2026 05 26 HackerNews

2026-05-26 Hacker News Top Stories [#](https://supertechfans.com/cn/post/2026-05-26-HackerNews/#2026-05-26-hacker-news-top-stories)

1. The Magnifica Humanitas advocates that technology and AI must be constrained by morality and the rule of law, rejects superhumanism, and builds a peaceful civilization based on love and justice through dialogue, education, and multilateral cooperation.

2. Google's push for conversational AI search has sparked resistance, with the article recommending alternatives like Kagi, DuckDuckGo, Startpage, &udm=14, Brave, and Ecosia to meet personalized needs for privacy, ad-free browsing, and controllable AI.

3. A comparison of Go and Rust for backend teams highlights that migration is primarily driven by maintainability and security rather than performance. While Rust has a steep learning curve, its long-term benefits are significant, suggesting a gradual introduction.

4. Geohot argues that treating AI agents as programmers is an expensive mistake, as LLMs only distribute fitting without addressing underlying issues, and should be used as assistants rather than replacements. True programming agents require a world model.

5. Pope Leo XIV emphasizes that AI must serve all humanity rather than a privileged few, calls for disarmament and strong regulation, opposes transhumanism, and promotes labor-based policies and AI education for youth.

6. The author reflects on their self-taught journey in Android development, warning that while LLMs can aid debugging and review, they may weaken thinking and learning. Continuous improvement through trial and error, collaboration, and sharing is encouraged.

7. By mapping Minsky register machines to Jira automation, the article demonstrates that Jira is Turing complete, though limited by cloud resources and execution chains, theoretically possessing unbounded computational power.

8. AI is essentially pattern matching and lacks contextual and constraint understanding, making it unsuitable for architecture design. Engineers should lead design, with AI serving as an implementation assistant and ensuring clear accountability.

9. A trial of a 100:80:100 four-day workweek by 15 Australian companies showed no decline in productivity and significant stress reduction, achieved through streamlined meetings and automation, requiring tailored processes.

10. Dutch authorities seized 800 servers and arrested two individuals suspected of providing infrastructure for Russian intelligence cyberattacks and disinformation campaigns in the EU, violating sanctions laws and denying knowledge.

1. Magnifica Humanitas [#](https://supertechfans.com/cn/post/2026-05-26-HackerNews/#1-%e5%b0%8a%e8%b4%b5%e7%9a%84%e4%ba%ba%e6%80%a7-magnifica-humanitas)

https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html

This article is Pope Leo XIV's encyclical "Magnifica Humanitas" on protecting human dignity in the age of artificial intelligence. It explores the application and guiding significance of Catholic social doctrine in the new era, focusing on the profound impact of contemporary technological developments, particularly AI, on human society.

The encyclical begins by posing a choice: to build a "Tower of Babel"-like division or to construct a city where God and humans coexist. Each generation bears the responsibility to shape the era, safeguard human dignity, promote justice, and foster brotherhood. Christ's Incarnation reveals the true meaning of humanity, and the Church has historically guided people towards abundant life.

It reviews the development of Catholic social doctrine, starting with Leo XIII's "Rerum Novarum," emphasizing that the Church cares not only about eternal life but also about real-world social issues. Social doctrine, grounded in scripture and tradition, integrates science to provide principles of thought, standards of discernment, and guidelines for action, helping people address contemporary challenges.

Regarding technology and AI, the article asserts that technology itself is not humanity's enemy but an extension of human freedom and autonomy. Technological progress improves living conditions but also brings risks and challenges. AI, as a valuable tool, requires responsible governance, transparency, and vigilance against potential misuse. It critiques narratives of superhumanism and posthumanism, emphasizing human limits, inner life, and dignity, advocating for the true meaning of Christian humanism's "transhumanism."

During periods of social transformation, the article stresses the importance of truth, work, and freedom. Truth is a public good, crucial for democracy and communication, necessitating the establishment of an educational alliance in the digital age, with schools playing a central role. Work dignity must be recognized, and society should create hope for families and youth in the face of unemployment and economic challenges. Freedom must guard against dependency and commercialization, breaking new forms of servitude.

