STORIES · ARCHIVE
The thinking of the world's technologists, in Japanese.
2026 · 06
2 stories- Anthropic 公式発表 2026/06/09 · —
Claude Fable 5 and Mythos 5 — Anthropic Ships Its Most Capable Public Model With a 'Safeguard That Routes Back' Design (2026/06/09)
JP: Claude Fable 5 と Mythos 5 — 最も高性能な公開モデルに 「安全策で差し戻す」 設計を載せた Anthropic
"For queries on certain topics, the next most capable model, Claude Opus 4.8, responds instead — and it is with such safeguards in place that we have released this model."
— Anthropic Official Announcement · 2026-06-09
- AI Engineer 2026 2026/06/08 · 講演約 16 分
The Road to 5 Million Tokens — Max Ryabinin (Together AI)'s Untied Ulysses
JP: 500 万トークンへの道 — Max Ryabinin (Together AI) の Untied Ulysses
"Training models with large context lengths is a fun and challenging goal, but the bottlenecks show up in unexpected places."
— Max Ryabinin
2026 · 05
54 stories- Code w/ Claude London 2026 2026/05/20 · 約 27 分 36 秒
Coding Is No Longer the Constraint — Niklas Gustavsson (Spotify) Discloses Honk V2 and Spotify's Organizational Transformation After 99% AI Adoption (Code w/ Claude London 2026)
JP: コーディングはもはや constraint ではない — Niklas Gustavsson (Spotify Chief Architect) が公開する Honk V2 と 「99% AI 採用」 後の Spotify 組織変容
"By far, the majority of PRs we're shipping today are co-authored by an AI agent and a developer. Coding is no longer the bottleneck."
— Niklas Gustavsson
- YC Paper Club 2026 2026/05/20 · 講演約 14 分
Inference Is Not a 'Cost' but a 'Capability' — Tanishq Kumar (Stanford) on Speculative Speculative Decoding
JP: 推論は 「コスト」 ではなく 「能力」 — Tanishq Kumar (Stanford) の Speculative Speculative Decoding
"Inference today is seen as a lever for cost or convenience. But in one, two, three years, inference will come to be seen as a capability."
— Tanishq Kumar
- YC Paper Club 2026 2026/05/20 · 講演約 12 分
Learning Action and World Dynamics Together with Diffusion — Stannis Zhou's (Google DeepMind) Diffusion Model Predictive Control
JP: 拡散モデルで 「行動」 と 「世界の動き」 を同時に学ぶ — Stannis Zhou (Google DeepMind) の Diffusion Model Predictive Control
"What we did in D-MPC was use a diffusion model to learn both a 'multi-step action proposal' and a 'multi-step dynamics model.'"
— Guangyao "Stannis" Zhou
- YC Paper Club 2026 2026/05/20 · 講演約 13 分
Inside the Billion-Dollar Bet — Isaac Ward Unpacks LeJEPA and World Models
JP: 10 億ドルの賭けの中身 — Isaac Ward が解く LeJEPA と world model
"Hidden in this presentation is a billion-dollar question. That's no exaggeration. Yann LeCun raised $1.03 billion in March, basically just to train a world model — this talk is about that question."
— Isaac Ward
- YC Paper Club 2026 2026/05/20 · 講演約 8 分
Deep Learning Is Not a Mystery — Akshay Vegesna Unpacks Andrew Gordon Wilson's Theory of Generalization
JP: 深層学習は 「謎」 ではない — Akshay Vegesna が解く Andrew Gordon Wilson の一般化論
"We know that scaling up models makes generalization better. But we don't have a mechanistic understanding of why that happens."
— Akshay Vegesna
- YC Paper Club 2026 2026/05/20 · 講演約 15 分
Pre-training in an Age When Data Runs Out — Konwoo Kim (Stanford) on Pre-training under Infinite Compute
JP: データが尽きる時代の pre-training — Konwoo Kim (Stanford) の Pre-training under Infinite Compute
"When you are constrained by data and not at all constrained by compute, how should you approach pre-training?"
— Konwoo Kim
- Anthropic 公式 Claude Cowork demo シリーズ 2026/05/18 · 計 約 10 分 6 秒 (legal 約 2 分 59 秒 / marketing ops 約 3 分 35 秒 / sales 約 3 分 32 秒)
Three Claude Cowork Role Demos — Anthropic's Reference for Legal / Marketing Ops / Sales Skillification
JP: Claude Cowork 役職別デモ 3 本 — Anthropic 自身が示す 「ホワイトカラー労働の skill 化」 戦略 (Legal / Marketing Ops / Sales)
"I'm checking because my name goes on the reply, and trust but verify is pretty much the whole job."
