<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by AngleFeint on Medium]]></title>
        <description><![CDATA[Stories by AngleFeint on Medium]]></description>
        <link>https://medium.com/@anglefeint?source=rss-ceb284c16b37------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*E8vt4g1UGCQgdra5</url>
            <title>Stories by AngleFeint on Medium</title>
            <link>https://medium.com/@anglefeint?source=rss-ceb284c16b37------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 01:16:32 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@anglefeint/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Reflections on the $6.3]]></title>
            <link>https://anglefeint.medium.com/reflections-on-the-6-3-886f2e6c9cae?source=rss-ceb284c16b37------2</link>
            <guid isPermaLink="false">https://medium.com/p/886f2e6c9cae</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[crypto]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[security]]></category>
            <category><![CDATA[cypherpunk]]></category>
            <dc:creator><![CDATA[AngleFeint]]></dc:creator>
            <pubDate>Sat, 07 Nov 2026 08:33:52 GMT</pubDate>
            <atom:updated>2026-11-07T14:41:24.865Z</atom:updated>
            <content:encoded><![CDATA[<h3>Reflections on the $6.3 Billion Binance Theft: The True Ideals of the Cypherpunks and What AI Will Bring to the Crypto World</h3><p>On cypherpunk ideals, centralized exchanges, and how personal AI agents could reshape cryptocurrency, privacy, and security.</p><p><strong><em>Originally published on</em></strong><em> </em><a href="https://anglefeint.com/en/blog/cypherpunk-ideals-ai-and-the-future-of-crypto/">https://anglefeint.com/en/blog/architecture-grows-with-your-business/</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*u1qWrESPicaVSXUo-5Qhrw.png" /></figure><p>Last week, I saw some big news on the internet: hackers had stolen $6.3 billion in cryptocurrency from Binance. It was the largest cryptocurrency theft in history.</p><p>I really don’t like the North Korean regime, especially its hackers. But I think CZ and Binance have enough money to cover the losses. Honestly, the news made me think about quite a few things.</p><p>I’ve known about cryptocurrency for a long time. One day, I started learning more about Q-day. During my research, I came across a new word: *cypherpunk*. I became curious about what cypherpunks were, what cypherpunk culture was, and what the cypherpunk mailing list was all about.</p><p>Then I found an interesting timeline:</p><p>| Year | Person | Technology or Idea |<br>| — — — — — — | — — — — — — — — | — — — — — — — — — — — — — — — — — — |<br>| 1980s | David Chaum | Anonymous electronic cash technology |<br>| 1989 | David Chaum | DigiCash |<br>| 1996 | Nick Szabo | Smart contracts |<br>| 1997 | Adam Back | Hashcash (Proof of Work) |<br>| 1998 | Wei Dai | b-money |<br>| Around 1998 | Nick Szabo | Bit Gold |<br>| 2004 | Hal Finney | RPOW (Reusable Proof of Work) |<br>| 2008 | Satoshi Nakamoto | Bitcoin |</p><p>And so on.</p><p>So many technologies that people call “new” today actually have a long history behind them.</p><p>Then I found some old writings from the cypherpunk community:</p><ul><li>[A Cypherpunk’s Manifesto](<a href="https://www.activism.net/cypherpunk/manifesto.html">https://www.activism.net/cypherpunk/manifesto.html</a>)<br>- [The Crypto Anarchist Manifesto](<a href="https://www.activism.net/cypherpunk/crypto-anarchy.html">https://www.activism.net/cypherpunk/crypto-anarchy.html</a>)<br>- [Virtual Communities](<a href="https://nakamotoinstitute.org/library/virtual-communities/">https://nakamotoinstitute.org/library/virtual-communities/</a>)</li></ul><p>In the third article, I found this sentence:</p><p>&gt; “Transparent links and nodes are the norm in a police state, such as the U.S.S.R., Iraq, China, and so forth.”</p><p>That’s interesting. Even back in 1994, the writer already knew that China was a police state. Wow, that really shocked me.</p><p>After reading more about cypherpunks and their ideas, I began to feel that their thinking came from the confrontation of the Cold War: the Free World vs the Communist World.</p><p>When we look back at history, we can see that the cypherpunks grew up during the Cold War era.</p><p>During the Cold War, the communist world relied heavily on spies, secret police, and surveillance. There were communist spies all over the world, including in the United States. Intelligence operations and political control were deeply rooted in the communist system.(The communist tradition has long been defined by espionage and betrayal. Backed by 50 million German gold marks, Lenin returned to Russia to wipe out the Tsar’s entire family — sparing neither the young children nor their servants and cook. In this realm, the infamous ‘swallows’ — female honey-trap operatives — became another signature export of communist tradecraft.)</p><p>The Free World had to fight against the Communist World. To win that fight, Western governments also developed powerful intelligence agencies, surveillance systems, and ways to intercept communications. In some ways, they learned some of the same bad habits as their opponents.</p><p>But there’s a very important difference.</p><p>The Free World still gave people room to do what they believed was right. Even when people disagreed with their own governments, they could speak out, organize, and resist. In the communist world, governments monitored everyone and didn’t allow people to resist. If they discovered resistance movements or organizations, they could imprison or even kill the people involved.</p><p>But in the Free World, even if you resisted the government, you still had space to do so.</p><p>As the saying goes, “Where there is power, there is resistance.”</p><p>When the Free World began adopting some of the surveillance methods of the Communist World, resistance grew from within the Free World itself. And I think the cypherpunks were part of that resistance.</p><p>They didn’t believe that governments and big companies would protect their freedom simply out of kindness. They wanted to build technology that could protect freedom and privacy without depending on the goodwill of those in power.</p><p>Many cypherpunks were influenced by libertarianism, and from their writings, we can see deep opposition to authoritarianism and totalitarianism was a central part of their ideals.</p><p>And here’s something I find ridiculous.</p><p>The cypherpunks came from the Free World. They hated authoritarianism and totalitarianism, yet they wrote code to challenge the governments of the Free World itself. They didn’t blindly trust their own governments just because those governments claimed to stand for freedom.</p><p>But look at what happened in the real world. The cryptocurrency industry has grown enormously in the Free World, while governments in the communist world are afraid of cryptocurrency and place severe restrictions on it.</p><p>And here’s the irony: the cypherpunks challenged the governments of the Free World, but their technology ended up becoming a much bigger threat to authoritarian and totalitarian regimes, including communist ones. Isn’t that funny?</p><p>I still remember that my first company used to accept Bitcoin, until the Chinese government told it to stop. One of my coworkers once sold some cryptocurrency and exchanged it for Chinese yuan, and his bank account was frozen.</p><p>So when I read about cypherpunk history, I keep finding connections to the Cold War and the struggle between freedom and government control. That history helps explain why the Chinese government banned so many cryptocurrency activities.</p><p>North Korea is another interesting example. Its government tightly controls financial activity at home, while North Korean hackers have become notorious for stealing cryptocurrency abroad. Apparently, cryptocurrency is unacceptable for North Korean’s ordinary people, but very useful for hackers working for the government.</p><p>What a ridiculous situation.</p><p>The more I learned about cypherpunks, the more I understood their ideals. I think the struggle for personal freedom and privacy is one of the most important parts of their thinking.</p><p>And decentralization is an important way to achieve those goals.</p><p>**Decentralization is a means, not an end.**</p><p>## Why Did Centralized Cryptocurrency Exchanges Become So Big?</p><p>But in the real world, centralized cryptocurrency exchanges became some of the biggest companies in the crypto industry. Binance is one example.</p><p>I’ve seen people joke about this contradiction: cryptocurrency is supposed to be decentralized, but people keep their money in centralized exchanges. They say they don’t trust centralized institutions, but then they trust another centralized company with their assets.