Second wave speakers for AIE Europe and CFP for AIE World’s Fair are announced today, and OpenCode is confirmed for Miami! We’ll also be in Melbourne & Singapore. Editor: This is the latest in our guest post program, where we will publish AI Engineering essays worth considering, even if we don’t personally agree with them — having just shipped an AI review tool, this is one of those cases where I
I have met a lot of developers in my life. Lately, I asked myself: “What does it take to be one of the best? What do they all have in common?” In the hope that this will be an inspiration to someone out there, I wrote down the traits I observed in the most exceptional people in our craft. I wish I had that list when I was starting out. Had I followed this path, it would have saved me a lot of time
ついにBitNet Llama8Bが登場! CPUのみで爆速推論するLLM,BitNet.cpp｜shi3z
科学の世界では、それまでの常識が覆ることを俗に「パラダイムシフト」と呼ぶ。 しかし、もしもAIの世界にパラダイムシフトという言葉があるとしたら、今週の人類は一体何度のパラダイムシフトを経験しただろうか。 そのトドメの一撃とも言えるのが、BitNetのLlama8B版だ。 Lllama-8B構造で学習された最初のBitNetであり、全てを変えてしまうゲームチェンジャーでもある。CPUのみで秒間5-20トークンを出力する。超強力なLLM推論エンジンの出現だ。 BitNetとは、そもそも1.58ビットに相当する情報量で、本来は4ビット以上必要な大規模言語モデルの計算を劇的に高速化する技術である。 LLMの推論には通常は巨大な浮動小数点数(8ビットから16ビット)の、大量の乗算(掛け算)が必要なため、GPUなどの特殊な半導体を必要としていた。特にNVIDIAのGPUがこの目的にマッチしていたので今
There’s a kind of polite fiction at the heart of the software industry. It goes something like this: Estimating how long software projects will take is very hard, but not impossible. A skilled engineering team can, with time and effort, learn how long it will take for them to deliver work, which will in turn allow their organization to make good business plans. This is, of course, false. As every
今回はJetBrainsが公開した、Zig言語の作者アンドリュー・ケリー（Andrew Kelley）氏へのロングインタビューを取り上げます。 アンドリュー氏は2018年に勤めていた会社を辞め、プログラミング言語Zigの開発に専念してきた人物です。8年が経った今、ZigはGhosttyやTigerBeetle、さらにはUberのクロスコンパイルなどを支える存在になりました。一方で「PRやIssueへの厳格なAI禁止ポリシー」「年間67万ドルの非営利財団」「GitHubからの離脱」「10年経っても1.0が出ない理由」など、賛否を呼ぶ決断でも知られています。 この記事では、動画で語られた内容を流れに沿って整理しお届けします。AIに関心のあるエンジニアやビジネスパーソンにとって、ひとつの「対極の哲学」として読んでいただける内容です。 そもそもなぜZigを作ったのかインタビューはストレートな問いか
訳文；「そこにはなんの報酬もありません。このゲームが何を為していてどう機能しているのか、ただただ見ていたかったのです」ジェンキンズ、カーソン、ホッキング、『Outer Wilds』へつづく２，３の論考 - すやすや眠るみたくすらすら書けたら
翻訳の秋が今年もきました。また去年みたく面白い記事をいくつか見つけて勝手に紹介したいところです！ 去年アップした『訳文；「"好奇心駆動型の冒険"とでも言うべき特殊なタイプの冒険に報酬を与えるゲームをつくりたい、それが『Outer Wilds』の主目的です」Ａ・ビーチャム氏の論文より』で翻訳紹介した論考のなかで参照文献として挙げられていた文献のうち2つと、別記事1つの計3つを勝手に紹介します。 ヘンリー・ジェンキンズ著『GAME DESIGN AS NARRATIVE ARCHITECTURE（物語による建築物としてのゲームデザイン）』 ボニー・ルバーク取材『Clint Hocking Speaks Out On The Virtues Of Exploration(クリント・ホッキングが語る冒険の美徳)』 ドン・カーソン著『Environmental Storytelling: Creat
Understanding how Facebook disappeared from the Internet
Understanding how Facebook disappeared from the Internet2021-10-04 The Internet - A Network of Networks “Facebook can't be down, can it?”, we thought, for a second. Today at 15:51 UTC, we opened an internal incident entitled "Facebook DNS lookup returning SERVFAIL" because we were worried that something was wrong with our DNS resolver 1.1.1.1. But as we were about to post on our public status pag