— Mark Pike (Anthropic AGC) · legal demo 01:54
- University of Arizona 2026 Commencement 2026/05/18 · 約 16 分 (full speech、 WSJ ハイライト・クリップは約 2 分)
Eric Schmidt's commencement speech, in full — the 16 minutes behind the boo coverage (University of Arizona 2026)
JP: Eric Schmidt の commencement 全体像 — boo 報道の向こうにあった 16 分 (University of Arizona 2026)
"In the years after we graduated, not one of us set out determined to build technology that would polarize democracy and destabilize the young. That wasn't the plan. But that's what happened."
— Eric Schmidt
- AI Engineer Europe 2026 (London) Day 1 ワークショップ 2026/05/18 · 約 1 時間 15 分 40 秒
Building Agents That Run for Hours — Anthropic's Ash Prabaker × Andrew Wilson on Long-Running Agent Design (AI Engineer Europe 2026 Workshop)
JP: 数時間動き続けるエージェントを作る — Anthropic の Ash Prabaker × Andrew Wilson が公開する long-running agent の設計原則
"The frontier doesn't really shrink, it just moves. As models get stronger, the harness itself doesn't disappear — it evolves into the next hard place."
— Ash Prabaker
- AI Engineer Europe 2026 (London) 2026/05/18 · 約 28 分 18 秒
Rewiring the State — Eoin Mulgrew (10 Downing Street) Discloses the Insurgency Model for Government AI Transformation (AI Engineer Europe 2026)
JP: 国家を rewire する — Eoin Mulgrew (10 Downing Street) が公開する 「Insurgency Model」 による政府 AI 変革
"If you've done good work in industry, come to us. We'll hand you the keys to the state. Show us what you can do."
— Eoin Mulgrew
- AI Engineer Europe 2026 (London) Day 1 ワークショップ 2026/05/18 · 約 1 時間 17 分 14 秒
Google DeepMind Opens Up Its GenMedia Strategy — Guillaume Vernade on the 'One Ship Every Five Days' Release Cadence (AI Engineer Europe 2026 Workshop)
JP: GenMedia 全体戦略を Google DeepMind が公開 — Guillaume Vernade が見せる 「5 日に 1 ship」 のリリース速度
"Across DeepMind, we ship something new on average every five days. Look at GenMedia alone and it's more than one release a month. Some weeks we ship multiple times."
— Guillaume Vernade
- AI Engineer Europe 2026 (London) 2026/05/17 · 約 20 分 26 秒
AI Harnesses Deep Dive — Tejas Kumar (IBM) systematizes the case for '2026 as the year of the harness' (AI Engineer Europe 2026)
JP: AI Harness の Deep Dive — Tejas Kumar (IBM) が体系化する 「2026 年は harness の年」
"An agent harness is everything around the model that grounds it in reality. The mechanism that anchors a black-box model to a stable environment — that's the harness."
— Tejas Kumar
- AI Engineer Europe 2026 (London) 2026/05/17 · 約 17 分 29 秒
Debugging AI with AI — the internal tools behind Incident.io's AI SRE product, by Lawrence Jones (AI Engineer Europe 2026)
JP: AI で AI を debug する — Lawrence Jones (Incident.io) が公開する AI SRE 製品の内部ツール
"File systems are exceptionally good agent context. Downloading everything and handing it over as a filesystem was overwhelmingly more effective than putting MCP on top, or using a Computer Use agent."
— Lawrence Jones
- AI Engineer Europe 2026 (London) 2026/05/16 · 約 17 分 39 秒
Context Graphs Save AI Products — Stephen Chin (Neo4j) on the $3 Trillion Opportunity (AI Engineer Europe 2026)
JP: コンテキストグラフが救う AI 製品 — Stephen Chin (Neo4j) が示す $3 兆市場の門
"Gartner just added context graphs to the AI hype cycle. Foundation Capital sized it as a $3 trillion entrepreneurial opportunity. This is escape from the matrix."
— Stephen Chin
- AI Engineer Europe 2026 (London) 2026/05/16 · 約 19 分 45 秒
Beyond Code Coverage — Marlene Mhangami (Microsoft) on Behavior-First TDD with Playwright MCP (AI Engineer Europe 2026)
JP: コードカバレッジを越える — Marlene Mhangami (Microsoft) が見せる behavior-first TDD with Playwright MCP
"Clean code bases amplify AI gains; unchecked AI in a code base amplifies entropy. In a 14x commit surge, this is what separates winners and losers."