</p><p>After reading and thinking about cypherpunk history, I began to believe that we’re simply living through a transitional stage.</p><p>And the reason is easy to understand.</p><p>How many people in the world own cryptocurrency? And how many of them actually understand blockchain technology, digital money, and how to use cryptocurrency directly from their own computers or phones?</p><p>How many wealthy countries have completely open gambling markets? How many countries allow people to freely exchange their money for foreign currencies? How many people want to buy or sell something without others knowing about it? And how many people simply want to protect their privacy?</p><p>And yes, how many people want to launder money or carry out transactions that governments don’t allow?</p><p>These are different kinds of activities, of course. Some are ordinary financial needs, some are restricted, and some are illegal. But they all help explain why people might look for financial services outside the traditional system.I also remember the cypherpunks pointing out that laws are different from country to country. Something allowed in one country might be forbidden in another.</p><p>I think cryptocurrency exchanges became successful partly because they met needs that traditional financial systems didn’t fully satisfy. Gambling, speculation, currency exchange, cross-border transfers, privacy, and even money laundering all created demand for cryptocurrency.</p><p>But there was another very important reason: centralized exchanges made cryptocurrency easy for ordinary people to use.</p><p>Most people don’t want to learn how blockchains work. They don’t want to manage private keys, understand different networks, or learn complicated software just to buy or sell something. They want to open an account and get things done.</p><p>Centralized exchanges solved that problem, and they made a lot of money doing it.</p><p>Now imagine a different world. Imagine that countries allowed people to gamble freely, exchange currencies freely, and move their own money without unnecessary restrictions. Imagine that governments and big companies didn’t constantly collect personal information, monitor communications, or interfere with private financial activities.</p><p>In that world, would centralized cryptocurrency exchanges have become as big as they are today?</p><p>I don’t think so. If people could already satisfy those needs through ordinary financial services, why would so many of them need cryptocurrency exchanges?</p><p>Of course, centralized exchanges would still have customers. They make trading convenient and bring buyers and sellers together. But I believe their room for growth would be much smaller than it has been in our world.</p><p>And centralization has some obvious disadvantages.</p><p>Centralized exchanges require KYC, and users may face account restrictions, withdrawal delays, or other limitations. More importantly, they concentrate enormous amounts of cryptocurrency under the control of a single organization.</p><p>That’s a very attractive target for hackers.</p><p>Look at the recent Binance theft. When one company controls billions of dollars in cryptocurrency, hackers have an enormous reason to attack it. Some of those hackers may have access to huge amounts of money, technical expertise, and even state support.</p><p>Think about it from a hacker’s point of view. If you had the ability to organize a sophisticated attack, would you rather spend months or years trying to steal a few thousand dollars from an ordinary person, or attack a company that controls billions?</p><p>**The amount of money concentrated in one place determines how much effort an attacker is willing to invest.**</p><p>Now imagine that centralized cryptocurrency exchanges didn’t exist. Everyone kept their cryptocurrency in their own wallets, spread across millions of personal computers, phones, and other devices.</p><p>Would hackers still attack people? Of course. But each wallet would usually contain much less money than a major exchange, so the incentive to spend enormous resources attacking one particular wallet would be much smaller.</p><p>And this brings me to AI.</p><p>## AI Might Change the Crypto Industry Again</p><p>We’re already living in the AI age, and I believe some major changes are coming.</p><p>Earlier this year, I read some comments from CZ about AI and cryptocurrency. He said that AI agents would find it easier to use cryptocurrency because they couldn’t easily use traditional financial systems with KYC requirements.</p><p>I understand that argument, but I don’t completely agree.</p><p>Why couldn’t AI agents have their own identities in the future? Why couldn’t they own property or manage money through some kind of recognized identity? We don’t know how future laws and financial systems will work. So AI agents may well be able to pass KYC in the future, contrary to what CZ suggested.</p><p>Here’s another interesting thought.</p><p>We assume that tokens are like blood for an AI agent. The agent works and earns money, then uses that money to pay for its computing power, memory, storage, and other expenses. If it earns enough money, it can keep running, upgrade its own capabilities, and continue working.</p><p>If an AI can earn enough money to support itself, maintain itself, and improve itself, what exactly would separate it from a human being in economic terms?</p><p>That’s a much bigger question, and maybe I’ll write about it another day.</p><p>For now, I believe AI agents will find it easier to use cryptocurrency before they can independently use traditional money.</p><p>But there’s something I think CZ’s argument misses.</p><p>**Just because AI agents can use cryptocurrency doesn’t mean they’ll use centralized cryptocurrency exchanges.**</p><p>They might use Binance. They might use a DEX. They might trade directly on a blockchain, use Hyperliquid, or use some new trading system that doesn’t even exist today.</p><p>Why should we assume that AI will choose CEXs?</p><p>In fact, I believe DEXs, Hyperliquid and other forms of direct blockchain trading will grow faster than CEXs in the AI age.</p><p>I’m not saying CEXs will stop growing. I believe more people will use cryptocurrency in the future, so both CEXs and DEXs can grow. My point is that DEXs may capture a larger share of that growth.</p><p>Let me explain why.</p><p>Earlier, I mentioned that one of the biggest reasons people use centralized exchanges is convenience. Ordinary people don’t want to understand all the technical details of cryptocurrency.</p><p>But what happens when people have AI agents that can understand those details for them?</p><p>At the moment, in 2026, many personal computers and mobile devices still don’t have enough computing power to run the most capable AI models locally, This is often called edge AI or on-device AI. But look at how technology has developed over the years.</p><p>Hardware will become more powerful. AI models will improve. Local and cloud storage will become larger and cheaper. Network speeds will increase. And software will become easier to use.</p><p>I believe personal AI agents will become much more capable as these technologies improve.</p><p>Imagine having a capable AI agent running on your own computer. It could use your local computing power, access local or cloud memory, and work with software installed on your device. You could keep your own hot wallet or cold wallet, and use fingerprints, facial recognition, or other methods to authorize important actions.</p><p>I also believe cryptocurrency projects will start providing tools designed specifically for AI agents. They might offer Agent Skills, workflows, APIs, local wallet software, or combinations of local and cloud services.</p><p>Different cryptocurrencies could have their own specialized tools, and your personal AI agent could find and use those tools when needed.</p><p>You wouldn’t have to learn how every blockchain works. Your agent would do much of that work for you.</p><p>Imagine this.</p><p>You’re walking down the street, wearing a smartwatch. You tell your agent, “I want to exchange some Bitcoin for USDC. Find me a good price.”</p><p>Your agent searches the available markets and finds a way to trade. Maybe it uses a DEX, maybe it finds another person willing to exchange assets, or maybe it uses a blockchain protocol that allows two people to trade without trusting each other.</p><p>It could even use a trading method that hasn’t been invented yet.</p><p>Then your agent prepares the transaction. You check the details, confirm with your fingerprint or another secure method, and the trade is completed.</p><p>You didn’t need to remember a wallet address, open several websites, or learn how different blockchain protocols work. You didn’t even need to use a centralized exchange.</p><p>You just told your own agent what you wanted to do.