The article also discusses power culture and the civilization of love in the digital age. Power culture manifests as normalized warfare, unlimited force, and threats from AI weapons, with the crisis of multilateralism intensifying. Building a civilization of love requires collective effort, eliminating violent rhetoric, constructing peace through justice, considering the perspective of victims, restoring dialogue, diplomacy, and multilateral cooperation, and supporting this process with prayer and hope.

The conclusion emphasizes the mystery of the Incarnation, calling for joint construction of today's "building site," inspired by the hymn "Gloria." The entire encyclical reflects the Church's deep concern for human dignity in the technological age, urging global shared responsibility to promote social justice and human care.

• * *

https://news.ycombinator.com/item?id=48265206

• The development of industrial technology once promoted the rise of the middle class, partly because it required extensive manual labor, forming a large consumer base.
• The cost reduction of renewable energy and battery technologies is partly due to policy interventions and subsidies, such as Germany's Renewable Energy Act promoting technology adoption and price reductions.
• Policy plays a key role in the early stages of technological development but its influence may diminish once the technology is widely adopted and economic pathways are clear.
• After the widespread adoption of electricity, power companies did not monopolize societal power, partly due to strict government regulation and nationalization of the power industry.
• The history of the power industry shows how governments have limited the power of private power companies through antitrust laws and public utility regulations.
• Some technologies, such as cloning, are feasible but widely banned or restricted due to social ethical reasons.
• The use of antibiotics is regulated and restricted to prevent misuse, demonstrating constraints on technology for public interest.
• The control of harmful substances like chlorofluorocarbons (CFCs) is a typical case of technology being "tamed," with consumers hardly noticing sacrifices.
• Restrictions on weapons of mass destruction, including nuclear, biological, and chemical weapons, are more based on rational choices in game theory than pure humanitarianism.
• Regulation and control over harmful substances and behaviors (such as nicotine, trans fats, gambling, alcohol) reflect attempts by society to manage the negative impacts of technology and products.
• The regulation and restriction of technology often adjust repeatedly as society reconsiders historical experiences.
• Emerging technologies like AI may require stringent regulation akin to public utilities to prevent excessive concentration of power.

• * *

2. Search Engine Alternatives: Google Isn't Google Anymore [#](https://supertechfans.com/cn/post/2026-05-26-HackerNews/#2-%e6%90%9c%e7%b4%a2%e5%bc%95%e6%93%8e%e6%9b%bf%e4%bb%a3%e6%96%b9%e6%a1%88%e8%b0%b7%e6%ad%8c%e5%b7%b2%e4%b8%8d%e5%86%8d%e6%98%af%e8%b0%b7%e6%ad%8c-search-engines-alternatives-now-that-google-isnt-google-anymore)

https://techcrunch.com/2026/05/21/six-search-engines-worth-trying-now-that-google-isnt-really-google-anymore/

This article introduces Google's major upgrade to its search engine announced at the 2026 Google I/O conference, transitioning to a conversational AI-driven search experience. Users can opt for AI mode, receiving AI-generated search summaries and chat boxes to facilitate follow-up questions. This is the largest overhaul of Google search in 25 years, but many users have reacted coolly, viewing it as another example of tech companies forcing AI chatbots into all products, and noting controversy surrounding the previous launch of Google AI summaries. Additionally, Google was ruled illegal for monopoly practices by an American court in 2024, leaving users fatigued by its dominant position.

The article then recommends several alternative search engines:

1. Kagi: A paid, ad-free search engine that allows users to customize their search experience and filter websites, offering AI quick answers but allowing AI summaries to be turned off.
2. DuckDuckGo: A free search engine that generates revenue through keyword-related ads, does not collect user data, and supports disabling AI features.
3. Startpage: Acts as a proxy for Google, protecting user privacy by removing personal information before requesting search results from Google, and supports disabling AI features.
4. &udm=14: A simplified version of Startpage that automatically removes AI summaries from Google searches, with open-source code for self-hosting.
5. Brave: Offers both a browser and search engine, supports Chrome extensions, allows users to apply third-party search filters, and supports toggling AI features.
6. Ecosia: Also offers a browser and search engine, emphasizing environmental sustainability by using 80% of its revenue for tree planting, prioritizing transparency and actual impact.