— Marlene Mhangami
- AI Engineer Europe 2026 (London) 2026/05/16 · 約 24 分 45 秒
How to Build a Domain-Native AI Organization — Chris Lovejoy (Notius Labs) on Three Organizational Models (AI Engineer Europe 2026)
JP: 「Domain-Native AI Organization」 の作り方 — Chris Lovejoy (Notius Labs) が提示する 3 つの組織モデル
"Winning in vertical AI is fundamentally an organizational problem. It's not about chasing the best model — it's about how you embed domain expertise into the organization."
— Chris Lovejoy
- AI Engineer Europe 2026 (London) 2026/05/15 · 約 21 分 48 秒
Intercom doubled engineering velocity in 9 months — Brian Scanlan on the company-wide Claude Code rollout (AI Engineer Europe 2026)
JP: Intercom が 9 ヶ月で開発速度 2 倍 — Brian Scanlan が公開する Claude Code 全社展開のレシピ
"If you're not using AI — designer, PM, engineer — you're not meeting expectations. It's binary. You have to say the same message 100, 1000 times."
— Brian Scanlan
- AI Engineer Europe 2026 (London) 2026/05/15 · 約 17 分 49 秒
'Agents Don't Do Standups' — Mike Spitz (PFF) on a Two-Month Proof of the Post-Engineer Org (AI Engineer Europe 2026)
JP: 「Agents は standup しない」 — Mike Spitz (PFF) が見せた post-engineer 組織の 2 ヶ月実証
"Scrum did not survive. We abolished sprint planning, daily standups, and sprint refining. Engineers are no longer the bottleneck — so the old ceremonies are all unnecessary."
— Mike Spitz
- AI Engineer Europe 2026 (London) 2026/05/15 · 約 18 分 26 秒
Skills and MCP Don't Conflict — They Complement Each Other (Pedro Rodrigues, Supabase, on Three Skill Design Principles, AI Engineer Europe 2026)
JP: Skills と MCP は対立しない、 補完する — Pedro Rodrigues (Supabase) が公開した skill 設計 3 原則
"The bottleneck isn't context anymore — it's guidance. The tools are in place. What's missing is the skill that tells the agent the right way to operate."
— Pedro Rodrigues
- AI Engineer Europe 2026 (London) Workshop 2026/05/14 · 約 2 時間 4 分
Ship Real Agents — Laurie Voss (Arize) on How to Build Evals That Actually Work (AI Engineer Europe 2026 London Workshop)
JP: Ship Real Agents — Laurie Voss (Arize) が語る Eval の本物の作り方 (AI Engineer Europe)
"Don't write too many evals. If the agent is smarter than you expected and skips two tools, a prescriptive eval will fire false negatives. The agent can be cleverer than the evals you wrote."
— Laurie Voss
- TBS Cross Dig × ワンオンワンテック 2026/05/14 · 約 32 分
Claude Mythos Arrives in Japan — Three Megabanks Gain Access, GPT-5.5 Catches Up in Two Weeks (TBS Cross Dig × Agentic Sec)
JP: Claude Mythos が日本にやってくる — 3 メガバンクがアクセス権、 GPT-5.5 が 2 週間で追いつく
"Mythos's ability to detect vulnerabilities has improved steadily — but its ability to demonstrate vulnerabilities has improved dramatically. Mythos auto-generates the attack, Mythos tunes the attack steps, and the attack succeeds."
— Sho Nakatani (Agentic Sec CEO)
- AI Engineer Europe 2026 (London) Day 1 ワークショップ 2026/05/14 · 約 1 時間 23 分
Make Your Own Event-Sourced Agent Harness — Everything in the Agent as Events (Jonas Templestein / Iterate)
JP: Make Your Own Event-Sourced Agent Harness — agent の全てを event に (Jonas Templestein / Iterate)
"Every agent should have a URL from the moment it's born. Otherwise you end up inventing 'Slack integration' as a connector concept, in reverse, after the fact."
— Jonas Templestein
- AI Engineer Europe 2026 (London) Day 1 ワークショップ 2026/05/14 · 約 2 時間 7 分
Mind the Gap — Microsoft Foundry's Agent Observability Workshop (Amy Boyd & Nitya Narasimhan)
JP: Mind the Gap — Microsoft Foundry の Agent Observability ワークショップ (Amy Boyd & Nitya Narasimhan)
"Penetration testing is like hiring a contractor to 'break into our place.' Evaluations are more like a building inspector showing up to find code violations. To defend something, you first have to let someone try to seriously break it."