</p><p>And because you control your own wallet and software, you could also have more control over your financial privacy and identity. Your agent could help you use different tools without giving every service unnecessary access to your personal information.</p><p>This is why I believe AI could bring us into a new stage of Web3. More people will use cryptocurrency, and AI agents themselves may become important users of cryptocurrency too.</p><p>But I don’t believe all that growth will go to centralized exchanges. I expect decentralized exchanges and direct blockchain trading to grow faster.</p><p>**AI could make the original ideals of the cypherpunks much easier for ordinary people to enjoy.**</p><p>Think about it. The cypherpunks wanted people to have more freedom and privacy in the digital world. They wanted individuals to communicate, exchange information, and use digital money without depending entirely on governments or large companies.</p><p>The goal is freedom, privacy, and giving individuals more control over their own digital lives.</p><p>Every generation builds on the work of the people who came before it. The cypherpunks developed ideas and technologies that helped make Bitcoin possible. Later generations built cryptocurrencies, wallets, exchanges, and other applications.</p><p>And now AI might help bring those ideas to many more people.</p><p>I think that’s a very interesting future.</p><p>## Hackers Might Return to the Old PC Era</p><p>There’s another part of this future that I find interesting: security.</p><p>Today, large centralized exchanges are attractive targets because they control enormous amounts of cryptocurrency. Hackers may spend months planning an attack because a successful operation could bring them billions of dollars.</p><p>But if more people start managing their own cryptocurrency through personal AI agents and local wallets, the money will become more widely distributed.</p><p>A single personal wallet might contain much less money than a large exchange, so the incentive for expensive, highly targeted attacks could decrease.</p><p>Hackers won’t simply disappear. They’ll probably change their methods.</p><p>Instead of attacking one company that controls billions of dollars, hackers might try to compromise software used by millions of people. They could attack popular wallet applications, distribute malicious Agent Skills, or find vulnerabilities in software that many personal AI agents depend on.</p><p>One successful attack could affect a huge number of users.</p><p>Doesn’t that sound familiar?</p><p>It reminds me of the old PC era, when computer viruses spread through personal computers, malicious downloads, and software vulnerabilities.</p><p>In the future, hackers might once again focus on attacking large numbers of individual computers and devices, except this time those devices could contain AI agents that manage money.</p><p>Of course, there would also be some important differences. If people use different wallets, different software, and separate private keys, a hacker who compromises one system might not be able to steal everything.</p><p>That diversity could make certain attacks much harder. But widely used software could also become a new target.</p><p>So the battle between hackers and security engineers would continue, just in a different form.</p><p>And that’s what I find so interesting about the whole story.</p><p>The cypherpunks wanted to give individuals more control over their digital lives. Then centralized cryptocurrency exchanges became huge because they made cryptocurrency easier to use. Now AI agents might make it easier for individuals to manage their own cryptocurrency again.</p><p>At the same time, hackers might move from attacking huge centralized financial systems to attacking software running on personal computers, just as they did decades ago.</p><p>In one sense, we’re moving into a completely new future. In another, we’re returning to problems we’ve already faced before.</p><p>**That’s how human history often feels to me. In one sense, we’re moving forward into the future; in another, we’re moving backward.**</p><p>## P.S. The New Cold War</p><p>There’s also something from my university days that came back to me while reading about cryptography and Cold War.</p><p>I once developed a game that needed random numbers, But I was never satisfied with the random number generators I was using. I asked some people for advice and learned about techniques for generating randomness that had been studied in cryptography, including during the Cold War.</p><p>Eventually, I tried something unusual: dropping a needle onto different surfaces, recording the sounds it made, and using those recordings as a source of randomness.</p><p>I didn’t know much about cypherpunks at the time. Looking back, I find it interesting that I had already encountered one of the technical problems cryptographers had been thinking about for decades.</p><p>And Now, We are in a New Cold war</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=886f2e6c9cae" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Architecture Grows with Your Business]]></title>
            <link>https://anglefeint.medium.com/architecture-grows-with-your-business-50e9ee80eeeb?source=rss-ceb284c16b37------2</link>
            <guid isPermaLink="false">https://medium.com/p/50e9ee80eeeb</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[technical-debt]]></category>
            <category><![CDATA[overengineering]]></category>
            <category><![CDATA[engineering-mangement]]></category>
            <category><![CDATA[software-architecture]]></category>
            <dc:creator><![CDATA[AngleFeint]]></dc:creator>
            <pubDate>Sun, 01 Nov 2026 13:41:09 GMT</pubDate>
            <atom:updated>2026-11-01T13:41:09.347Z</atom:updated>
            <content:encoded><![CDATA[<p>Build what your business needs today, and leave room for it to change.</p><p><strong><em>Originally published on</em></strong><em> </em><a href="https://anglefeint.com/en/blog/architecture-grows-with-your-business/">https://anglefeint.com/en/blog/architecture-grows-with-your-business/</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*61yv5DbXjhij9z37UYcKeQ.png" /></figure><p>When I worked at some of China’s largest internet companies, I often heard big phrases in meetings: “best practices for this technology,” “the optimal architecture,” and “the next-generation infrastructure platform.”</p><p>At the time, cloud services were gaining popularity, but the companies I worked for still built much of their infrastructure themselves. They had dedicated infrastructure departments, with large teams developing and maintaining internal systems.</p><p>Imagine a small AWS inside every large internet company.</p><p>Some of those systems solved real problems. Others created a burden that lasted much longer than the people who built them.</p><p>Employees joined and left. Internal politics, performance reviews, and promotion requirements influenced technical decisions. Building a framework could bring recognition and influence, while maintaining it years later brought much less attention.</p><p>At one company, a leader in my department once took me aside and told me that my architecture focused too much on serving the business. He explained that, in his own technical platform, he deliberately coupled business functionality from different teams to the underlying infrastructure. This made those parts harder to separate and made the teams more dependent on him.</p><p>He was describing an intentional design choice. The dependencies strengthened his position, while leaving the company responsible for the complexity they created.</p><p>That conversation stayed with me. I doubt it was an isolated case, even within that department, let alone across the industry.</p><p>An architecture built to preserve someone’s position can make it harder for the business to change direction. When the company needs to separate a service, replace a component, or stop using a platform, those deliberately constructed dependencies become obstacles.</p><p>This is one reason I question the enthusiasm for building internal frameworks.</p><p>A department might develop its own framework even though a mature open-source alternative already exists and is widely used. Why should we expect our version to be better? More importantly, why should we expect it to remain better over time?</p><p>Even if the internal version has an advantage at launch, that advantage requires continuing investment. A widely adopted open-source project gets tested across many companies and real-world situations. An active community can discover bugs, contribute fixes, improve documentation, and adapt the software to new requirements.