【Claude Code】Agent Teamを役割ではなく「4つの性格」で組んだら、議論の性質が変わった
はじめに — Agent Teamをどう「編成」するか Claude CodeにAgent Teamが登場して、複数のエージェントを並列に動かせるようになりました。 最初に考えたのは「役割で分ける」ことでした。フロントエンド担当、バックエンド担当、テスト担当。タスクを分割して、それぞれに任せる。効率的に見えます。 これはタスクを並列に処理したい場面では有効です。ただ、設計判断など 「考える」ことが中心のタスクでは、別のアプローチもあるのではないかと感じました。 「役割」ではなく「考え方」で分けたらどうなるか——そこで試したのが 「性格で分ける」 というアプローチでした。 4つの性格定義 役割の代わりに、以下の4つの「思考スタイル」をエージェントに割り当てています。 実際にはプロンプトで思考の方向性を定義しており、便宜上「性格」と呼んでいます。 The Pragmatist（実用主義者） "
It turns out we've all been using MCP wrong. Most agents today use MCP by directly exposing the "tools" to the LLM. We tried something different: Convert the MCP tools into a TypeScript API, and then ask an LLM to write code that calls that API. The results are striking: We found agents are able to handle many more tools, and more complex tools, when those tools are presented as a TypeScript API r
エドワード・サイードのオリエンタリズム批判エドワード・サイードの1978年の著作『Orientalism』（参考1、参考2、補足）は、西洋が東洋（Orient）を「他者」として構築し、帝国主義的な支配を正当化するための知識体系が築かれており、それがオリエント学のような高尚とされる論説として体系的に築かれているために、ヨーロッパの言論空間で中東を語る語彙を使おうとすると否応なく帝国主義的な意識に引きずられてしまうとしてして批判した書物である（これに対する反論はこちらが分かりやすいので興味があればどうぞ：反論はともあれ、オリエンタリズムという言葉は現在は《悪い意味でのエキゾチシズム》という意味で定着している）。 My contention is that Orientalism is fundamentally a political doctrine willed over the Orie
merge_vs_rebase_vs_squash.md I get asked pretty regularly what my opinion is on merge commits vs rebasing vs squashing. I've typed up this response so many times that I've decided to just put it in a gist so I can reference it whenever it comes up again. I use merge, squash, rebase all situationally. I believe they all have their merits but their usage depends on the context. I think anyone who sa
Photos from Lars Tunbjörk’s “Office” seriesConsider Toto. If you spend much time in American public bathrooms, or rather if you’re simply a particularly attentive patron of American public bathrooms, you’ll probably have noticed Toto’s toilets at some point or another: they’re distinguished by a quite memorable serif-font “TOTO” logo. Toto toilets aren’t quite dominant in American bathrooms, since
Tetris is a classic time-waster, both in and outside of the office. What good is any computing device if it can’t play this game? Tokyo System House certainly thought so, and ported it to the NEC mini5 line of CP/M-based word processors. Let’s preserve it for future generations and then see what it’s like! I’ve been trying to get this game for a bit. First, I had been looking at the online old-gam
大実験！ChatGPTは競プロの問題を解けるのか (2024年5月版) - E869120's Blog
1. はじめに 2024 年 5 月 14 日、OpenAI 社から新たな生成 AI「GPT-4o」が発表され、世界に大きな衝撃を与えました。これまでの GPT-4 よりも性能を向上させただけでなく1、音声や画像のリアルタイム処理も実現し、さらに応答速度が大幅に速くなりました。「ついにシンギュラリティが来てしまったか」「まるで SF の世界を生きているような感覚だ」という感想も見受けられました。 しかし、いくら生成 AI とはいえ、競技プログラミングの問題を解くのは非常に難しいです。なぜなら競技プログラミングでは、問題文を理解する能力、プログラムを実装する能力だけでなく、より速く答えを求められる解法 (アルゴリズム) を考える能力も要求されるからです。もし ChatGPT が競技プログラミングを出来るようになれば他のあらゆるタスクをこなせるだろう、と考える人もいます。 それでは、現代最強の