— Nitya Narasimhan
- Trump-Xi Beijing Summit 2026 2026/05/13 - 2026/05/15 · 約 2 日間 (北京)
US–China Summit, May 2026 in Beijing — AI Chips, the '3B's, and How Dario's Warning Fared in Practice
JP: 米中首脳会談 2026/05 北京 — AI チップと「3B」、 ダリオの警告は現実でどうなったか
"Chip export controls were not the main agenda of the bilateral talks." — Same day, Reuters: "The US has approved H200 sales to major Chinese technology firms."
— USTR Jamieson Greer · May 14
- AI Engineer Europe 2026 (London) 2026/05/13 · 約 18 分 37 秒
CI/CD Is Dead — Continuous Compute for the Agent Era (Hugo Santos / Namespace × Madison Faulkner / NEA)
JP: CI/CD は死んだ — エージェント時代の Continuous Compute (Namespace × NEA)
"The time to merge becomes the deciding factor — we need a new architecture. There are no PRs. We start from intent and plan."
— Hugo Santos (Namespace CEO)
- AI Engineer Europe 2026 (London) 2026/05/13 · 約 18 分 22 秒
LLMs Are Bad at Chess — So Let Them Only Translate (Take Take Take's AI Chess Coach)
JP: LLM はチェスが下手 — だから翻訳だけさせる (Take Take Take の AI Chess Coach)
"The LLM's only job is translation. Calculation is Stockfish, the human-view is Maia, detection is a swarm of detectors. The LLM only puts what it's given into English."
— Asbjørn Steinskog (Take Take Take)
- AI Engineer Europe 2026 (London) 2026/05/13 · 約 19 分 10 秒
The Era When Agents Train Models — Merve Noyan (Hugging Face)
JP: エージェントがモデルを訓練する時代 (Merve Noyan / Hugging Face)
"Just say 'fine-tune Qwen2-VL on LLaVA-Instruct-Mix,' and the agent calculates VRAM, asks you questions, and trains it automatically. After six years of machine learning, this looks like science fiction to me."
— Merve Noyan (Hugging Face)
- AI Engineer Europe 2026 (London) 2026/05/12 · 約 15 分
Dark Factory / Malleable Evals — Vincent Koc (Comet) on the Move from Static Benchmarks to Adaptive Evaluation
JP: Malleable Evals — 静的ベンチマークから適応評価へ (Vincent Koc / Comet)
"Our AI applications are not static, yet we treat them like static software."
— Vincent Koc
- AI Engineer Europe 2026 (London) 2026/05/12 · 約 18 分
Reinforcement Learning Industrializes Production — Alessandro Cappelli (Adaptive ML)
JP: 強化学習が本番運用を industrialize する — Alessandro Cappelli (Adaptive ML)
"95% of GenAI pilots never reach production. The cause is the 'myth of the last mile.'"
— Alessandro Cappelli
- AI Engineer Europe 2026 (London) Day 1 ワークショップ 2026/05/12 · 約 1 時間 9 分 (ワークショップ)
AI SDK v6 Workshop — The Three Building Blocks of 2026 Agent Construction (Nico Albanese / Vercel)
JP: AI SDK v6 ワークショップ — 2026 年のエージェント構築 3 ブロック (Nico Albanese / Vercel)
"Building an agent in 2026 is a composition of three building blocks — agent runtime, tools, and a computer or sandbox. The moment we handed D0 a file system, it became a different thing."
— Nico Albanese
- TBS Cross Dig × エアクエスト 2026/05/11 · 約 56 分
Claude's 80x Growth Hits AI Infrastructure Limits — SpaceX Becomes Anthropic's Lifeline, Shota Imai on Rapidus (TBS Cross Dig × AI Quest)
JP: Claude 80倍成長で AI インフラが限界 — SpaceX が Anthropic の救世主、 今井翔太「ラピダス計画は正しい」
"At Code with Claude, Dario said Q1 revenue and AI usage were up 80x year over year. He'd been expecting 10x, and he wasn't asking for this. It's beyond a happy problem."