</p><p>Meanwhile, the original internal developers might leave. Their replacements need to learn the system and understand decisions that were never properly documented. Other teams build dependencies on it. Eventually, upgrading or replacing it can involve substantial work and risk.</p><p>What began as a technical achievement can become technical debt.</p><p>I have seen too many unnecessary frameworks built, adopted, and then inadequately maintained. “The open-source solution doesn’t fit our needs” was a familiar explanation. Sometimes that was true. Sometimes configuration, a small extension, or an upstream contribution could have solved the problem.</p><p>Before committing to an internal alternative, we should be able to answer three questions: Which limitation is blocking the business? What meaningful advantage will our own solution provide? Can the company sustain the responsibility after the original team moves on?</p><p>There are also times when building our own technology is necessary.</p><p>A new business model, a genuinely new technical product, or a unique operating environment may require capabilities that existing tools cannot provide. In those situations, forcing the business to fit the available tools can hold the company back. Developing the missing capabilities may be essential to its growth.</p><p>Refusing that work can be as damaging as building something unnecessary.</p><p>Recognizing the difference takes judgment — and honesty with ourselves. We need to distinguish a capability the business genuinely needs from a platform we would like to own. And we need to reconsider that judgment as the business and the available tools change.</p><p>This brings me to a broader question: what makes a good architecture?</p><p>I don’t think there is one standard architecture that every company should follow. The architectures we study today emerged as people responded to particular businesses, constraints, and tools. We should understand those conditions before adopting their answers.</p><p>Early YouTube is an example I keep coming back to. Its infrastructure had rough edges, and a small engineering team kept fixing bottlenecks as the service grew. The business became valuable while its architecture was still evolving. It did not need a finished answer to every future problem before it could succeed.</p><p>Through my own career, I have arrived at a simple belief: a good architecture can evolve with the business.</p><p>This does not mean we should stop reading papers or learning from other engineers. Their experience helps us make better decisions. But studying someone else’s architecture does not relieve us of the responsibility to understand our own business.</p><p>I am happy with a sixty-percent solution if it supports the business today and leaves us room to improve tomorrow. I would choose it over a supposedly perfect architecture that takes too long to build or becomes too expensive to change.</p><p>The remaining forty percent may become necessary later. Some of it may never matter. As the business develops, we may discover that we were measuring perfection against the wrong requirements in the first place.</p><p>A working system gives us something real to learn from. We can see where users struggle, where costs grow, and which limitations actually hold the business back. That experience helps us decide what deserves the next investment.</p><p>The ability to evolve includes more than adding features. It also includes upgrading dependencies, removing unnecessary components, replacing an internal system, or handing a responsibility over to an external service.</p><p>Sometimes, a temporary system is enough. If it supports the business today and replacing it later is affordable, it may already be doing its job. Every system does not need to last forever.</p><p>Today, cloud platforms and mature open-source projects provide many capabilities that companies once had to build themselves. That changes the economics of architecture.</p><p>For a company selling infrastructure, developing better infrastructure may be the business itself. For another company, the same investment needs to justify its place through the business results it enables.</p><p>That calculation starts before development begins and continues throughout the system’s life. It includes building and testing, training the teams that use it, integration, maintenance, debugging, upgrades, and eventually migration or retirement. Savings on servers are only one part of the account. We also need to consider what those people could have accomplished elsewhere.</p><p>Before optimizing a process, we should ask whether the business still needs that process at all.</p><p>Does every large company still need a large infrastructure department? I don’t think so. Its size should follow the needs of the business, the capabilities available elsewhere, and the responsibilities the company can sustain.</p><p>Architecture evolves; it cannot be fully mapped out in advance. We can make plans for the next three, five, or even ten years, but our architecture should leave room to adapt as the business changes.</p><p>When someone claims their architecture will meet every need for the next decade, they are essentially claiming to predict the future. And if you can really predict the next ten years, surely you have better things to do with that gift than design software architecture.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=50e9ee80eeeb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why I Write]]></title>
            <link>https://anglefeint.medium.com/why-i-write-3b771f2b7a76?source=rss-ceb284c16b37------2</link>
            <guid isPermaLink="false">https://medium.com/p/3b771f2b7a76</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[censorship]]></category>
            <category><![CDATA[totalitarianism]]></category>
            <category><![CDATA[1984]]></category>
            <dc:creator><![CDATA[AngleFeint]]></dc:creator>
            <pubDate>Wed, 21 Oct 2026 05:31:55 GMT</pubDate>
            <atom:updated>2026-10-21T05:31:55.767Z</atom:updated>
            <content:encoded><![CDATA[<p>Simplified Chinese Is Dying</p><p><strong><em>Originally published on</em></strong><em> </em><a href="https://anglefeint.com/en/blog/why-i-write/">https://anglefeint.com/en/blog/why-i-write/</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wgpmJQSd2pYpr-YG_2gWzA.png" /></figure><p>Throughout human history, many languages have been created, used by people, and eventually disappeared.</p><p>And now, I think — or perhaps I am sure — that Simplified Chinese is dying.</p><p>I do not mean that people will stop speaking Chinese, or that Simplified Chinese will disappear.</p><p>I mean another kind of death.</p><p>When I was a student, I loved surfing the Internet. Compared with today, those years were wonderful. In the Simplified Chinese world, there were many websites, forums, and all kinds of online communities. The Internet was much freer than it is today. You could talk about anti-CCP ideas online, and you could search for this kind of information using Chinese search engines.</p><p>In the early days of the Chinese Internet, we could access Google, Wikipedia, and YouTube. We also had PPLive — I remember watching some Korean pop shows there — Tianya, Tieba, and countless forums across the Chinese Internet. We could freely discuss many topics that are no longer allowed in today’s Chinese Internet.</p><p>Even around 2013 and 2014, we could still download and use many international apps such as Instagram, WhatsApp, LINE, Telegram, and others.</p><p>Was this a gift that the CCP gave to the Simplified Chinese world?</p><p>No.</p><p>The CCP simply did not have enough power or capability to fully control and suppress freedom in mainland China. It was not a gift. And that is just a fact.</p><p>In the late 1980s and early 1990s, Eastern Europe changed dramatically, and the Soviet Union collapsed. China at that time was still poor, and the CCP was far less powerful than it is today. It had far fewer technological and financial resources. There were many things it wanted to control but simply could not.</p><p>During this period and the years that followed, many Internet products were created.</p><p>Decades later, what has the Simplified Chinese world become?</p><p>When Chinese people want to talk about politics or other serious subjects, sometimes they cannot even use Simplified Chinese directly. They have to use pinyin instead. Sometimes even that does not work. An Internet service may prevent them from posting the information, delete what they have written, or even delete their account entirely.</p><p>Over the years, a huge amount of information in Simplified Chinese has been created and then disappeared because of totalitarianism.</p><p>I have seen this happen many times.