is clear— they view platforms like PumpFun as competitors. Now, Binance is actively competing with Hyperliquid, PumpFun, Jupiter, etc.— this is a war between on-chain and off-chain.
Qiao: Everyone is targeting the same verticals— trading and stablecoins. In trading: crypto-native platforms and Robinhood. In stablecoins: Circle, Tether, and traditional fintech companies.
Robinhood's Strategic Move Towards Tokenization
Imran: Robinhood recently held an Apple-like launch event emphasizing their commitment to "everything tokenized." With their vast distribution network, they are in a favorable position in the market. They also want to tokenize real estate— similar to a tradable OpenSea-style MLS— allowing users to trade risk exposure to real estate assets.
Qiao: But not physical real estate— they might offer derivatives tied to real estate, which aligns with their strengths.
Imran: That's right— they deeply understand their user base. Additionally, they launched a cash delivery service— you can request physical cash instead of going to an ATM in person. At first, I wondered who else was using cash, but many people actually are— contractors, barbers, etc.— mainly for privacy reasons.
Qiao: I understand that appeal. Banks ask some invasive privacy questions when you try to withdraw large amounts of cash. It's frustrating. I just don't quite understand why Robinhood wants to enter this space.
Robinhood's Cash Delivery and Banking Services
Imran: You know, this might be another angle— that could be because the ATM business is still very profitable. If you can capture that...
Qiao: Are they trying to become banks? They want people to deposit money. Yes, they want people to deposit so they can lend and earn the spread, essentially like a bank.
Imran: That could be a potential reason— essentially, "take your money out of the bank and give it to us." Then add this cash delivery service. But they also offer other incentives. Recently, they announced a promotion— I'm actually considering this. If you transfer your brokerage account to them, they'll give you an extra 2% bonus. So I'm really considering moving it over. It's not a bad deal. So they are pushing hard— traditional business, banking services, traditional brokerage accounts, and now even competing with crypto trading platforms. Their reach extends into three areas.
Qiao: I think Robinhood is the biggest threat to Coinbase.
Robinhood vs. Coinbase: The Battle of Platform Distribution
Imran: I think Robinhood is the biggest threat to crypto startups. Because all the services they're offering now— whether tokenizing X or Y— are exactly what we want our founders to build. If Robinhood becomes the main distribution point, it essentially cuts off the middlemen— in this case, the startups. So if we look to the future and examine the current competitive landscape, it actually feels very small. And when you factor Robinhood and other major players into the mix, it becomes even smaller.
Qiao: When you say "small," do you mean there are only a handful of players competing for the same opportunities?
Imran: Yes. Hyperliquid.
Qiao: Binance, Coinbase...
Qiao: Pump. Right now, the launch trading volume handled by Pump is about 30%, right? That's quite significant— quite large.
Imran: Yes, it's significant. I'm not sure of the latest numbers, but the 24-hour trading volume is around $294 million. Annual revenue is roughly $30 billion. In contrast, Raydium's AMM trading volume is $332 million. However, trading volume has slightly declined— it's now close to $200 million. Essentially, Pump has surpassed Raydium in trading volume. That's the point I want to make.
Liquidity Fragmentation Between Trading Platforms and Aggregators
Qiao: What data are you looking at? DeFi Llama?
Imran: Yes. Raydium's trading volume mainly remains stable, primarily due to previously issued tokens.
Main.
Imran: Right, but that might change. Pump is gaining attention. Jupiter launched Meteora (an AMM), and there's also Moonshot. They're all part of the same founder network. Jupiter acts as an aggregator. This market is getting crowded, and everyone is chasing the same marginal users. The real challenge now is expanding the market— how do we onboard new users into crypto? That's the next frontier. Before that, it's a game of waiting for the next wave of meaningful innovation. An interesting experiment is Noise on MegaETH. They use Kaido as a trend oracle— kind of like fantasy sports, but for trend trading. This is part of our long-discussed theory of "monetizing attention."
Qiao: Yes, this isn't fresh; it's just a repackaging of the same idea. There's nothing transformative yet.
Imran: Almost.
Circle's Strategic Dilemma: Dependence on Coinbase, Global Competition, and On-Chain Transformation
Qiao: Circle's IPO valuation is between $3 billion and $5 billion— slightly lower than I expected, but maybe that's just my bias.
Imran: Reasonable. What's striking in the application is that Circle pays 50% of its reserve income— from government bonds and other income-generating assets— to Coinbase. Last year, Circle's revenue was about $1.8 billion, with $800 million paid to Coinbase. That's huge. It shows a clear power dynamic: Coinbase holds the user relationships and is the primary distribution channel for USDC. It's like Pump has its end users and uses them for AMM.
