So all the parts are still made in China. What Apple etc does is to ship these parts to India. I guess this hurts Apple than its does China.
Given China and India is in the same continent, I wonder if it makes sense for Apple to shift the factory from east part China to west - the dessert or farming region (I imagine). If that happens it can be blessing for Chinese people since they would get the employment which it would never get otherwise.
Another, new, huge, open-weight foundation model coming out from a Chinese AI lab. How many do we have something similar outside China today? It’s astonishing. They’re supposed to have limited GPU supply but it doesn’t seem to matter.1
This article portrays the life at a “AI town” in China which has strong ties to both an “Ivy league”-esque University there and the industry players like Alibaba. It’s kind of like San Francisco but they’re doing it in open-weight ways.
For these catching up players there are strategic values to be open-weight, and there is no way to see this as a pure altruism. But it doesn’t matter.
Well, it doesn’t matter if you’re one of the researchers or engineers working on these big frontier models. There are so many players in their own country! They are competitors but they must be friends too. And they’re all trolling (or even dominating) the US closed model players. How exciting it is to be in such a possibly once-in-a-lifetime moment. I cannot imagine.
Japanese IME stopped working on new Obsidian installation on my corp laptop so here you go… ↩︎
Roughly every five years, the cost to build a factory for making such chips doubles, and the number of companies that can do it halves. 25 years ago, there were about 40 such companies and the cost to build a fab was about $2-4 billion. Today, there are either two or three such companies left (depending on your optimism toward Intel) and the cost to build a fab is in excess of $100 billion. Project these trends forward another ten years and you can expect a single factory to cost nearly half a trillion dollars, and the number of companies that can do it should drop to less than one.
人類これ以上 CPU 作るのムリなのでは、という気分になるが、
Highly parallel chips designed from the ground up for extremely high defect tolerance could perhaps be manufactured with a vastly lower-quality, and by extension vastly cheaper chip fab.
I’m currently raising some pre-seed funds, aiming to build high-performance, high-efficiency, general-purpose CPUs designed from first principles. I’m also increasingly convinced that, with the right compromises and specialization, a startup-scale fab may be viable and is at least well-worth the experiment.
The deal was simple: We’ll pay you 20-50% below market rate, but in exchange, you get stability, reasonable work-life balance, and most importantly, no layoffs. This wasn’t written in any employee handbook, but everyone understood it. It was the Microsoft way.
By every objective measure, Microsoft is thriving—our market performance, strategic positioning, and growth all point up and to the right. (…) And yet, at the same time, we’ve undergone layoffs.
This is the enigma of success in an industry that has no franchise value.
きみたちめっちゃ franchise value で生きてるじゃんか!!検索みたいにエーアイに殺される心配もないじゃんか!!何いってんだ!!なでらっち、最近ちょっと Zuck 氏みたいなサイコパスになってきてない? Sam Altman にフラレて性格変わっちゃったのかい?
Federighi said he was more concerned the public would believe Apple was making too many compromises to get the software running on iPhones, the people said.
しかしこんなにコードいっぱい書かないとダメとは随分 low-level だな・・・とおもっていたら、DISCLAIMER: This is a copied version of ... with the intention of working around a key bug とのことで、普通は SDK を使えば良い。
Mobile App: Compose Multiplatform (Kotlin Multiplatform) or Flutter (Dart) using Material Design libraries and principles, when sharing code between Android and iOS. Jetpack Compose (Kotlin JVM) with Material Design principles or SwiftUI (Swift) for native apps targeted at either Android or iOS, respectively.
シレっと Flutter 推すのやめてくれませんかね・・・
Parallelism: Execute multiple independent tool calls in parallel when feasible (i.e. searching the codebase).