</p><p>In today’s Simplified Chinese world, every website and app allowed to operate in China has effectively become part of the Chinese government’s system of control. At the same time, much of the information on these services is inaccessible to people with foreign phone numbers.</p><p>As a result, the Simplified Chinese Internet is becoming isolated from the rest of the world.</p><p>I think that in a communist country, memory itself can become illegal.</p><p>If people study this carefully, they may find fewer records of real everyday life in the Soviet era than in the era of Tsarist Russia. The same thing is happening in the Simplified Chinese world.</p><p>People do not remember what happened a hundred years ago. They do not know what society was really like before China’s Reform and Opening Up. They are even losing their memories of 2020–2022, as if the lockdowns had never happened.</p><p>As one of the Party’s slogans in 1984 puts it:</p><p>“Who controls the past controls the future: who controls the present controls the past.”</p><p>Not only can thought become illegal.</p><p>Memory can become illegal.</p><p>When the ecosystem of a language is dying, the language itself is dying. Simplified Chinese is now in this situation.</p><p>I think a language has depth.</p><p>What people write, discuss, create, and record becomes part of that depth. It is what people leave behind. It is proof that they were here.</p><p>When those things keep disappearing, the language becomes thinner and thinner.</p><p>There is another thing I have noticed as an engineer.</p><p>A free Internet is naturally decoupled. People can create their own websites, forums, communities, and products. They can grow, change, fail, or disappear independently.</p><p>Censorship works in the opposite direction.</p><p>Censorship is like a huge middleware layer inserted between what people freely create and the users who want to access it. It tries to decide what is allowed to reach users and what is not.</p><p>As more and more products, communities, forms of expression, and special cases appear, this middleware grows with them.</p><p>It becomes a huge and chaotic coupling layer.</p><p>Censorship becomes technical debt.</p><p>And the entire ecosystem has to keep paying interest on that debt.</p><p>When information, memories, arguments, records, and everything else within a language keep disappearing, what remains is only a thin surface.</p><p>That surface becomes thinner and thinner.</p><p>Let me tell you a story.</p><p>When I was a child, older people told me that Chinese society had been much more open and free in the 1980s and 1990s.</p><p>I listened to them, but I had never lived through that time myself.</p><p>Years later, I grew up and started working.</p><p>Then something strange happened.</p><p>Younger students started telling me about the time when I was young.</p><p>They told me how lucky my generation had been. They said that back then, we could use Google and access many parts of the global Internet directly. To them, the Internet of my youth seemed much freer than the one they had grown up with.</p><p>And suddenly, I realized that I had become the older generation in the same story.</p><p>I did not tell them that my time was freer.</p><p>They told me.</p><p>Isn’t that ironic?</p><p>The freer China of the 1980s and 1990s was something I knew through the memories of people who had lived through it.</p><p>Now, the freer Internet that I lived through is becoming a memory too.</p><p>When the records of a time disappear, a living past gradually becomes a story — something one generation tells the next.</p><p>This is what I mean when I say that a language becomes thinner.</p><p>What was once there becomes a memory.</p><p>Then even the memory begins to disappear.</p><p>So I want to write down some truths and preserve them for people in the future.</p><p>This is why I write.</p><p>This piece is a tribute to George Orwell’s Why I Write.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3b771f2b7a76" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Time Starts Again]]></title>
            <link>https://anglefeint.medium.com/time-starts-again-d4f1c7d679e7?source=rss-ceb284c16b37------2</link>
            <guid isPermaLink="false">https://medium.com/p/d4f1c7d679e7</guid>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[time-management]]></category>
            <category><![CDATA[internet]]></category>
            <dc:creator><![CDATA[AngleFeint]]></dc:creator>
            <pubDate>Wed, 21 Oct 2026 05:27:58 GMT</pubDate>
            <atom:updated>2026-10-21T05:27:58.415Z</atom:updated>
            <content:encoded><![CDATA[<p>For the first time in years, the world feels new again.</p><p><strong><em>Originally published on</em></strong><em> </em><a href="https://anglefeint.com/en/blog/time-starts-again/">https://anglefeint.com/en/blog/time-starts-again/</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fS2Dgt4Lh0m1RVXiWVLb7Q.png" /></figure><p>I still remember the first time I touched a computer. I wasn’t even old enough to be in primary school. It was in my dad’s office.</p><p>Back then, computers were still very new to a lot of people. One of my dad’s colleagues showed me a video game on that computer. Even after all these years, I can still remember that moment and the image on the screen: a man jumping across a stage.</p><p>When I was in primary school, I started playing StarCraft, Army Men, C&amp;C: Red Alert, Counter-Strike, Diablo II, Age of Empires, Virtual Life 2, and, you know, a lot of other video games. During those primary school years, I learned Flash. There was a famous website called 闪客帝国 (Shanke Diguo, literally “Flash Empire”) back then, and at school we used Logo to learn programming.</p><p>The Matrix — yes, you guessed it, the famous movie. I first watched it when I was still in primary school. I watched that movie again and again. How many times? Want to guess? You know what? I don’t know either.</p><p>When I was in junior high school, I joined the computer club at our school and continued using Flash to build little Flash projects, haha. We were still playing computer games like StarCraft, Diablo II, Counter-Strike, and Virtual Life 2, but the world was beginning to change. New kinds of games were being created, like The Legend of Mir 2, and platforms like HF Gaming Platform were appearing. We started using the Internet, instead of just LANs, to play video games with our friends.</p><p>In high school, I started using the Visual Basic programming language to write some software. And speaking of games — something huge appeared and suddenly became popular all over the world. That game was called World of Warcraft. You can’t imagine how popular WoW was. Almost every cybercafe I went to was full of WoW players. Then Dota came out, and it became popular around the world too.</p><p>At university, I studied C/C++, Java, PowerScript, JavaScript, C#, Brainfuck, and more. I used IRC for secure chats with friends from all over the world. We had a computer club there too, where we used Silverlight — a technology from the past, a little like Flash, using C# and XAML — to build a game. That is still the only game I have ever built. I joined ACM programming competitions and some programming competitions held by companies. I used LabWindows/LabVIEW and C/C++ to develop a missile guidance system. During university, we even developed an online storage system similar to Dropbox.</p><p>Looking back at those early years of my life, I still remember what the Internet felt like when I was a child, as it became more and more popular around the world. People speaking all kinds of languages were building their own websites. It was a freer age. Even under the Chinese Communist Party (CCP), you didn’t need government approval just to build a website. You could simply build one yourself. There were also many websites that told you things the CCP didn’t want to talk about, and you could find them through Baidu or other Chinese search engines. We could even discover good games that the CCP didn’t want us to play, like Command &amp; Conquer: Generals. A classmate could tell you about the game, and then you could find it on the Internet yourself.</p><p>Because China had such a large population when the Internet became popular around the world, there were a huge number of Simplified Chinese websites, and Simplified Chinese content made up a significant part of the Web.</p><p>It was a good time for everybody.</p><p>When I was a student, we could use Wikipedia, YouTube, Instagram, Telegram, Facebook, Google, LINE, WhatsApp, Gmail, Google Maps, and so many other websites and applications.</p><p>I still remember the first time I used Google Maps with my older siblings. We tried to find our school and our home on the map.</p><p>That was amazing.</p><p>But you know, there are always some bad guys who don’t want the good times to last.