Qiao: Indeed. If government bond yields decrease, that poses a survival risk for Circle. Perhaps that's part of the reason they're pushing for an IPO now. It's also worth noting— Coinbase refunds most of the USDC earnings (about 4%) back to users. So Circle pays Coinbase for distribution, and Coinbase uses it as a user acquisition tool, ultimately profiting through higher-margin trading and other businesses.
Imran: Circle seems to be fighting a losing battle. While they currently have a regulatory moat, traditional financial competitors are emerging.
Qiao: USDC dominates in the US, but globally, USDT remains king.
Imran: Yes, a founder from Colombia told me that USDC is basically irrelevant in Latin America. Despite offering better products, it hasn't gained traction.
Qiao: That's the network effect of currency. USDT has existed for ten years, and everyone trusts that brand— especially on Tron, where USDT has a significant share.
Imran: So what is Circle's ultimate goal? They sit between regulators and Coinbase, and while USDC has been adopted in DeFi and among some founders, is that enough?
Qiao: From the outside, Circle seems to be in a tough business. I'm not sure what their next step is.
Imran: They seem to be leaning more towards on-chain infrastructure.
Qiao: The best thing they've built is CCTP— the cross-chain USDC bridge. It's fast, with settlement times of under a minute, and it's basically free. Since it's centralized, they just need to move database entries internally. Honestly, I trust CCTP more than some decentralized bridges.
Imran: I'm also using Debridge and Wormhole— both are quite good. But Circle's solution might be the best in terms of speed and reliability. There are also two notable acquisitions— Kraken acquired NinjaTrader, a major traditional derivatives platform, and there are rumors that Coinbase made an offer for Deribit. To me, this reflects a broader trend: existing companies are expanding beyond pure crypto trading, which is no longer enough to meet growth expectations. Innovation seems stagnant, so trading platforms are doubling down on adjacent fields. Interestingly, some are turning to traditional finance, while others— like Robinhood— are returning to crypto with a refocused attitude.
Qiao: Yes, I feel the same. The line is clearly starting to blur.
Market Sentiment Check: Saylor, GameStop, and the Liquidity Cycle
Imran: Quick market sentiment check— Michael Saylor just raised another $2 billion, and GameStop followed up with a $1.6 billion offering.
Qiao: Saylor's moves are hardly news anymore. But yes, I checked his risk— his liquidation risk is small unless Bitcoin's price drops below $20,000 within months.
Imran: Right, he holds about 3% of the total BTC supply. He's aggressive but also safe.
Qiao: So what phase are we currently in the cycle? A lot may depend on what Trump says today— policies may change rapidly. As for stocks, I believe only a handful of US stocks are reasonably valued. Google is one of them— it's reviving in AI, and the valuation is still reasonable. I'm also optimistic about Tesla. I bought my Cybertruck at the end of last year, and in just a few months, FSD has almost become capable of hands-free driving. That made me decide to double down. Musk also plans to launch the Optimus humanoid robot in 2026 to 2027. For me, Tesla is the most liquid venture investment in AI, autonomous driving, and robotics.
Imran: Agreed. If you're looking for broader exposure in fintech and tokenization, Robinhood is also interesting. I might lean towards it over Coinbase right now, although that might change based on acquisitions.
Escalating AI Competition: Google's, Tesla's, and China's Involvement
Imran: I've seen videos of humanoid robots in public spaces in China— while they're not general-purpose, they're still impressive.
Qiao: Yes, it might be specific task robots. But I agree— China is clearly involved.
Imran: If investors want to tap into Chinese innovations, there are now more options.
Qiao: And the prices of Chinese tech stocks are far below those of US stocks. I hold some— Tencent and Pinduoduo (PDD) are my favorites. Tencent is doing serious AI work in terms of data scale. PDD is founder-led, growing quickly, and still feels like a startup.
Imran: Makes sense. So, the tech competition between China and the US is unfolding in real-time.
Qiao: In the AI model space— have you tried Mistral? We use it internally at Alliance. For deep research tasks, it clearly outperforms ChatGPT. Perplexity is also quite good in research— it quickly became one of my go-to tools.
AI Assistants Showdown: Gemini, Mistral, and Perplexity
Imran: A few days ago, someone compared Mistral to Google's Gemini on Twitter. They said if it took five minutes to do deep work with Mistral, it only took two and a half minutes with Gemini. So I've been using Gemini more.
Qiao: Gemini is indeed quite good. Its performance is surprisingly robust.