Overall, the article aims to help users find alternative search tools that better meet privacy protection and personalized needs in the context of Google search's significant transformation and user dissatisfaction.

• * *

https://news.ycombinator.com/item?id=48266051

• Kagi's search results are precise, supporting AI-assisted and personalized features, with a good user experience.
• Kagi currently relies mainly on third-party index data and has not yet fully developed its own general search index.
• Kagi employees indicate they are actively developing their own index, including specialized indexes for specific fields such as programming.
• Kagi's existing Teclis index focuses on "niche networks" and has not covered large-scale general network indexing.
• Building a large-scale search index faces challenges, including dealing with massive SEO spam content, with Kagi developing related detection technologies.
• Whether search engines will conduct content reviews for political or commercial reasons is a focal point of user concern, with some users opposing any form of review.
• Content review by search engines and choosing partners are different issues, involving complex ethical and geopolitical considerations.
• Users have moral and political concerns about search engines purchasing data from specific countries, such as Russia.
• Different users have varying views on whether to avoid collaborating with companies from certain countries, involving complex discussions of geopolitics and business ethics.

• * *

3. Migrating from Go to Rust [#](https://supertechfans.com/cn/post/2026-05-26-HackerNews/#3-%e4%bb%8e-go-%e8%bf%81%e7%a7%bb%e5%88%b0-rust-migrating-from-go-to-rust)

https://corrode.dev/learn/migration-guides/go-to-rust/

This article is a guide on migrating from the Go language to the Rust language, primarily aimed at backend service developers. The author notes that Go and Rust overlap in performance and type systems but that the core of migration lies in differences in correctness guarantees, runtime trade-offs, and developer experience. The article emphasizes that Go is suitable for rapid iteration and has a vast ecosystem, while Rust achieves stricter compile-time checks and higher robustness through ownership mechanisms, type systems, and lack of garbage collection.

The article provides a detailed comparison of toolchain commands for Go and Rust, noting that Rust's cargo tool is more integrated, covering functions such as building, testing, formatting, static analysis, and security audits, whereas Go relies more on third-party tools. The author also analyzes fundamental differences in memory management, error handling, and concurrency models between the two, arguing that Rust shifts many checks that Go relies on runtime or convention to compile time through its type system, thereby reducing runtime errors.

The author openly criticizes Go's language design but acknowledges its success and widespread use, aiming to remain objective in the article. The primary motivation for migrating to Rust is not performance bottlenecks but rather Go's cumbersome error handling, risk of null pointers, and lack of advanced type features. The article also mentions that Rust has a steep learning curve, but once mastered, developers will find that the compiler handles much of the work previously requiring manual memory and checks.

Overall, the article provides practical comparative perspectives and migration advice for development teams considering moving from Go to Rust, highlighting Rust's advantages in correctness and control in backend services and how to implement migration gradually.

• * *

https://news.ycombinator.com/item?id=48259808

• For web backend work, the Go language is more suitable due to its goroutines and extensive mature libraries, especially those used internally by Google which have been thoroughly tested, whereas Rust's libraries are relatively less mature.
• Rust's error handling system is complex, with multiple incompatible error types, making error propagation cumbersome, while Go's error handling, though verbose, is unified and simple.
• Rust's "?" syntax simplifies error handling, but the lack of a unified error type remains a pain point.
• Many new languages fail to unify fundamental types (such as error types, boolean types, multi-dimensional arrays, etc.) in their early stages, leading to multiple incompatible implementations later on, affecting fields such as numerical computation and graphic modeling.
• Modern AI-assisted programming makes Rust's complexity less of an issue, with Rust code performing excellently in terms of performance and reliability.
• Currently, the Rust code generated by mainstream LLMs is of poor quality, often misusing string types and poorly designed global variables, and lacking maturity in lifecycle management, requiring intervention from experienced developers for correction.
• Strict code inspection tools (such as clippy) can significantly improve the quality of AI-generated Rust code.
• A simpler language is not necessarily better for AI programming; languages with more safety checks and constraints can enable AI to produce higher-quality code.
• Code generated by LLMs generally requires substantial refactoring to reach a maintainable and efficient level.
• With AI assistance, Rust development efficiency has improved significantly, especially for specific tasks (such as SQL repository interfaces), without relying on traditional ORM tools.
• Code review has become the current bottleneck; while AI can accelerate code writing, code review for complex languages still requires significant manual effort.
• In contrast, Go code, due to its weak type system and verbose error handling, is harder to read and review, whereas Rust, although having a steep learning curve, has high information density, making the code more readable.