</p><p>In 2010, Google left China after a cyberattack originating from China and amid its conflict with the Chinese government’s censorship requirements. I saw the news while I was using iGoogle and its drag-and-drop gadgets, so I remember that day clearly.</p><p>Later, we couldn’t access Google anymore.</p><p>To me, the reason was clear: the remnants of the Soviet Union had refused to abandon their evil plans and had begun putting them into action…</p><p>They used so-called “Internet sovereignty” to dismantle the various organizations and communities people had formed through the Internet. Everything changed. Over the years, Simplified Chinese content became a smaller and smaller part of the Web. Even the Internet itself became boring. Information accumulated on the Simplified Chinese Internet over the past ten or twenty years has been wiped away — as if it had never existed at all, just like in 1984. A lot of Chinese websites disappeared, while others deleted huge numbers of old Chinese posts.</p><p>In China, even buying a SIM card requires real-name registration, and when people post something on major Internet platforms, their IP location is displayed. The Chinese Internet has become increasingly separated from the global Internet.</p><p>But the new Axis continued to move. We could no longer freely use many foreign websites and applications. Even foreign-made games became harder or impossible to access officially. If people wanted to create a website, they needed to register it with the government. The same eventually became true for mobile apps. And in the Simplified Chinese world, if you wanted to publish a game for Chinese players, you also needed government approval.</p><p>It felt as if the Internet had stopped moving and had become so boring for everyone.</p><p>Then, in 2022, Russia and its friends put their plans into action with the invasion of Ukraine. A new era seemed to begin.</p><p>A new Cold War was beginning.</p><p>But suddenly, AI began showing us the future.</p><p>The regimes of the new “Axis” this time are all very weak. AI will be the protagonist, just as “Star Wars” — the Strategic Defense Initiative (SDI) — became a symbol of technological competition during the last Cold War.</p><p>The AI war has already begun.</p><p>Everybody is building their own AI products or using AI to do something.</p><p>Everybody is talking about AI, just as people talked about computers and the Internet many years ago.</p><p>The future is rhyming with the past.</p><p>A different world is coming.</p><p>When the Internet began, I was a child. I enjoyed it, but I didn’t build anything of my own.</p><p>But now, I know AI will become boring one day too. Maybe in 30 years. But I don’t want to find myself writing another retrospective article like this one.</p><p>This time, I will be part of it, not just enjoy it. I will build something of my own.</p><p>It feels as if I am back in the past, once again watching that man jump across a stage on a computer screen.</p><p>Time starts again.</p><p>And the gears of fate begin to turn.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d4f1c7d679e7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The High Cost of Hiring Big Names]]></title>
            <link>https://anglefeint.medium.com/the-high-cost-of-hiring-big-names-eafe1bb0a915?source=rss-ceb284c16b37------2</link>
            <guid isPermaLink="false">https://medium.com/p/eafe1bb0a915</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[corporate-culture]]></category>
            <category><![CDATA[management]]></category>
            <dc:creator><![CDATA[AngleFeint]]></dc:creator>
            <pubDate>Thu, 15 Oct 2026 06:59:38 GMT</pubDate>
            <atom:updated>2026-10-15T06:59:38.902Z</atom:updated>
            <content:encoded><![CDATA[<p>Favoritism, Factionalism, Misallocation of Resources, and the Erosion of Fairness</p><p><strong><em>Originally published on</em></strong><em> </em><a href="https://anglefeint.com/en/blog/the-high-cost-of-hiring-big-names/">https://anglefeint.com/en/blog/the-high-cost-of-hiring-big-names/</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wgpmJQSd2pYpr-YG_2gWzA.png" /></figure><p>Throughout my career, I’ve witnessed many good decisions — and plenty of bad ones.</p><p>In my opinion, one of the worst mistakes a company can make is hiring people simply because they carry prestigious titles.</p><p>I’ve seen this happen in both large corporations and startups.</p><p>For example, when I joined a startup, its core business was already growing rapidly. The company then decided to hire several university professors and senior executives from well-known companies.</p><p>After one of these executives joined, the company started building a cloud computing business.</p><p>But cloud computing had little to do with the company’s existing operations. There wasn’t even a closely related business that could justify such an expansion.</p><p>So why did the company need him in the first place?</p><p>Hiring university professors was even more ridiculous.</p><p>Just because someone has taught students at a university doesn’t mean they know how to build reliable infrastructure, develop successful applications, or run a profitable business.</p><p>And when these people joined the company, they often brought their former colleagues and students along with them.</p><p>Some of these students had just graduated from university, yet they were given surprisingly high positions and salaries.</p><p>Naturally, many existing employees felt this was unfair.</p><p>After all, the company’s success hadn’t been built by these newly hired executives, their students, or their former colleagues. It had been built by the employees who had been working hard long before they arrived.</p><p>I once asked the company why it made these decisions.</p><p>I was told that bringing in prestigious people would make the capital markets value our company more highly, among other things.</p><p>But guess what?</p><p>Looking back years later, I think those arguments were just bullshit.</p><p>The company’s business wasn’t built by those famous executives or professors.</p><p>And this shit doesn’t just happen at startups. Throughout my career, I’ve seen large tech companies make the same mistakes.</p><p>Just look at China’s major tech companies. Over the years, many of them established departments with impressive names like “Research Labs” or “AI Institutes.” They hired university professors and other big names to lead these organizations.</p><p>In recent years, some of these initiatives have been acknowledged by management as failures.</p><p>But by then, substantial company resources had already been wasted on decisions that, in hindsight, should never have been made.</p><p>Meanwhile, the employees who never worked in those prestigious labs, but instead spent their days building and maintaining the company’s actual business, were often treated as ordinary, lower-level employees.</p><p>Is that fair?</p><p>Looking back at how far the market valuations of many Chinese tech companies have fallen from their peaks, I can’t help but wonder how much these kinds of management mistakes contributed to their decline.</p><p>The success of these companies wasn’t built by prestigious titles.</p><p>It was built by thousands of ordinary employees who did the actual work.</p><p>And when I left China to begin a new chapter of my life, I discovered that the same shit was happening in American companies, in much the same way.</p><p>Yet American companies seem largely unaware of how damaging these practices can be.</p><p>I think the difference comes down to where the two countries stand in their economic cycles.</p><p>China’s era of rapid growth, driven by decades of reform and opening up, has largely come to an end. After decades of expansion, the slowdown has made organizational mistakes much harder to ignore.</p><p>America, on the other hand, is the new Rome of our civilization.</p><p>And this new Rome is still far from reaching its peak.</p><p>Its cycle of prosperity has lasted far longer, and continued economic growth allows companies to overlook inefficiencies and mistakes that might be impossible to ignore elsewhere.</p><p>Growth can hide a lot of bad decisions.</p><p>And that leaves me with a question:</p><p>If we can avoid these mistakes and build companies the right way in this new Rome, how far could we go?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=eafe1bb0a915" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Awkward Future of iOS/Android Cross-Platform Frameworks in the Age of AI]]></title>
            <link>https://anglefeint.medium.com/the-awkward-future-of-ios-android-cross-platform-frameworks-in-the-age-of-ai-9086adfd149f?source=rss-ceb284c16b37------2</link>
            <guid isPermaLink="false">https://medium.com/p/9086adfd149f</guid>