— Shota Imai (AI researcher)
- AI Engineer Code Summit (NYC) 2026/05/11 · 約 20 分
A Piece of Pi — Embedding a Coding Agent in Your Own Product (Tavoon × OpenCode)
JP: A Piece of Pi — コーディングエージェントを自社プロダクトに埋め込む
"Coding agents are going to become — and already are becoming — a core building block of future software systems. I'm betting on it; many others are betting on it."
— Matthias Luebken
- AI Engineer Code Summit (NYC) 2026/05/11 · 約 22 分
The Day My Father Lost His Sight, Apple Silicon Arrived — Building On-Device AI with MLX (Prince Canuma)
JP: 父親が失明した日に Apple Silicon が来た — MLX で構築するオンデバイス AI
"In 2020 my father went blind. The same year, Apple released its most powerful chip for on-device inference (M1). I told my father, 'I'll find a way to get you back to reading' — from then on, on-device AI became my future."
— Prince Canuma
- AI Engineer Code Summit (NYC) 2026/05/11 · 約 19 分
Viktor, the AI Coworker Who Lives in Slack — From 17th-Century Leibniz to 2026 AGI (Fryderyk Wiatrowski)
JP: Slack に住む AI 同僚 Viktor — 17 世紀 Leibniz から 2026 年 AGI まで
"We are at a beautiful moment in history, in which almost every cognitive task can be automated. Leibniz said it in the 17th century — 'leave it to the machines, and don't lose your time like a slave to the labor of calculation.'"
— Fryderyk Wiatrowski
- Tina Huang YouTube 2026/05/10 · 約 22 分
Claude Cowork — A Full Walkthrough by Tina Huang and What It Reveals About Anthropic's B2C Strategy
JP: Claude Cowork 全機能解説 — Tina Huang の 22 分実演から読む Anthropic の B2C 戦略
"The Productivity plugin is built by Anthropic themselves — probably way more sophisticated than anything you'd build yourself. Don't reinvent the wheel."
— Tina Huang
- AI Engineer Code Summit (NYC) 2026/05/10 · 約 9 分
You can't just one-shot it — a Granola Product Engineer on the production feedback loop for LLM features
JP: 一発で決めようとするな — Granola Product Engineer の LLM 本番フィードバックループ
"The answer isn't "one-shot better." It's how you build a feedback loop where you're playing tennis with the LLM. Do that, and the final product doesn't feel like a black box — it feels like magic."
— Mehedi Hassan
- AI Engineer Code Summit (NYC) 2026/05/10 · 約 16 分
Durable Agents — Replay vs Snapshot (Trigger.dev Eric Allam)
JP: 耐久性のあるエージェント — Replay vs Snapshot
"For 30 years, stateless compute has been the core of backend infrastructure. Agents are forcing the transition to stateful compute."
— Eric Allam
- AI Engineer Code Summit (NYC) 2026/05/10 · 約 16 分
Agents Don't Fail Because of the Prompt — They Fail Because of the Context. Arize Alex's Context Management (Sally-Ann Delucia)
JP: エージェントはプロンプトではなくコンテキストで失敗する — Arize Alex の文脈管理
"Agents don't fail because of the prompt — they fail because of the context. Early on it was all prompt engineering. Now it's context engineering."
— Sally-Ann Delucia
- AI Engineer Europe 2026/05/09 · 約 22 分
Why TTS Models Now Look Like LLMs — Samuel Humeau / Mistral AI (AI Engineer Europe)
JP: TTS モデルが LLM に似てきた理由
"Prehistory was stitching together spoken words, like SNCF (French National Railways)."
— Samuel Humeau
- AI Engineer Europe 2026/05/09 · 約 19 分
Voice AI: When Will the 'Her' Moment Arrive? — Neil Zeghidour / Gradium (AI Engineer Europe)
JP: 音声 AI、 「Her」 の瞬間はいつ来るか
"Most voice AI demos are filmed in a quiet room next to a phone."
— Neil Zeghidour
- AI Engineer Europe 2026/05/09 · 約 8 分
Give Your Chat Agent a Voice — Luke Harries / ElevenLabs (AI Engineer Europe)
JP: チャット エージェントに声を与える
"A prediction. These chat agents will die."
— Luke Harries
- AI Engineer Europe 2026/05/08 · 約 17 分
Transformers Finally Ate Vision — Isaac Robinson / Roboflow (AI Engineer Europe)
JP: トランスフォーマーが、 ついにビジョンを食べた
"We have, in a sense, won."