The article first introduced the basic instructions and working principle of the Minsky register machine: two unbounded counters and a finite state machine, which perform computation through incrementing and decrementing counters and conditional jumps. The author then mapped this model to the automation environment of Jira, where registers correspond to the number of different types of issues associated with an Epic, the program counter corresponds to the state of the Epic, instructions correspond to automation rules, and the clock is implemented by automated triggers or external re-triggers.

In the specific implementation, the author designed an addition program that adds the value of register A to register B. By creating specific workflow states and automation rules, the program performs operations to add and delete Bug and Task type associated issues, and controls the program flow through changes in the Epic state. Experimental results showed that the final state of the Epic and the number of associated issues accurately reflected the result of the addition operation.

Additionally, the article introduced the use of Jira's "convert issue type" feature to simplify Minsky machine operations and achieve the calculation of the Fibonacci sequence. This program uses three states and three registers, generating the Fibonacci sequence through type conversion and state transitions, demonstrating more complex computational capabilities.

The conclusion section notes that although there are execution chain depth and resource limitations in the Jira cloud environment, theoretically, the automation language of Jira supports unbounded register machine computation, thus possessing Turing completeness. This explains why complex Jira automation rules are essentially programs.

Investigations show that WorkTitans and MIRhosting were the most commonly used networks for pro-Russian cyber attacks targeting Danish government agencies during the 2025 November local elections. Nesterenko denies any knowledge of the situation and claims to have terminated his cooperation with the Neculiti brothers after sanctions took effect. MIRhosting also stated that they are conducting an internal investigation and have temporarily suspended services for WorkTitans.

Nesterenko was born in Nizhny Novgorod, Russia, and was once a child prodigy pianist. In 2004, he founded the parent company of MIRhosting, Innovation IT Solutions Corp., which had hosted websites for hacker organizations during the 2008 Russo-Georgian War. Nesterenko emphasizes that his company does not support cybercrime or sanctions evasion, stating that his arrest has caused serious harm to him and his company.

Less information is available about Zinad, who has been low-profile since last year, has blocked his LinkedIn account, and has not responded to multiple contact requests. Nesterenko claims that Zinad is not an employee of MIRhosting but rather assisted with some matters under normal business relations between companies. However, previous emails show that Zinad was a member of the company's legal team, and public records indicate that he was the official contact person for the MIRhosting office in Almere. Multiple attempts to reach Zinad have failed, and he has maintained an evasive stance.

• * *

https://news.ycombinator.com/item?id=48266906

• These seized servers are not legitimate hosting companies but front companies for Russian intelligence agencies, primarily serving intelligence activities and not open to regular users.
• Some users reported using similar services, noting poor quality and frequent user bans, but they did provide legitimate temporary VPS services.
• There is speculation that the seized companies inherited customers and equipment from sanctioned entities, suggesting business ties with sanctioned entities.
• Some emphasize that sanctioned companies are not equivalent to government institutions and should not be equated simply.
• It is argued that providing website and customer services does not prove legitimacy; such hosting services mainly support illegal activities.
• Security professionals express amazement at the professionalism and complexity of cybercrime infrastructure, noting that those involved have the capability for legitimate work but choose to support criminal activities.
• Some share the motivations of individuals arrested for cybercrime, suggesting they derive a sense of superiority and thrill from illegal activities, ultimately being caught due to mistakes.
• The discussion touches on regional stereotypes, with some pointing to similar criminal tendencies among certain populations in Eastern Europe, while others oppose generalizing by region or ethnicity.
• Regarding regional differences in crime rates, some view it as observation based on statistics rather than racial discrimination, requiring evidence.
• Others believe that Eastern Europe has a wealth of technical talent, but historical opportunities have limited some into the realm of cybercrime.