Harness design for long-running application development
Published Mar 24, 2026 Harness design is key to performance at the frontier of agentic coding. Here's how we pushed Claude further in frontend design and long-running autonomous software engineering. Written by Prithvi Rajasekaran, a member of our Labs team. Over the past several months I’ve been working on two interconnected problems: getting Claude to produce high-quality frontend designs, and g
Consider SQLite
If you were creating a web app from scratch today, what database would you use? Probably the most frequent answer I see to this is Postgres, although there are a wide range of common answers: MySQL, MariaDB, Microsoft SQL Server, MongoDB, etc. Today I want you to consider: what if SQLite would do just fine? For those who are unfamiliar, SQLite is a implementation of SQL as a library — this means t
Go Scheduler
Go Scheduler Contents Introduction Compilation and Go Runtime Primitive Scheduler Scheduler Enhancement GMP Model Program Bootstrap Creating a Goroutine Schedule Loop Finding a Runnable Goroutine Goroutine Preemption Handling System Calls Network I/O and File I/O How netpoll Works Garbage Collector Common Functions Go Runtime APIs Disclaimer This blog post primarily focuses on Go 1.24 programming
OpenAI API ドキュメント 日本語訳｜#2 GET STARTED 後編｜Nobue Otsu｜地方で老舗パン屋を第三者承継
OpenAI API ドキュメントの日本語訳をこちらでまとめます。文字量の多いドキュメントなので、セクションごとに記事を分割しています。 今回は「GET STARTED 」のセクションからLibraries 、Models、TutorialsそしてUsage policiesを抜粋した後編です。 基本 DeepLで翻訳して、気になるところだけ書き換えています(ほぼ気になるところがないのが、DeepLのすごいところ)。原文との突き合わせができるようにはじめに原文を入れてますので、間違いなど見つけられましたら、ぜひご指摘ください。ご指摘箇所は随時反映させていただきます。 原文のリンクが有効になってますので、それぞれ必要な場合は原文リンクの方を参照ください。 前回のおさらいはこちら Python library｜Python ライブラリWe provide a Python library, w
The Coming Loop
I don’t prompt Claude anymore. I have loops running that prompt Claude and figuring out what to do. My job is to write loops. — Boris Cherny Over the last months I have watched more and more people build something on top of coding agents that feels meaningfully different from just using a coding agent. Some of this happens on top of Pi which is cool to see for sure! The pattern is the same everywh
As someone makes more money, expenses once considered luxuries can suddenly become seen as necessities: It’s called lifestyle creep. In the world of software development, we can suffer from a similar affliction: stack creep. Where hardware limitations once restricted developers to a minimalist approach, increased processing power, memory, and storage have led many down a more maximalist path. It’s
no slop grenade
What's a slop grenade? Pasting a massive AI-generated response into a chat or email where a human would write one sentence. It destroys the medium itself. Nobody writes essays in Slack. It's only possible because of AI copy-paste. It's like calling someone and asking "What time is the meeting?" and they read you a 10-page analysis of calendar management best practices. You asked a simple question.