Imran: Yes, I've started switching. Every few weeks, I feel the need to change my workflow based on the latest tools.
Qiao: Me too.
Imran: There's currently no moat. So I've been using Gemini more frequently. Last night, I tried their new feature— real-time conversations. This is part of the latest updates. You can actually brainstorm with AI in real time. For example, I wanted to plan a birthday party— it helped me brainstorm, find a venue, estimate costs, and I even had it help me make calls.
Qiao: Did it do that?
Imran: Not yet. But it got me to a point where all I needed to do was make that call myself. It automated everything to that point. Super useful.
Qiao: Can ChatGPT do that too? Or not quite?
Imran: From what I understand, Gemini feels faster and more interactive. Have you tried Sesame AI? It's the most advanced voice model I've seen— it feels like you're talking to a real person.
Qiao: Who's behind Sesame?
Imran: Not too sure— no big company has claimed that yet. But I've tried both of their models, and they left a strong impression. I had my wife try it out, and she found it really eerie. She said it felt like talking to a real person. It even simulates breathing— it's amazing.
Qiao: Turing test passed. That's crazy.
Imran: Yes, very compelling.
Qiao: What you mentioned about Google is the only US stock I currently hold— Google. Its forward PE is around 17— not cheap, but still reasonable compared to the other "seven giants." For the first time in over a decade, Google is back at the forefront. They lost to Facebook in social media and couldn't keep up with Apple in mobile. But in AI? This is Google's turf. This is deep software technology, and Google excels at it. I wouldn't be surprised if by the end of this year, Google has the best AI model.
Imran: It's said that Google co-founder Sergey Brin has returned to Google and is working overtime to ensure they win in AI— at least, that's what I've seen on Twitter. What I like about Google is that when I conduct deep research, it pulls information from sources like YouTube and podcasts— real-time content. This gives it a unique advantage. Sometimes I think, "This research would actually be better on Google than on ChatGPT or Mistral."
Qiao: That's a good point. I still use a combination of all the top AIs— ChatGPT, Perplexity, Claude, Gemini, DeepSeek, etc. I just switch based on context. And while I can't always articulate the reason clearly, I instinctively know which tool to use in each case. It's like I've built my own mental neural network for choosing the right AI assistant. By the way, OpenAI just raised another round of funding— $300 million at a $30 billion valuation.
Imran: Exactly, it's massive. Clearly high-profile.
Qiao: The AI competition is intense— at both the foundational model level and the application level. All the major players— the "seven giants"— are heavily investing in AI, except for Apple. Apple may also be investing heavily, but they haven't released any products yet. Meanwhile, Amazon just released their own AI assistant product for their browser.
Imran: Exactly. They're officially entering this space, competing with ChatGPT, Anthropic, and others.
Qiao: I also saw a recent interview with Jeff Bezos— he is clearly all in on AI.
Imran: Yes, he didn't explicitly say he's back at Amazon, but when asked about Amazon's biggest focus, he immediately answered: "AI."
Qiao: That says it all.
Imran: I switch based on the best tools every few weeks. Recently, I tried Gemini's new real-time conversation feature— it helped me plan a birthday party, from location to pricing. It was very interactive.
Qiao: Can ChatGPT do the same?
Imran: It could be, but Gemini feels faster and more agile. Have you tried Sesame? It's a voice model that feels very human-like— even simulating breathing.
Qiao: That's crazy.
Imran: That's impressive. In AI, Google is back in the game. Their forward PE is reasonable, and they are finally playing a role in the forefront of competition.
Qiao: Agreed. They stumbled in social and mobile, but AI is their turf— deep software. I wouldn't be surprised if Google leads in AI by the end of the year.
Imran: Sergey Brin is reportedly back at Google, pushing hard. Plus, Google's advantage in pulling real-time data (like YouTube, podcasts, etc.) gives it an edge in research applications.
Qiao: Yes, I rotate tools for tasks— ChatGPT, Gemini, Perplexity, Claude, DeepSeek, etc. This has become instinctual. Also, OpenAI just raised $300 million at a valuation of $30 billion. All major players are deeply diving into AI except Apple. Amazon recently launched a browser-based assistant, and even Bezos seems to be re-engaging.
Imran: With Sergey and Bezos back, it feels like the most competitive and exciting tech race we've seen in years.
Will AI make coding obsolete? The clash of productivity and expertise.