Certainly有不少人尝试过，但问题在于每个Jira实例都像是一个多层嵌套的自定义属性的混乱结构，这些属性是由于旧的失败迁移和新的组织策略演变而来的。

而且很多时候，API可以执行UI不允许的操作，但大家都依赖UI来进行操作，因此你最终会陷入一些奇怪的问题中，因为你没有注意到需要将custom_field_5537与custom_field_442配对，否则它不会出现在其他人的仪表板上。此外，它声称custom_field_10995是一个整数类型的字段，并且在XML中返回整数，但在创建任务时（而不是更新！）你必须使用一堆未文档化的魔法常量字符串，否则会收到无用的错误信息。网页界面则不会这样做（HTML和请求中只是整数），而且只有80%的字符串与下拉菜单中的显示文本匹配。

Automating Jira是我经历过的最糟糕的编程体验。我完全相信更简单的配置存在，而且它们可能相当容易，但天啊。

遗憾的是，这一切努力仍然非常值得。强烈推荐。

Groxx

Here is the translation of the provided Markdown text:

There is a lot to read here. I am curious where the reflections on the ‘mystery of the person’ will go: a brief search doesn’t show further mention. The encyclical seems to focus more on exhortations for us, humans, than on the nature of AI. Probably wise at this stage. I feel it is not AI that is either positive or negative, but its use of it, and the call to the growth of private industry as more powerful among nation-states is a strong statement for an institution like the Vatican to make:

Technological power thus takes on an unprecedented, predominantly “private” aspect, which makes it even more challenging to discern, govern, and direct such power toward the common good.

vintagedave

A form of uniformity that eliminates diversity, choosing homogeneity over integration.

Unrelated to AI, but in this era of anti-diversity, fairness, and inclusivity, this well supports the breadth of humanity.

So, we must avoid "Babel Syndrome," the worship of profit at the expense of the weak, a uniformity that erases differences, and the pretense that a single language—be it digital—can transform everything, including the mystery of the person, into data and performance.

There is a lot to read here. I am curious where the reflections on the ‘mystery of the person’ will go: a brief search doesn’t show further mention. The encyclical seems to focus more on exhortations for us, humans, than on the nature of AI. Probably wise at this stage. I feel it is not AI that is either positive or negative, but its use, and the call to the growth of private industry as more powerful among nation-states is a strong statement for an institution like the Vatican to make:

Technological power thus takes on an unprecedented, predominantly “private” aspect, making it even more challenging to discern, govern, and direct such power toward the common good.

• * *

Migrating from Go to Rust [#](https://supertechfans.com/cn/post/2026-05-26-HackerNews/#migrating-from-go-to-rust-1)

https://news.ycombinator.com/item?id=48261529

This is a strange document that attempts to serve as both a migration guide and an advocacy document for Rust.

Ultimately, if you have to ask, the Rust vs. Go consideration boils down almost completely to “do you want a managed runtime or not.” A generation of Rust programmers has convinced itself that “managed runtime” is bad, that not having one is an important feature. But that’s obviously false: there are more programming domains where you want a managed runtime than ones where you don’t.

That’s not an argument for defaulting to Go in all those cases! There are plenty of subjective reasons to prefer Rust. I miss match when I write Go (though I don’t miss tokio and async Rust). They’re both perfectly legitimate choices in virtually any case where you don’t have to distort the problem space to fit them in (e.g., trying to write a Go LKM would be a strange move).

The Rust vs. Go debate is a strange and awkward backwater of our field. Large portions of the industry are happily building entire systems in Python or Node, and smirking at the oddballs arguing over which statically typed compiled language to use. Python vs. (Rust|Go) is a real question. Rust vs. Go isn’t.

tptacek

This is a strange document that attempts to serve as both a migration guide and an advocacy document for Rust.

Ultimately, if you have to ask, the Rust vs. Go comparison almost entirely comes down to whether you want a managed runtime. A generation of Rust programmers has convinced themselves that a managed runtime is bad, and that not having one is an important feature. But that’s obviously false: there are more programming domains where you need a managed runtime than where you don’t.