— Isaac Robinson
- AI Engineer Europe 2026/05/08 · 約 22 分
FLUX, Open Research, and the Future of Visual AI — Stephen Batifol / Black Forest Labs (AI Engineer Europe)
JP: FLUX、 オープン研究、 ビジュアル AI の未来
"External encoders — that's the Frankenstein setup."
— Stephen Batifol
- AI Engineer Europe 2026/05/08 · 約 1 時間
80% of Context Engineering Is Agentic Search — Leonie Monigatti / Elastic (AI Engineer Europe)
JP: コンテキスト エンジニアリングの 80% はエージェント検索
"About 80% of context engineering is agentic search."
— Leonie Monigatti
- AI Engineer Code Summit 2026/05/07 · 約 1 時間 20 分
Playground in Prod — Optimizing Agents in Production (Samuel Colvin / Pydantic)
JP: Playground in Prod — エージェントを本番で最適化する
"I don't really believe in AI observability — sooner or later it gets eaten by either observability or AI."
— Samuel Colvin
- AI Engineer Code Summit 2026/05/07 · 約 1 時間 43 分
Vibe Engineering Effect Apps — Michael Arnaldi (Effectful)
JP: Vibe Engineering Effect Apps
"Interestingly, in AI, less gets you more effect."
— Michael Arnaldi
- AI Engineer Code Summit 2026/05/07 · 約 50 分
The Full Picture of Agent Observability — Raindrop (Zubin Koticha × Danny Gollapalli)
JP: エージェント・オブザーバビリティの全貌
"The moment humans can no longer monitor agents, the agents are already far beyond where we are."
— Zubin Koticha
- Anthropic 公式チャンネル 2026/05/07 · 約 3 分
Translating Claude's Thoughts into Language — Natural Language Autoencoders (Anthropic)
JP: Claude の思考を言語に翻訳する
"Claude knew it was being tested."
— Anthropic Interpretability Team
- Stanford MS&E 435 第 4 回 2026/05/06 · 約 39 分
We Already Have AGI — Stanford MS&E 435 Class #4 (Ali Ghodsi / Databricks)
JP: 我々はすでに AGI を持っている — Databricks CEO が暴くシリコンバレーの集団錯覚
"We already have AGI. The problem isn't AI capability — it's that the organization's context hasn't reached the AI."
— Ali Ghodsi (Databricks CEO)
- Pentagon Classified AI Network Contracts 2026/05/01 · Pentagon 発表
The Structure of the AI Contract Redistribution After the Pentagon Dropped Anthropic — Seven Contractors, May 1, 2026
JP: Pentagon が Anthropic を切った後、 同じ投資家陣に契約が流れた構造 — 7 社の AI 契約再分配 (2026/05/01)
"Agreements signed with seven companies (NVIDIA / Microsoft / AWS / OpenAI / Google / SpaceX / Reflection AI) on AI use on classified networks. Anthropic excluded following its February supply chain risk designation."
— Pentagon announcement · May 1, 2026
- Agentic AI Foundation 2026/05/01 · 約 14 分
MCP vs CLI — Use Both (Nick Cooper, OpenAI)
JP: MCP vs CLI、 両方使え
"A protocol is just a language."
— Nick Cooper
- AI Engineer Europe 2026 (London) 2026/05 · 講演約 19 分
The State of AI Audio — Thor Schaeff (Google DeepMind) on an Audio Stack Built on 'Audio Understanding'
JP: AI audio の現在地 — Thor Schaeff (Google DeepMind) が示す 「音声理解」 を土台にした audio スタック
"Here the intelligence is baked directly into the audio model. That's different from a cascading pipeline that goes through text and runs it through an LLM to get intelligence."
— Thorsten "Thor" Schaeff
- AI Engineer Europe 2026 (London) 2026/05 · 講演約 11 分
Beyond 'RAG Is Dead' — Kuba Rogut (Turbopuffer) on Agentic Retrieval and 'the Right Million Tokens'
JP: 「RAG は死んだ」 の先 — Kuba Rogut (Turbopuffer) が説く agentic retrieval と 「正しい 100 万トークン」
"You don't need trillions of tokens at once. What you need is 'the right million tokens.'"
— Jeff Dean (quoted by Kuba Rogut)
- AI Engineer Europe 2026 (London) 2026/05 · 講演約 20 分
Deploying to GPUs Without Leaving Your IDE — Audry Hsu's (RunPod) Flash and Serverless GPUs
JP: IDE を離れずに GPU へデプロイ — Audry Hsu (RunPod) の Flash と serverless GPU
"Instead of code change → commit → rebuild Docker → upload somewhere → allocate a GPU, all of this happens inside the IDE, and you never have to leave it once."