Qiao: The CEO of Replit recently said, "I no longer think you should learn to code." This sparked a lot of controversies. But I think what he means is that you no longer need to learn coding— it's no longer a barrier. Within Alliance, I asked tech founders how AI tools like ChatGPT, Cursor, and Windsurf have improved productivity. Most reported a 2 to 4 times boost— huge gains. Additionally, more non-technical founders are now using Cursor to quickly develop their front ends. Previously, they had to explain their visions to technical partners, and a lot was lost in translation. Now they can prototype their ideas directly and hand them off to technicians for refinement. This is a massive shift— it allows visionaries to go directly from idea to product.
Imran: Me too. I'm not technical, but I recently built a Telegram bot using Mistral and deployed it on AWS. It opened my eyes. If I can do that, imagine what today's builders can do.
Qiao: Indeed. At this pace, you don't need a large team or massive funds. A few people and AI tools might be enough to build a unicorn.
Imran: I spoke with a founder from Proof— they are building rapid development tools for crypto projects. Their target users are founders building MVPs, which makes a lot of sense. These tools drastically shorten feedback cycles, making it easier to test ideas and iterate quickly. This could significantly accelerate the pace of innovation.
Qiao: Another big opportunity is consumer-facing micro-apps. With tools like Cursor, anyone can build a highly personalized app. One of our founders spent two hours building a language learning app for himself— and it works better than Duolingo because it's built entirely around his needs.
Imran: That's a great example. Independent developers like @levelsio are also doing this— he posted about rebuilding Skype features and is now making $10,000 a month. This shows you can solve a niche problem and become profitable immediately— especially with a global user base.
Qiao: Indeed. These tiny opportunities used to often be overlooked. But as AI tools lower the barriers, even non-tech people can build applications that can genuinely be profitable— $10,000, $50,000 a month— without needing big teams or funding. This is the next wave.
For Tokenizing Liquidity of Micro-Startups
Imran: I think we're entering the "micro-unicorn" season— we'll see hundreds, even thousands of small-scale startups emerge. These companies will solve significant problems for small user groups. But they may remain small-scale. They may never become billion-dollar companies. In the end, there might be some consolidation, but for now, this is the shape of the future.
Qiao: And the question is, these small startups will never go public. So if they— or their investors— want liquidity, what do they do?
Imran: That's exactly why they should be tokenized. It's a perfect use case. Tokenization could become the liquidity engine for micro-startups. In fact, I see another trend. I'd love to hear your thoughts. There's a guy— I forgot his name— but he launched an app called Cal AI. I downloaded it. Basically, you take a picture of your food, and it automatically calculates calories, fat content, etc. Then, it gives you a weekly health plan based on your calorie intake. He's only 18. Rejected by all universities. He shared his story online— posted his Oura Ring data, etc.
Qiao: Yes, I saw that.
Next-generation builders choose to skip school.
Imran: A teenager recently launched a product and reportedly reached $10 million in annual recurring revenue— despite being rejected by all universities. And now, more and more young people are choosing to skip college and dive straight into building.
Qiao: For this generation, traditional education is starting to feel optional.
Imran: Indeed. It's not that education isn't important— but building what people want may be more critical. I saw Nik mention on Twitter that perfectly summarizes this:
"Working at Google or attending Harvard doesn't signify status— because those decisions come from small, biased committees."
"True status comes from something that can't be faked: building products used by millions, taking a company public, or saving lives."
"You don't need that job at Google— true status is when Google has to acquire your company."
Qiao: I completely agree. We're seeing younger and younger founders. In our next batch, there are kids dropping out of Princeton University in their sophomore year— they are incredibly smart.
Imran: They might not even know what they're building yet, but the energy is undeniable. I think this trend will continue.
Qiao: And education itself is evolving. We're working with a team to build AI tools that help kids learn to read earlier. There's also a new startup school called Alpha School. Have you heard of it?
Disrupting Traditional Education with AI Tutors
Qiao: Have you heard of Alpha Schools in Texas? There, kids learn through AI tutors, spending only two hours a day on academics— yet they outperform students in traditional schools.
Imran: I saw that. Their students ranked in the top 2% nationally on tests.
Qiao: I'm not surprised. AI as a tutor is incredibly effective. I believe that within 5 to 10 years, AI will completely disrupt K-12 education. Kids will no longer need to endure eight-hour lectures; they'll learn the same material in much shorter time frames— AI tutors will personalize teaching according to their pace and interests. This will be far more efficient than traditional classroom teachers.
Imran: I can't stand the current school model— eight hours of inefficient learning. Students are passive, and most of the time they're not productive.
Qiao: Exactly. And Alpha School goes even further: 8-year-olds are launching startups, 10-year-olds are giving TED-style talks, and 12-year-olds are studying Harvard Business School case studies. This is real applied learning— not just rote work.