そんなことできるんだ!たしかにサーチが並列に動いてる時あるよな。実装みておきたいところ。
few-shot 的なものは <example> タグで囲んでいる。へえ。
plan を立てろと序盤に書いてあり、Plan 提案例が <example> に入っている。 しかしそこでは plan を箇条書きして許可を求めあとは agentic に動いてしまうので、ステップごとに一旦止まるとか計画をファイルに書くとかしたい場合は GEMINI.md で上書きしないとダメそうだな。
が、よくみると環境変数 GEMINI_SYSTEM_MD を使うと .gemini/system.md or お好みのパスから上書きできるらしい。いいじゃない!自分の Obsidian helper はコード関係の system prompt まったく必要ないからな。
こうした現実を考えると、今の所 model agnostic にいいものを作るのは大変そう。これは前にも思ったことだけれど, まだ色々荒削りなので抽象化をするのは早いってことじゃないかな。cline とか goose とかがんばってるけれど、しばらくは claude code なり gemini-cli なり model 固有のツールを使うほうが現実的に思える。
My mother and her husband are COBOL devs for a US state government. She works on the health insurance side for teachers and other state employees. Think claim processing.
Lots of batch jobs running at night. Their alert system is an actual human who calls my mom when jobs fail in the middle of the night.
My father is 75 and he still works, has his own software development company with his own back-office program written in cobol. He started a company back in 1991 with two other cobol programmers, they are retired now, and while almost all of the code they wrote has been replaced with c# code by younger programmers there are still some parts of the code written in cobol that he still maintains.
…if you woke up on a Casper mattress, worked out with a Peloton, Ubered to a WeWork, ordered on DoorDash for lunch, took a Lyft home, and ordered dinner through Postmates only to realize your partner had already started on a Blue Apron meal, your household had, in one day, interacted with eight unprofitable companies that collectively lost about $15 billion in one year.
But, the Tax Cuts and Jobs Act (TCJA) of 2017 amended Section 174, effective for tax years beginning after December 31, 2021. Starting in 2022, R&D expenditures must be capitalized and amortized over 5 years for domestic research (and 15 years for foreign research… which is pretty untenable.) This change eliminated the option to immediately deduct R&D costs, increasing tax liability for companies with significant research budgets in the short term.
The short version is: this rule change has increased taxable income for businesses in the short term, as they can no longer deduct R&D costs immediately.
These young engineers - squandering their opportunities to learn how things actually work - would briefly glance at the AI-generated code and/or explanation messages and continue producing more code when “it looks okay.”
Hyprland provides the latest Wayland features, dynamic tiling, all the eyecandy, powerful plugins and much more, while still being lightweight and responsive
Mountain View Castro St. のコーヒー屋としては Red Rock Coffee が圧倒的な人気を誇っており、たとえば WhatsApp 発祥の地としても知られている:
Brian Acton, one of the co-founders of WhatsApp, describes how he and his business partner had been working on their endeavor for years at Red Rock, describing it as a community resource and one of the best cafes in the area.
hipify_torch is a python utility to convert CUDA C/C++ code into HIP C/C++ code. It is NOT a parser; it does a smart but basic search-and-replace based on CUDA-to-HIP mappings which are specified in the hipify_torch module. It can also “hipify” the header include statements in your source code to ensure that it’s the hipified header files that are included.
Other changes on Galaxy Z Fold 7 can be found in the camera array, with a new 200MP camera at the core and backed up by a 12MP sensor with an ultrawide lens, and a third 10MP sensor for telephoto.
The ISOCELL HP2 packs 200-million 0.6-micrometer (μm) pixels in a 1/1.3” optical format, a sensor size that is widely used in 108MP main smartphone cameras.
画素がでかいわけではないので、たいした嬉しさはなさそうだな。むしろ画像サイズがデカくてストレージを圧迫する弊害があるのでは。フルサイズで撮ることは少なく、大抵は bin するのだろうけれど。
Earlier on Sunday, Trump’s treasury secretary said Musk should focus on running his companies and keep himself out of politics, a day after the world’s richest person – and a former White House adviser – announced the formation of a new political party.
Yeah, pretty accurate. Perhaps even eerily so. Upon first reading, I was surprised when it nailed age range, location, family size, etc.—details that I didn’t expect to be represented in a data set that I thought mostly consisted of Hacker News posts. Shows you how much can be inferred by a simple trail of “likes.”