AI 2027
Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, Romeo Dean We predict that the impact of superhuman AI over the next decade will be enormous, exceeding that of the Industrial Revolution. We wrote a scenario that represents our best guess about what that might look like.1 It’s informed by trend extrapolations, wargames, expert feedback, experience at OpenAI, and previous forecasting
まえがき 「……敢えて言うなら、医学と同様、翻訳もまた芸術なのだ。ただし、それは科学に基づいた芸術である。……」――ジョルジュ・ムーナン『翻訳の理論』 翻訳。すなわち、ある言語で表された物事を、また別の言語に置き換える行いは、本質的に困難を孕んでいる。これは、シニフィエあるいはシニフィアンといった大層な言語学用語を持ち出さずとも、感覚として了解できることであろう……鳩を撃つのに大砲は不要なのだ。現実にあるもの――それはカギカッコ付きの「現実」であって、たとえそれが実在であろうが非実在であろうが構わないのだが――つまり、脳が処理したイメージを、さらに言語という記号で表してみせる一連の動作によって、われわれは会話をし、文章を紡ぐ。それが塊になると、戯曲になったり小説になったり、あるいは詩歌になったりする。発信者が抱くもやもやとした概念、情景、人物、何だってよいが、それをそのままの形で受信者に伝
At over thirty years old, HTTP is still the foundation of the web and one of the Internet’s most popular protocols—not just for browsing, watching videos and listening to music, but also for apps, machine-to-machine communication, and even as a basis for building other protocols, forming what some refer to as a “second waist” in the classic Internet hourglass diagram. What makes HTTP so successful
Writing Good Unit Tests
In my first job as a software engineer the project I was working on grew out of control. As often happens, use cases were added without proper refactoring and with more users we were drowning in support tickets and nothing seemed to work. It was clear we needed to push for quality so we started adding unit tests. The tech lead at the time mandated high test coverage. But here's the reality: writin
Today we’re excited to announce the release of TypeScript 5.0! This release brings many new features, while aiming to make TypeScript smaller, simpler, and faster. We’ve implemented the new decorators standard, added functionality to better support ESM projects in Node and bundlers, provided new ways for library authors to control generic inference, expanded our JSDoc functionality, simplified con
Simple.css Framework
Are you using Simple.css? If you are, it would be great if you considered buying me a coffee to say thanks. Things like this really help open source software thrive. You can Buy Me A Coffee or even sponsor me on GitHub. ❤️ Simple.css is a CSS framework that makes semantic HTML look good, really quickly. Simple.css is mostly classless, which means that you can integrate Simple.css with plain HTML a
AWS deleted my 10-year account and all data without warning
After 10 years as an AWS customer and open-source contributor, they deleted my account and all data with zero warning. Here's how AWS's 'verification' process became a digital execution, and why you should never trust cloud providers with your only copy of anything. On July 23, 2025, AWS deleted my 10-year-old account and every byte of data I had stored with them. No warning. No grace period. No r
Here Are the Attack Plans That Trump’s Advisers Shared on Signal - The Atlantic
So, about that Signal chat. On Monday, shortly after we published a story about a massive Trump-administration security breach, a reporter asked the secretary of defense, Pete Hegseth, why he had shared plans about a forthcoming attack on Yemen on the Signal messaging app. He answered, “Nobody was texting war plans. And that’s all I have to say about that.” At a Senate hearing yesterday, the direc
npx skills install jpoehnelt/skills/agent-dx-cli-scale Human DX optimizes for discoverability and forgiveness. Agent DX optimizes for predictability and defense-in-depth. These are different enough that retrofitting a human-first CLI for agents is a losing bet. I built a CLI for Google Workspace — agents first. Not “built a CLI, then noticed agents were using it.” From Day One, the design assumpti
AI 2027
We predict that the impact of superhuman AI over the next decade will be enormous, exceeding that of the Industrial Revolution. We wrote a scenario that represents our best guess about what that might look like.1 It’s informed by trend extrapolations, wargames, expert feedback, experience at OpenAI, and previous forecasting successes.2 (Added Nov 22 2025: To prevent misunderstandings: we don't kno
Published Sep 11, 2025 Agents are only as effective as the tools we give them. We share how to write high-quality tools and evaluations, and how you can boost performance by using Claude to optimize its tools for itself. The Model Context Protocol (MCP) can empower LLM agents with potentially hundreds of tools to solve real-world tasks. But how do we make those tools maximally effective? In this p