Imran: This is the most exciting case of educational innovation I've seen. The whole system needs to change; it feels like the future.
AI Health Assistants and the End of Primary Care Physicians
Qiao: I see Apple is exploring an AI doctor product— that makes sense. The Apple Watch collects heart, sleep, and other health data, and they are in a good position. Besides Apple, there are huge opportunities for startups to build AI health assistants. One of our team members has been regularly using ChatGPT for health advice— regarding heart health, diet, sleep, etc. Imagine an AI pulling data from your wearables, medications, and lab results to provide tailored health advice. That's incredibly powerful. A recent personal example: I took my son to an allergist. The doctor was rushed and didn't answer all my questions. So I asked ChatGPT at home— it gave the same diagnosis and better suggestions. It felt like having a second, more patient doctor.
Imran: Indeed. This has already happened. My uncle in India fell ill, and no one at home could understand the medical documents. I uploaded them to ChatGPT and got a clear summary and action plan, which I then relayed to them. This clarity was something that actual doctors never provided.
Qiao: That's why I think some types of doctors— especially primary care physicians— may ultimately be replaced by AI.
Imran: Telemedicine is similar. It's a step in that direction, but AI will go further— providing scalable personalized care. At the very least, AI can offer second opinions and reduce misdiagnoses.
Qiao: I wonder which startups are leading in this space right now. There may be many, but no standout leaders yet. Maybe Apple will be the first to get there.
Legal, Analyst, and Junior Roles Face Threats
Imran: I mean, lawyers in general— especially junior lawyers— most of their work can already be handled by AI.
Qiao: Exactly, it's the junior roles that are the most vulnerable.
Imran: Yes, I can definitely see this industry taking a significant hit. Even us— junior analysts or researchers— we no longer really need them. So I think entering these types of industries will become more challenging for those looking to build a career.
Qiao: They have to find new ways to build reputation.
Imran: One way might be to showcase your thinking on Twitter. I've seen people find jobs that way— just keep tweeting, sharing sharp insights. I read a discussion and think, "This person should absolutely work somewhere," and sure enough, weeks later they announce they're hired.
Qiao: It will be interesting to see how this evolves. The broader trend I see is that AI is replacing many junior roles across multiple industries— not just doctors and lawyers, but in our field as well. Remember when we talked about hiring venture capital analysts? We no longer need them. A lot of the work they would do— market research, startup analysis— we can accomplish just by talking to AI.
Generalists vs. Experts in the Age of AI
Qiao: Interestingly, I think Kyle tweeted something along the lines of "AI's enhancement is greater for generalists than for specialists." I get his point, but conversely: in any specific industry, the most experienced individuals— true domain experts— still cannot be replaced by AI. They have unique insights, edge cases, and experience-based intuitions that AI lacks. And those people are the experts.
Imran: Yes, it's a diversified shift. Generalists can deeply focus on one area, while specialists can expand across fields. So I believe AI empowers both sides. It's not one against the other. For me, this ultimately comes down to agency. If you're willing to learn, adapt, and delve deeply or broadly based on circumstances, you'll succeed. If you lack that drive, you'll fall behind— especially as others more actively leverage AI.
Qiao: Yes. Naval recommended a book about parenting that touched on this point— I can't recall the title at the moment.
Imran: Yes, I remember. He shared two main points. I actually have that book now— it's really interesting.
Teaching Kids Agency Through AI and Books
Qiao: The core idea in the book recommended by Naval is to teach kids agency early on. Let them make their own decisions— even if it's not ideal in the short term— so they can learn from experience and understand the reasons behind choices. This is far more effective than just saying, "Do this, don't do that."
Imran: Naval shared an example: he lets his kids eat whatever they want— as long as they read and do math for an hour each day. Eventually, one of his daughters stopped eating ice cream on her own, realizing it wasn't healthy. That impressed me. I adopted part of that model. I bought Kindles for my kids and told them, "If you read, you can watch a bit of TV." We usually don't allow them to spend too much time on screens, but now they read almost two hours every day without prompting. They love it.
Qiao: That's good. My approach is similar— two hours of math and language each day, then they can freely explore what interests them.
Imran: Me too. My daughter loves reading and uses Khan Academy independently. My son is a builder— we recently got a robot kit powered by NVIDIA's new GPU, and we’re assembling it together. It's an opportunity for teaching and learning to happen side by side.
Qiao: This is exactly the way of learning I want to encourage.
Imran: Exactly, that's my thinking about the future— practical, curiosity-driven learning.
Original Link