            <category><![CDATA[cross-platformdevelopment]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[mobile-development]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[AngleFeint]]></dc:creator>
            <pubDate>Fri, 09 Oct 2026 15:29:37 GMT</pubDate>
            <atom:updated>2026-10-09T15:39:43.373Z</atom:updated>
            <content:encoded><![CDATA[<p><em>The Cross-Platform Illusion: Engineering, Bureaucracy, and the Age of AI</em></p><p><strong><em>Originally published on</em></strong><em> </em><a href="https://anglefeint.com/en/blog/cross-platform-frameworks-in-the-age-of-ai/"><strong>https://anglefeint.com/en/blog/cross-platform-frameworks-in-the-age-of-ai/</strong></a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fS2Dgt4Lh0m1RVXiWVLb7Q.png" /></figure><p>I spent more than a decade working at some of the largest internet companies in mainland China. Later, I moved overseas and joined a global technology company operating worldwide, but not in mainland China. After that, I started my own company overseas.</p><p>These experiences gave me a different perspective on the engineering culture and technical decisions I had witnessed in China.</p><p>My doubts about cross-platform frameworks began in 2024.</p><p>At the time, I was building an ASR/TTS mobile app in my spare time. Instead of relying on cloud computing, I wanted to use on-device computing power, as both iOS and Android were already supporting increasingly capable machine learning APIs.</p><p>I had also started using Cursor for AI-assisted development.</p><p>While researching cross-platform solutions for my app, I came to a conclusion: iOS/Android cross-platform frameworks would become increasingly irrelevant in the AI era.</p><p>And perhaps their value had been overstated even before AI.</p><p>Recently, I came across Shopify’s engineering article, [Native Is Now the Future of Mobile at Shopify](<a href="https://shopify.engineering/back-to-native">https://shopify.engineering/back-to-native</a>), describing its decision to move from React Native back to native development.</p><p>It reminded me of the doubts I had about cross-platform frameworks back in 2024 and inspired me to finally write down my own thoughts.</p><p>Shopify claims React Native was the right choice in 2020. I am not so convinced.</p><p>Shopify itself acknowledges spending significant time and resources on performance optimization, framework improvements, and dependency maintenance. But these are only some of the costs of cross-platform development. What about the hidden costs of maintaining a developer ecosystem, organizational overhead, and lost business opportunities?</p><p>Did Shopify ever fully calculate the true economic cost of that decision? I seriously doubt it.</p><p>There are several reasons behind my conclusion:</p><p>* <strong>AI coding agents</strong> are dramatically reducing the cost of developing and maintaining native applications.<br>* <strong>On-device AI</strong> makes native development more attractive, while cross-platform frameworks often lag behind new platform capabilities.<br>* <strong>Hidden costs</strong>, including framework maintenance, internal developer ecosystems, cross-team coordination, and organizational politics, are frequently underestimated.<br>* <strong>The New Cold War and the AI Arms Race</strong>: We are now in a new Cold War, and AI has become one of its most important strategic battlegrounds, much like the Strategic Defense Initiative (Star Wars) during the previous Cold War. As competition intensifies, time-to-market and user experience become increasingly important relative to development costs.</p><p>But before discussing AI, I want to share something I observed in China’s internet industry.</p><h3>Write Once, Run Anywhere?</h3><p>At several large Chinese internet companies where I worked, cross-platform frameworks were often promoted as a way to reduce development costs.</p><p>But their ambitions went far beyond sharing code between iOS and Android.</p><p>Companies wanted to write code once and run it everywhere: iOS, Android, H5, WeChat Mini Programs, Alipay Mini Programs, and sometimes even more platforms.</p><p>The argument sounded simple: write once, run anywhere.</p><p>But reality was much more complicated.</p><p>* <strong>Write once? Not necessarily.</strong> Sometimes developers ended up writing code three times: once for iOS, once for Android, and once for the cross-platform framework. Supporting Web or mini-program platforms could introduce even more work.<br>* <strong>Performance and user experience.</strong> Significant engineering effort was often required to approach native performance, and some limitations could not be fully eliminated.<br>* <strong>Platform API delays.</strong> New iOS and Android capabilities were not always immediately available through cross-platform frameworks.<br>* <strong>Coordination overhead.</strong> Cross-platform development introduced additional dependencies between business teams, framework teams, and native platform teams.<br>* <strong>Organizational politics.</strong> Some engineering projects seemed more focused on technical achievements, team expansion, and promotions than on actual business value.<br>* <strong>Unnecessary iterations.</strong> Did the business really need to change the app so frequently? Or were some projects simply work created for the sake of work?</p><p>And every additional layer of complexity comes with a cost.</p><p>Even if a framework reduces development time, how do we measure the performance, user experience, and business opportunities sacrificed along the way?</p><h3>The Hidden Cost: An Internal Developer Ecosystem</h3><p>One of the most underestimated costs of building a cross-platform framework is maintaining its developer ecosystem.</p><p>Apple and Google invest heavily in their developer ecosystems because these ecosystems directly strengthen their platforms and businesses.</p><p>But when an internet company builds its own cross-platform framework, it effectively creates another developer platform inside the company.</p><p>It must maintain native API bridges, UI components, build tools, debugging tools, testing infrastructure, documentation, SDK integrations, and backward compatibility.</p><p>And this internal ecosystem must constantly follow the evolution of iOS, Android, and sometimes Web, mini-programs, and HarmonyOS.</p><p><strong>The company is no longer just building an application. It is maintaining a platform within a platform.</strong></p><p>Using an established framework like React Native can reduce this burden because Meta and the open-source community maintain much of its ecosystem. But large companies often end up building extensive internal infrastructure around it anyway.</p><p>So how much money does this ecosystem actually save after accounting for its own development and maintenance costs?</p><p>And is it being built because the business needs it, or because an engineering organization needs something to build?</p><h3>Hot Updates and Manufactured Demand</h3><p>There is another reason why cross-platform and dynamic frameworks became so popular in China’s internet industry: **hot updates.**</p><p>Many companies wanted to update their apps without going through the traditional app store review and release process every time.</p><p>This was especially attractive to companies running endless marketing campaigns, promotional events, and frequent product experiments.</p><p>But looking back, I have a question.</p><p>Did users really need so many changes?</p><p>Did those campaigns and constant iterations actually create enough business value to justify their costs?</p><p>Or were some of them simply manufactured by large organizations that needed more projects, more activities, and more achievements?</p><p>A framework that makes unnecessary development faster does not necessarily make a company more efficient.</p><p><strong>It may simply make the company better at producing unnecessary work.</strong></p><p>And no matter how sophisticated its engineering infrastructure becomes, a company cannot solve fundamental business problems by shipping more features.</p><p><strong>Engineering efficiency does not necessarily translate into business efficiency</strong>.</p><h3>Exam-Driven Engineering and Manufactured Complexity</h3><p>There is another problem I observed in China’s internet industry: an exam-driven engineering culture.</p><p>China’s exam-oriented education system trains people to solve difficult problems, but rarely encourages them to question whether those problems should exist in the first place.</p><p>This mindset can carry over into engineering organizations.</p><p>Some engineers are excellent at solving complicated technical problems, but rarely question whether those problems need to be solved.</p><p>Sometimes, they even create problems where none exist.