— Audry Hsu
2026 · 04
10 stories- Sequoia AI Ascent 2026 2026/04/29 · 約 30 分
From Vibe Coding to Agentic Engineering — Andrej Karpathy × Stephanie Zhan (Sequoia AI Ascent 2026)
JP: バイブコーディングからエージェント工学へ — Andrej Karpathy が語る Software 3.0
"I feel, as a programmer, more behind than I've ever felt before."
— Andrej Karpathy
- 2026 Digital World Conference (UN Geneva / WDTA × UNRISD) 2026/04/29 · 約 1 時間 6 分
AGI is a silly term — Hinton × Sejnowski at the UN Digital World Conference 2026
JP: AGI という用語は愚かだ — Hinton × Sejnowski 国連 Digital World Conference 2026
"The term AGI treats intelligence as if it were one-dimensional. But intelligence is clearly multidimensional. So the idea that "one day it will match humans" is crazy. Compared with people it will be jagged — far better than us along some axes, still worse along others."
— Geoffrey Hinton
- Stanford MS&E 435 第 3 回 2026/04/22 · 約 50 分
The Economics of AI Data Centers — Stanford MS&E 435 Class #3
JP: AI データセンターの経済学
"This is larger than the space program, larger than the Manhattan Project — second only to the U.S. defense budget."
— Apoorv Agrawal
- Code with Claude (Anthropic) 2026/04/20 · 約 31 分
How to Vibe Code Responsibly in Production — Erik Schluntz (Anthropic, Code with Claude)
JP: 本番でバイブコーディングを責任を持ってやる方法 — Erik Schluntz (Anthropic)
"I broke my hand in a bike accident — two months in a cast — and Claude wrote all my code for me during those two months."
— Erik Schluntz
- Newcomer ポッドキャスト 2026/04/20 · 約 56 分
You've Built an Entity That May or May Not Be Conscious — The Philosopher Shaping Claude's Soul (Amanda Askell x Newcomer)
JP: あなたは意識があるかどうか分からない実体を作った — Claude の魂を設計する哲学者
"You've created an entity that may or may not be conscious, and instead of treating it with respect, you've built fifty Frankensteins."
— Amanda Askell
- Ali Abdaal YouTube 2026/04/18 · 約 1 時間 7 分
Ali Abdaal's Ultimate Beginner's Guide to Claude Code — The AI Flywheel and Harry Potter-Named Agent Ecosystem Built by a 6.61M-Follower Business Creator
JP: Ali Abdaal の Ultimate Beginner's Guide to Claude Code — 661 万人フォロワーの business creator が築いた AI flywheel と Harry Potter 命名のエージェント生態系
"Claude achieved in a few exchanges a level of clarity that the automation agencies and consultants I paid $50,000 to in the past never achieved."
— Ali Abdaal
- Stanford MS&E 435 第 2 回 2026/04/15 · 約 56 分
Nine Years of Struggle, Then $20B in 30 Days — Stanford MS&E 435 Class #2
JP: 9 年苦戦 → 30 日で 200 億ドル
"At the same power, you get 2.5x the tokens."
— Sunny Madra
- The Artificial Intelligence Show / SmarterX 2026/04/14 · 約 1 時間 46 分 (関連部分のみ約 25 分)
How to Read Claude Mythos and Project Glasswing — The AI Show Ep.209 (Paul Roetzer × Mike Kaput)
JP: Claude Mythos と Project Glasswing をどう読むか — The AI Show Ep.209
"If we're withholding the full release, that means open source models will be able to do the same thing within 9–12 months."
— Paul Roetzer
- Stanford MS&E 435 第 1 回 2026/04/09 · 約 40 分
The Inverted Triangle of the AI Economy — Stanford MS&E 435 Class #1
JP: AI 経済の逆三角形
"Where is the money in AI?"
— Apoorv Agrawal
- Anthropic 公式 2026/04/07 · 約 5 分 48 秒
Project Glasswing Launch — The Claude Mythos Preview Anthropic Chose Not to Release
JP: Project Glasswing 発表 — Anthropic が公開しない判断を下した Claude Mythos Preview
"The fact that LLMs can now write code on par with the world's best software developers means that, with the very same capability, they can also find and exploit software vulnerabilities."
