The first wave of AI disruption will not be unemployment. It will be organizational embarrassment. AI will expose every meeting, approval, report, and role that only exists because coordination was expensive. The real disruption starts inside the workflow.

AI is starting to scare people. Not because it failed. Because it keeps getting better. Models are becoming more capable, more autonomous and harder to predict at a speed institutions were never designed for. Now some of the people leading the AI race are talking about slowing down. That should get your attention. The question is no longer: “How smart can AI become?” It is: “How much control will humans still have when it does?”

AI’s biggest story isn’t a new model. The people racing to build the frontier are now asking to slow it down. Amodei. Altman. Musk. Hassabis. My take: this is not the end of the AI boom. It’s the start of the governance war. Who controls the brake may control the race.

Chatgpt Down Its like an internet link problem nowadays.

OpenAI’s 18% sequential growth looks solid… until you put it next to Anthropic more than doubling revenue in the same quarter and flipping to profitability. The underdog just became the revenue leader.

China just green-lit 10,000 Nvidia H200s each for ByteDance and Tencent.The catch? These chips are already two generations behind the ones Chinese companies are still banned from buying. This isn’t closing the AI gap — it’s just measuring how wide it still is. The post is from @coinbureau

21-year-old Harvard dropouts just raised nearly $2B and hit a $21B valuation on an AI chip idea from 4 months ago… with their own in-office data center and Jane Street already signed as customer #1 while poaching Nvidia talent.The barrier to building the next generation of silicon just collapsed.This is the new game. Follow if you want the real signal on AI hardware before the next $20B kid shows up.

Higgsfield may be showing where enterprise AI video is heading. ~$20M → $200M → $700M annualized revenue Valuation: $1.3B → $5.4B in 8 months But the bigger signal: businesses now generate most of its revenue. AI video is shifting from creative tool → business infrastructure. The real question isn’t how many videos AI can generate. It’s: How much of the global marketing budget can AI software absorb?

水素競争は「技術」から「ルール」へ 水素競争は、もう「誰が良い技術を持つか」だけではない。 日本は9月、第8回水素閣僚会議を開催する。 各国政府、企業、国際機関が集まり、水素・アンモニアの社会実装を議論する。 重要なのはここだ。 次の産業競争では、 技術を作る国より、ルールを作る国が強い。 日本は水素で「作る側」から 世界の標準を決める側へ行けるか。

水素競争は「技術」から「ルール」へ 水素競争は、もう「誰が良い技術を持つか」だけではない。 日本は9月、第8回水素閣僚会議を開催する。 各国政府、企業、国際機関が集まり、水素・アンモニアの社会実装を議論する。 重要なのはここだ。 次の産業競争では、 技術を作る国より、ルールを作る国が強い。 日本は水素で「作る側」から 世界の標準を決める側へ行けるか。

日本のスポーツは「試合」ではなく「輸出産業」へ 日本のスポーツ政策が変わり始めている。 政府は、国内リーグなどの海外展開を支援する補助金の二次公募を開始。 対象は映像配信、海外イベント、プロモーション、インバウンド施策。 1件最大1,500万円。 日本のスポーツは「試合を開催する産業」から、 海外でファンと需要を作る輸出産業へ。 次の勝負は、競技力だけではない。 世界でどう“売るか”だ。

AI doesn’t just remove tasks. It collapses the time between signal and decision. Companies still running monthly reviews and multi-layer approvals are about to look slow in ways that used to be invisible. The real competitive edge is no longer “having AI.” It’s redesigning the operating model so decisions move at the speed of the data. How many of your current approval steps would still exist if coordination were free?