This doesn’t mean you should default to Go in all those cases! There are plenty of subjective reasons to prefer Rust. I miss match when writing Go (though I don’t miss tokio and async Rust). In virtually any case where you don’t have to distort the problem space to fit them in, they are both perfectly legitimate choices (e.g., trying to write a Go LKM would be a strange move).

The Rust vs. Go debate is a strange and awkward corner of our field. Large portions of the industry are happily building entire systems in Python or Node, and smirking at those who argue over which statically typed compiled language to use. Python vs. (Rust|Go) is a real question. Rust vs. Go isn’t.

• * *

Magnifica Humanitas [#](https://supertechfans.com/cn/post/2026-05-26-HackerNews/#magnifica-humanitas-3)

https://news.ycombinator.com/item?id=48267427

I have found that, for many statements about what AI should do, I would actually be happier if the letters “AI” were replaced with “companies.”

Lerc

I have found that, for many statements about what AI should do, I would actually be happier if the letters “AI” were replaced with “companies.”

• * *

I spent 50 hours drawing a line graph [#](https://supertechfans.com/cn/post/2026-05-26-HackerNews/#i-spent-50-hours-drawing-a-line-graph)

https://news.ycombinator.com/item?id=48257572

I build furniture, and while I do my design work digitally for remote clients, I do my shop drawings by hand.

One super helpful tip I got from an actual trained draftsman is to use harder pencil lead for your layout and construction lines. Like 6H to 9H. You’ll get a much lighter line to erase later. It’ll also hold a finer point for longer.

I prefer lead holders to wooden pencils. They take 2mm lead, and you sharpen them with a lead pointer. K&E pointers are readily available on eBay, as are the abrasive cups that do the actual sharpening. The plastic trash can ones will get the job done, but are unsatisfying from a tactile standpoint.

Finding a decent lead holder is tricky. The Alvin one I bought is too loose and the lead slips up into it. The Staedtler one doesn’t close tightly at the tip and supports the lead well enough to prevent breaking. The Prismacolor one is satisfactory, and I inherited a vintage one that I love from the aforementioned draftsman.

I recommend an erasing shield to make revising your pencil work without erasing too much. Another person I know with an art background tipped me off to putting tracing paper over your main drawing to iterate on details before committing them to paper to reduce erasing.

Drafting vellum is pretty forgiving of erasing, but it has a toothier surface that can get a little dingy if you’re working on a drawing for a while. I’ve never tried Bristol board; I don’t need immaculate drawings for reproduction, just good enough ones to build from.

Happy drawing. It’s an immensely satisfying process for me. If you’re detail oriented, you’ll likely find it enjoyable too.

mauvehaus

I build furniture, and while I do my design work digitally for remote clients, I do my shop drawings by hand.

One super helpful tip I got from an actual trained draftsman is to use harder pencil lead for your layout and construction lines, such as 6H to 9H. This will give you a much lighter line that is easier to erase later. It will also maintain a finer point for a longer time.

I prefer lead holders to wooden pencils. They use 2mm lead and are sharpened with a lead pointer. K&E pointers are readily available on eBay, as are the abrasive cups used for sharpening. Plastic trash can sharpeners will work, but they are unsatisfying from a tactile perspective.

Finding a decent lead holder is tricky. The Alvin one I bought is too loose, and the lead slips into it. The Staedtler one doesn’t close tightly at the tip and doesn’t support the lead well enough to prevent breakage. The Prismacolor one is satisfactory, and I inherited a vintage one from the aforementioned draftsman that I love.

I recommend using an erasing shield to revise your pencil work without erasing too much. Another person with an art background suggested placing tracing paper over your main drawing to iterate on details before committing them to paper, which reduces the need for erasing.

Drafting vellum is forgiving of erasing, but it has a rougher surface that can become dirty if you work on a drawing for a while. I haven’t tried Bristol board because I don’t need perfect reproductions; I just need drawings that are good enough to build from.

Happy drawing. It’s an immensely satisfying process for me. If you’re detail-oriented, you’ll likely find it enjoyable too.

