くなぎさ ネト/AI/健康 retweeted
いや、スペースXですらあんなに失敗してるのに死亡事故や労災聞かないが?と思って調べたら、宇宙関連で区切っても桁違いに労災事案が起きてるらしく、闇深くてそっ閉じした。この話はおしまい。皆さん本当に長生きして御安全に過ごしてくださいませ。死して賞は貰えないのだよ我々は汗
くなぎさ ネト/AI/健康 retweeted
one moving camera, and nobody labeled the depth.
Depth Anything V4 reconstructs a dynamic 4D scene from monocular video with Riemannian Flow Matching on 4D Gaussian Splatting. no human-annotated depth labels. it outperforms Depth Anything V3 and per-scene 4D-GS.
arxiv.org/abs/2608.18388
#PhysicalAI #ComputerVision
みんな当たり前のようにリバースエンジニアリングしてて草生える
俺も匿名でやろうかなぁ
I turned my DJI Mavic into a home security drone with REA & Opus 5.5.
this f*cking is insane..
I reverse-engineered my DJI drone with Opus 5.5
no cloud. no GPS on the target. it all runs on the phone. iPhone or Android.
here's the build step by step:
step 1 → plug a spare phone into the Mavic remote and paste the prompt below into Claude Code on Opus 5.5
step 2 → Opus checks your drone against the DJI Mobile SDK v5 list. not on it?
stop there. no firmware hacks
step 3 → it builds a small app on the official DJI SDK: Swift for iPhone, Kotlin for Android
step 4 → sync with the alarm: Home Assistant sees it trigger → push to your phone → one tap "Launch"
step 5 → eyes: a tiny YOLO model on the phone's chip. Core ML on iPhone, LiteRT on Android
step 6 → memory: a Kalman filter. he runs behind a fence → LOCKED → COASTING → re-locks on the same ID
step 7 → hands: Virtual Stick at 10 Hz. box center → yaw, box size → distance, never closer than 8 m
step 8 → at the property line it stops, climbs to 30 m and keeps the camera on him until he's out of sight
step 9 → lands back on the pad. the clip with track, distance and timestamps is already on your phone
the alarm wakes you up. the drone already knows where he went...
Find full prompt below
Then, read the article below on how to reverse-engineer anything with Opus 5.5.
クラス4レーザーを道端で実験して大丈夫なわけないだろ頭湧いてんのかこの杉原海ってやつは。
これが創業初期から資金調達数億円してるスタートアップのCEOなのヤバすぎる。
宇宙での電力転送を目的とする人達が、試験用のレーザーを川岸で発射した件
CEO曰く「大丈夫です!!」
では「大丈夫」か否か、こちらから動画や写真をご覧になり、皆様でご判断ください
これは私が保存した物では無く、CEOご本人が投稿された物ですので、誤解なきよう
search.yahoo.co.jp/realtime/…
くなぎさ ネト/AI/健康 retweeted
宇宙での電力転送を目的とする人達が、試験用のレーザーを川岸で発射した件
CEO曰く「大丈夫です!!」
では「大丈夫」か否か、こちらから動画や写真をご覧になり、皆様でご判断ください
これは私が保存した物では無く、CEOご本人が投稿された物ですので、誤解なきよう
search.yahoo.co.jp/realtime/…
鰤もナルトも好きだけど、超長期連載の弊害として話が発散して後期が微妙だったのは認めざるを得ない事実。
だけど、今あまり冷笑されないのはファンと比べて冷笑するアンチはいつまでも冷笑する熱量がないからなんかな。
くなぎさ ネト/AI/健康 retweeted
Yann LeCun has changed the game for robotics.
Research he co-authored tackles a problem inside AI world models: their internal representations can make planning harder than it needs to be.
A world model predicts what happens after an action. But even useful visual features can form curved paths in its internal “map,” making distance to a goal misleading.
The researchers introduced Temporal Straightening.
During training, it encourages those internal paths to become straighter, so the planner gets a more useful measure of progress.
The results:
More stable gradient-based planning
Higher success rates across tested goal-reaching tasks
Code available for others to explore
The approach changes how the model represents movement, rather than rewriting every RL algorithm to work in curved space.
For robotics, that’s a useful direction: build world models whose internal maps help an agent choose its next move.