November 1, 2023 Hello! I’m slowly working on explaining git. One of my biggest problems is that after almost 15 years of using git, I’ve become very used to git’s idiosyncracies and it’s easy for me to forget what’s confusing about it. So I asked people on Mastodon: what git jargon do you find confusing? thinking of writing a blog post that explains some of git’s weirder terminology: “detached HE
axios Compromised on npm - Malicious Versions Drop Remote Access Trojan - StepSecurity
Read the full behind-the-scenes story of how StepSecurity detected and helped remediate this attack: Behind the Scenes: How StepSecurity Detected and Helped Remediate the Largest npm Supply Chain Attack axios is the most popular JavaScript HTTP client library with over 100 million weekly downloads. On March 30, 2026, StepSecurity identified two malicious versions of the widely used axios HTTP clie
Don't write clean code, write CRISP code — Bitfield Consulting
I’m sure we’re all in favour of “clean code”, but it’s one of those motherhood-and-apple-pie things that no one can reasonably disagree with. Who wants to write dirty code, unless maybe it’s for a porn site? The problem, of course, is that few of us can agree on what “clean code” means, and how to get there. A rule like “methods should only do one thing”, looks great on a T-shirt, but it’s not so
Developers are increasingly relying on AI coding assistants to accelerate our daily workflows. These tools can autocomplete functions, suggest bug fixes, and even generate entire modules or MVPs. Yet, as many of us have learned, the quality of the AI’s output depends largely on the quality of the prompt you provide. In other words, prompt engineering has become an essential skill. A poorly phrased
How Big Tech Runs Tech Projects and the Curious Absence of Scrum
Project management is a topic most people have strong opinions on, and I’m no exception. To answer the question of how different companies run engineering projects, I pulled in help from across the industry. In this issue we’ll cover: Project management approaches across the industry. An overview of a survey with over 100 companies represented, plus key takeaways.Project management at Big Tech. Ho
Fish 4.0: The Fish Of Theseus
About two years ago, our head maintainer @ridiculousfish opened what quickly became our most-read pull request: #9512 - Rewrite it in Rust Truth be told, we did not quite expect that to be as popular as it was. It was written as a bit of an in-joke for the fish developers first, and not really as a press release to be shared far and wide. We didn’t post it anywhere, but other people did, and we go
Published Jun 13, 2025 Our Research feature uses multiple Claude agents to explore complex topics more effectively. We share the engineering challenges and the lessons we learned from building this system. Claude now has Research capabilities that allow it to search across the web, Google Workspace, and any integrations to accomplish complex tasks. The journey of this multi-agent system from proto
A bit over a year ago I wrote How I use LLMs as a staff engineer. Here’s a brief summary of what I used AI for last year: Smart autocomplete with Copilot Short tactical changes in areas I don’t know well (always reviewed by a SME) Writing lots of use-once-and-throwaway research code Asking lots of questions to learn about new topics (e.g. the Unity game engine) Last-resort bugfixes, just in case i
j次のブックマーク
k前のブックマーク
lあとで読む
eコメント一覧を開く
oページを開く
How to Kill the Code Review
The Best Programmers I Know | Matthias Endler
How I estimate work as a staff software engineer
ZigがAIを禁止する理由。GitHub離脱、確固たる運営方針。｜直也テック
Code Mode: the better way to use MCP
現代に残されたオリエンタリズムのはけ口としての日本｜✨わん🐶にゃん😺癒し動画✨
Merge vs. Rebase vs. Squash
Why Japanese companies do so many different things
NEC’s Tetris Processor
Marie Kondo your software stack with open source
【連載#1】サイエンス・フィクションのなかの言語実験たち（鯨井久志） 第1回 ハーラン・エリスン ｜作品社
The state of HTTP in 2022
Announcing TypeScript 5.0 - TypeScript
You Need to Rewrite Your CLI for AI Agents
Writing effective tools for AI agents—using AI agents
Confusing git terminology
The Prompt Engineering Playbook for Programmers
How we built our multi-agent research system
How I use LLMs as a staff engineer in 2026