</p><p>Software engineering has no perfect architecture, no perfect framework, and no universally correct solution. Almost any technical decision can be justified with enough arguments, benchmarks, and carefully selected trade-offs.</p><p>This creates plenty of room for organizational politics.</p><p>As long as a proposal sounds technically reasonable and satisfies the right people, an engineering team can justify almost anything.</p><p>Build a new framework. Rewrite an existing system. Introduce another abstraction layer. Create a new technical standard.</p><p>Then use the complexity of the project, the number of engineers involved, and its supposed technical achievements to secure promotions.</p><p><strong>The goal is no longer to solve problems. It is to manufacture problems worth getting promoted for.</strong></p><p>And because there is no perfect architecture, there is always another problem to solve, another framework to build, and another reason to keep the organization busy.</p><p>The real question — whether any of this creates enough value to justify its cost — is often conveniently ignored.</p><p><strong>When engineering complexity becomes a path to promotion, simplicity becomes a threat.</strong></p><h3>When Abstraction Creates More Abstraction</h3><p>Here is something almost absurd.</p><p>At one Chinese internet company, multiple cross-platform frameworks existed simultaneously.</p><p>Eventually, another framework was developed to make those existing frameworks compatible with one another.</p><p>Think about that.</p><p><strong>We built abstractions to reduce complexity, then built another abstraction to manage the complexity created by those abstractions.</strong></p><p>Ridiculous.</p><p>And this happened before AI coding agents began changing the economics of software development.</p><h3>The Economics of Cross-Platform Development</h3><p>All of this can be examined with a simple formula.</p><p>Let’s define the variables:</p><p>V_cross — Net economic value of adopting a cross-platform framework.<br>C_saved — Actual development and maintenance costs saved through code reuse across platforms.<br>C_complexity — Additional costs from abstractions, compatibility issues, debugging, and performance optimization.<br>C_ecosystem — Costs of maintaining developer tools, SDKs, testing infrastructure, and platform API compatibility.<br>C_organization — Additional costs from coordination, bureaucracy, organizational politics, and promotion-driven projects.<br>C_opportunity — Opportunity costs from delayed time-to-market, platform API lag, inferior user experience, and lost platform differentiation.</p><p>The formula is:</p><p>V_cross = C_saved — C_complexity — C_ecosystem — C_organization — C_opportunity</p><p>A cross-platform framework only creates positive economic value when the costs it actually saves exceed the additional costs it introduces.</p><p>And this was already a question worth asking before AI.</p><h3>The Economics of Cross-Platform Development in the AI Era</h3><p>Now, let’s see how AI changes this equation.</p><p><strong>First, AI coding agents are dramatically reducing the cost of native development.</strong></p><p>As a result, the development costs that cross-platform frameworks were originally designed to save are shrinking.</p><p>C_saved \ going down</p><p><strong>Second, the New Cold War and the AI arms race are making native platform capabilities increasingly valuable.</strong></p><p>On-device AI, or edge AI, will inevitably become an essential part of iOS, Android, and other computing platforms.</p><p>Native development offers faster access to new APIs, better performance, easier integration, and more direct access to platform-specific hardware and software.</p><p>But there is another question worth asking.</p><p><strong>Why should iOS and Android applications be identical in the first place?</strong></p><p>Windows and Linux applications do not necessarily look or behave the same. Why should mobile applications?</p><p>Different platforms have different hardware, operating systems, AI capabilities, and user expectations.</p><p>In the coming years, Apple and Google will continue developing their own on-device AI capabilities, models, hardware acceleration, privacy mechanisms, and system-level integrations.</p><p>Why should a product limit itself to the common subset of these capabilities?</p><p>In an era of intense technological competition, platform-specific differences can become competitive advantages rather than problems to be eliminated.</p><p>And when companies compete on performance, user experience, and speed of innovation, development costs carry relatively less weight.</p><p>This increases the opportunity cost of relying on cross-platform frameworks.</p><p>C_opportunity \ going up</p><p><strong>Third, AI does not eliminate the complexity introduced by cross-platform frameworks.</strong></p><p>AI coding agents can help developers build native modules, adapt new APIs, and maintain cross-platform frameworks faster.</p><p>But writing code is only part of the problem.</p><p>A cross-platform framework still introduces additional runtime layers, integration boundaries, compatibility requirements, and debugging complexity.</p><p>When something goes wrong, developers may need to determine whether the problem comes from the operating system, the native implementation, the framework, or the integration between them.</p><p>AI can reduce the cost of dealing with these problems, but it cannot make the architectural complexity disappear.</p><p>Meanwhile, AI is making native development cheaper without requiring those additional abstraction layers.</p><p><strong>AI reduces the cost of implementing abstractions, but it does not eliminate the complexity introduced by those abstractions.</strong></p><p>Therefore:</p><p>V_cross \ going down</p><p><strong>The economic value of cross-platform frameworks is being squeezed from both sides.</strong></p><p>Native implementation is becoming cheaper, while platform-specific capabilities are becoming more valuable.</p><h3>Reuse the Core, Not Necessarily the Entire App</h3><p>When I was building my ASR/TTS app in 2024, I eventually chose not to use a cross-platform application framework.</p><p>I wanted to work directly with native capabilities, including Core ML and the C/C++ inference ecosystem around Whisper.</p><p>Interestingly, [whisper.cpp](<a href="https://github.com/ggml-org/whisper.cpp">https://github.com/ggml-org/whisper.cpp</a>) is itself a cross-platform project.</p><p>But this is a fundamentally different kind of code reuse.</p><p>Sharing a C/C++ inference engine across operating systems is not the same as forcing an entire application through a shared UI framework, runtime, and developer ecosystem.</p><p>A native iOS app can use SwiftUI and Core ML. A native Android app can use Kotlin and Android-specific acceleration. Both can still reuse appropriate parts of the same inference core.</p><p><strong>Cross-platform code reuse is not the same as a cross-platform application framework.</strong></p><p>My argument is not against sharing code. It is against the assumption that sharing an entire application implementation is always economically beneficial.</p><p>And AI makes this distinction increasingly important.</p><h3>The Awkward Middle</h3><p>For simple applications, content-driven products, and frequently updated pages, why not just use Web technologies?</p><p>The Web already offers a mature, widely supported platform with an enormous developer ecosystem.</p><p>For rich applications, especially those deeply integrated with on-device AI, hardware acceleration, and operating system capabilities, why not use Native?</p><p>Cross-platform application frameworks are increasingly caught in the middle.</p><p>Of course, they still have value in certain situations. A small team building a moderately complex application with limited platform-specific requirements may benefit greatly from React Native or Flutter.</p><p>But that value should be demonstrated, not assumed.</p><p>And we should stop treating code reuse as an unquestionable engineering virtue.</p><p>Decades ago, Java popularized the idea of Write Once, Run Anywhere. Much of modern software engineering has followed the same direction: introducing abstractions to reduce the cost of supporting different platforms.</p><p>AI is changing the economics behind that decision.</p><p>We may once again build separate implementations for different platforms — not because we have forgotten the lessons of the past, but because the cost of implementation itself has changed.</p><p>The future may not be about eliminating platform differences.</p><p>It may be about embracing those differences while making their implementation dramatically cheaper.</p><p><strong>Reuse the core. Automate the implementation. Preserve the platform.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9086adfd149f" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>