UPDATE: Colombia’s earthquake disaster has become dramatically worse. The latest UNGRD balance: 294 DEAD. 3,935 injured. 320 missing. 353 rescued alive. The magnitude 7.4 earthquake struck near San José del Palmar, Chocó on August 10. Colombia’s Geological Service says it was the: LARGEST EARTHQUAKE RECORDED IN COLOMBIA THIS CENTURY. The scale of physical damage is becoming clearer: 14,493 homes destroyed 81,506 homes damaged 66 buildings collapsed 241 health facilities affected 2,612 educational facilities affected and damage across roads, bridges, water systems and airports. More than: 115,000 PEOPLE have been affected across: 448 MUNICIPALITIES. But another number needs care. UNGRD’s earlier Saturday update listed 320 people missing. Reuters’ later reporting says official estimates had fallen to roughly 140 missing, as authorities reconciled lists and located people previously unaccounted for. That is why rapidly changing disaster numbers must always carry a timestamp and source. The bigger picture is now unmistakable: EARTHQUAKE ↓ SEARCH + RESCUE ↓ MASS DISPLACEMENT ↓ INFRASTRUCTURE DAMAGE ↓ RECOVERY ↓ A NATIONAL RECONSTRUCTION CHALLENGE And the rescue phase is entering its hardest point. A Cali fire captain told Reuters that crews had only one site remaining where signs of life were still being detected; elsewhere, operations were increasingly shifting toward recovering victims.

「10ギガ」と実際の速度は別物 「10ギガ回線」でも、Wi-Fi実測の中央値は約285Mbps。 RBB SPEED TESTが6月の8万1420件の計測ログから10Gサービスを分析。 平均330.79Mbps、中央値284.88Mbpsだった。 もちろんルーター、端末、宅内環境の影響は大きい。 だからこそ重要なのは「最大10Gbps」という看板ではない。 通信競争は、**最大値より“毎日出る速度”**を競う段階ではないか。

日本コンテンツの海外展開が、 「完成品を輸出する」から「最初から世界と作る」へ。 ドワンゴのBEYOND BORDERs第1弾は、 初音ミク × Grimes × さたぱんP。 国内ボカロPと海外アーティストが共同制作し、米国向けA&R・マーケティングもLA拠点企業が担当する。 次の輸出力は、翻訳ではない。 制作工程そのものを国際化できるかだ。

Nvidia just drew the line.They will finance the first 5 GW of OpenAI’s campus. They will not underwrite the full 10 GW fantasy.That is not weakness. That is the first adult capital allocation decision in the entire AI infrastructure boom.Everyone else is still writing blank checks and calling it “partnership.” Nvidia kept the GPU lock-in and forced the market to price real risk. In a sector drowning in FOMO balance sheets, this is how you stay the picks-and-shovels king instead of becoming the mine’s co-signer.Who actually understands this is maturity, not retreat? Drop your take.

Apple just ran into the geopolitics of AI. AI boom → memory shortage → chip prices rise → Apple looks to China → Washington says: don’t. Apple is exploring memory chips from China’s CXMT. The bigger story? AI isn’t just reshaping software. It’s reshaping the global semiconductor supply chain. Who wins from this? 🇺🇸 Micron 🇰🇷 Samsung 🇰🇷 SK hynix 🇨🇳 CXMT Pick one.

AIの利用料金が、電気料金のように「時間帯」で変わり始める。 DeepSeekはV4 Pro / Flashにピーク・オフピーク料金を導入。 V4 Proの出力は、ピーク時3.96ドル／100万トークン。オフピークはその半額。 企業のAIコスト最適化は、もう「どのモデルを使うか」だけではない。 いつ推論するかまで設計する、AI版FinOpsの時代が来る。

Nvidia just taught the entire AI industry a lesson in capital allocation. They could have underwritten the full $250B OpenAI campus fantasy. Instead they capped exposure, kept the GPU lock-in, and forced the market to price real risk.This is the difference between selling the picks and owning the entire gold mine’s debt.Most of the sector is still in FOMO mode. Nvidia just switched to survivor mode. Who’s actually watching the balance sheets instead of the hype? Drop your take.

ミリ波の本命は「スマホ直結」ではない？ コミケ108で、28GHzミリ波の意外な使い方が見える。 ソフトバンクは、ミリ波を対応スマホだけに飛ばすのではなく、FWAのバックホールとしてWi-Fiスポットにも利用。 つまり、ミリ波非対応スマホも恩恵を受けられる。 次世代通信の価値は「最新端末だけを速くする」ことではない。 既存端末の“裏側”を高速化することかもしれない。

Three months ago, Ronny Chieng told Harvard graduates: “Your generation’s mission is to destroy AI.” It sounded extreme. Now AI leaders are warning about something similar: AI may make us faster — while making us think less. Maybe the real AI risk isn’t replacement. It’s dependency. Was he right? youtu.be/ORq_Hi5dB-g?si=TMmS…