• * *

DeepSeek makes the V4 Pro price discount permanent [#](https://supertechfans.com/cn/post/2026-05-26-HackerNews/#deepseek-makes-the-v4-pro-price-discount-permanent)

When he quoted Tolkien, my heart stopped. This passage might provide you with a suggestion on how to live a virtuous life:

When he quoted Tolkien's words, my heart stopped beating. These words might offer some guidance on how to live a virtuous life:

"The twentieth-century Catholic author J.R.R. Tolkien, through the voice of a protagonist in one of his novels, described our responsibility in this way: 'Our task is not to master all the tides of the world, but to do what is in us for the succour of those years wherein we are set, uprooting the evil in the fields that we know, so that those who live after may have clean earth to till.' The civilization of love will not arise from a single or spectacular gesture, but from the sum total of small and steadfast acts of fidelity that serve as a bulwark against dehumanization."

• * *

Memory has grown to nearly two-thirds of AI chip c… [#](https://supertechfans.com/cn/post/2026-05-26-HackerNews/#memory-has-grown-to-nearly-two-thirds-of-ai-chip-c-2)

https://news.ycombinator.com/item?id=48259811

An interesting implication of this is that AI inference and training has a path to a ~3x hardware cost reduction (and maybe ~2x total cost reduction) without any technical innovation whatsoever, we just need to wait for DRAM supply to meet demand (either by manufacturing scaling or just waiting for the current rate of manufacturing to fill the demand spike).

gpm

One interesting implication of this is that AI inference and training could see a ~3x reduction in hardware costs (and possibly a ~2x reduction in total costs) without any technological innovation, we just need to wait for DRAM supply to meet demand (either through increased manufacturing capacity or simply waiting for current production rates to meet the demand spike).

• * *

Microsoft open-sources “the earliest DOS source co… [#](https://supertechfans.com/cn/post/2026-05-26-HackerNews/#microsoft-open-sources-the-earliest-dos-source-co)

https://news.ycombinator.com/item?id=48254089

It is rare that I say this but, thanks MS! Arguably just as, if not more, important is the BASIC that they wrote. That was what they actually wanted to do. DOS just got them the contract with IBM. For decades MS was really a developer tools company with a side biz of writing operating systems and other misc software. They also open sourced that BASIC code too [1].

[1] https://opensource.microsoft.com/blog/2025/09/03/microsoft-open-source-historic-6502-basic/

jmward01

I rarely say this, but thanks Microsoft! Arguably, the BASIC they wrote is even more important, if not more so. That was what they truly wanted to achieve. DOS was merely a tool to secure the IBM contract. For decades, Microsoft was essentially a developer tools company that also produced operating systems and other miscellaneous software. They have also open-sourced that BASIC code [1].

[1] https://opensource.microsoft.com/blog/2025/09/03/microsoft-open-source-historic-6502-basic/

• * *

The Eternal Sloptember [#](https://supertechfans.com/cn/post/2026-05-26-HackerNews/#the-eternal-sloptember)

https://news.ycombinator.com/item?id=48263653

I think a lot of the problem with the current discourse is how black-and-white it is. Either you’re a luddite or “ai pilled”.

In most cases, LLMs can get you 80-95% of the way, sometimes less, sometimes more. And heck, sometimes, it just gets you somewhere wrong.

But it seems everyone is arguing about whether LLMs can be perfect software engineers in isolation running in a closet, and using that to say that LLMs do not have a massive potential in other scenarios.

Sometimes, I like to imagine how much more productive most organizations could be from the things that the internet gave us, even to this day. Most companies never really do even a fraction of what is possible. That helps to ground my view of LLMs as well.

The fault dear Brutus isn’t in our language models, but in ourselves.

cafkafk

I think a lot of the issue with the current discourse is how polarized it is. You're either a luddite or an "AI enthusiast."

In most cases, LLMs can handle 80-95% of tasks, sometimes less, sometimes more. And sometimes, they just get it wrong.

But it seems everyone is debating whether LLMs can be perfect software engineers operating in isolation, and using that to argue that LLMs lack significant potential in other scenarios.

Sometimes, I like to imagine how much more productive most organizations could be with the tools the internet has provided, even today. Most companies barely scratch the surface of what's possible. This perspective also grounds my view of LLMs.

The fault, dear Brutus, lies not in our language models, but in ourselves.