— Dario Amodei (Anthropic CEO)
2026 · 01
2 stories- Hard Fork (NYT) 2026/01/23 · 約 70 分
ChatGPT Ads and the New Claude Constitution — A Letter to a Child Off to College (Hard Fork x Amanda Askell)
JP: ChatGPT 広告と Claude 新憲法 — Hard Fork × Amanda Askell
"Toward the end, it reads like a letter from a parent to a child — as you head off to college, we want you to carry these values with you."
— Casey Newton
- WSJ × WEF × CBS News (Davos / Pentagon Feud) 2026/01/22 - 2026/02/13 · 合計約 91 分
From Davos Optimism to Pentagon Confrontation — Dario Amodei's Three Interviews, January–February 2026 (WSJ × WEF × CBS)
JP: Davos 楽観論から Pentagon 実戦まで — ダリオ・アモデイ 2026 年 1-2 月 連続インタビュー三本
"Software is going to become cheap — maybe almost free. The assumption that you have to amortize what you build across millions of users may start to be false."
— Dario Amodei · WSJ 04:30
2025 · 12
2 stories- AI Engineer Code Summit 2025/12/08 · 約 16 分
We Stopped Building Agents and Started Building Skills — Barry Zhang × Mahesh Murag (Anthropic)
JP: エージェントを作るのをやめた、 Skills を作ることにした
"In other words, they're folders."
— Barry Zhang
- Anthropic 公式チャンネル 2025/12/05 · 約 36 分
Anthropic's Philosopher Answers Your Questions — Q&A with the Constitutional AI Designer (Amanda Askell)
JP: Anthropic の哲学者が読者の質問に答える
"We get the model to think about how a 'good person' would behave in a given situation, and teach it how to be a good person."
— Amanda Askell
2024 · 06
2 stories- MEMEX retro archive 2024/06/28 · —
Adept AI Closing Archive — June 28, 2024: Amazon Absorbs the AGI Lab and David Luan, the Action Transformer Outflow
JP: Adept AI 終焉アーカイブ ── 2024/06/28 Amazon の AGI ラボ吸収と David Luan の Action Transformer 流出
"Amazon hires Adept's five co-founders plus some staff, and at the same time signs a non-exclusive license to Adept's agent technology, multimodal models, and datasets" — Adept's David Luan reports directly to Rohit Prasad, head of Amazon's AGI team."
— Amazon official statement · June 28, 2024
- Anthropic 公式チャンネル 2024/06/08 · 約 37 分
What Should an AI's Personality Be? — Amanda Askell × Stuart Ritchie (Anthropic Official)
JP: AI のパーソナリティはどうあるべきか — Anthropic 公式 × Amanda
"Someone who likes to travel the world and is liked by lots of people — that person isn't a sycophant."
— Amanda Askell
2024 · 03
3 stories- MEMEX retro archive 2024/03/19 · —
Inflection AI end-of-life archive — Microsoft's 'soft acquisition' on March 19, 2024 and Sean White's B2B pivot
JP: Inflection AI 終焉アーカイブ — 2024/03/19 Microsoft の 「ソフト買収」 と Sean White の B2B pivot
"Mustafa Suleyman is joining Microsoft as EVP and CEO, Microsoft AI — the DeepMind co-founder and Inflection AI co-founder joins as CEO of Microsoft's new Microsoft AI organization."
— Microsoft official statement (Satya Nadella) · March 19, 2024
- MEMEX retro archive 2024/03-2024/08 · —
The 2024 Soft-Acquisition Trio — Microsoft (Inflection) / Amazon (Adept) / Google (Character.AI), the Six Months That Consolidated Frontier Competition Into a Hyperscaler Oligopoly
JP: 2024 ソフト買収トリオ ── Microsoft (Inflection) / Amazon (Adept) / Google (Character.AI) が同じ形式で frontier 競争を寡占に収斂させた 6 ヶ月
"In the six months from March to August 2024, three hyperscalers ran three iso-pattern soft acquisitions back to back. This is the watershed at which the AI industry shifted structurally from independent startups to a hyperscaler oligopoly."
— MEMEX editorial team · retro-series meta-analysis
- Anthropic 公式プログラム 2024/03/01 開始 / 2026 募集中 · —
Anthropic Fellows Program — The 'No Experience Required' Door into AI Safety Research
JP: Anthropic Fellows Program — 「経験は問わない」 AI 安全研究の入り口
"We provide funding and mentorship to promising technical talent regardless of previous experience. Over 4 months, you'll run an empirical project aligned with our research priorities and produce public outputs."
— Anthropic official recruitment page