静态网站悖论
个人网站的两种不同实现方式:一种是复杂的内容管理系统(CMS),另一种是简单的静态 HTML 文件。文章指出,尽管大多数普通用户倾向于使用复杂的解决方案(如 WordPress),但实际上,只有少数专业软件工程师能够选择更简单的静态网站。
via HackerNews 2024 10 09
前两天刚好听朋友说 square space 已经涨到了近乎搞笑的 $25 月费,做不用来盈利的个人博客实在难以 justify。这篇文章中吐槽得很在点子上:
normal users are stuck with a bunch of greedy clowns that make them pay for every little thing, all while wasting ungodly amounts of computational power to render what could have been a static website in 99% of cases.
普通用户被困在了一群屁大点功能都要收费的贪婪小丑手里,与此同时浪费着人神共愤额度的算力来渲染 99% 的情况下都可以作为静态的网站。
当然原文中说的“只有少数专业软件工程师才能选择更简单的静态网站”略微夸张并不认同,因为静态站至少是比 self-host 的动态 CMS 少太多维护了。我的 backlog 里也一直躺了篇安利新手用静态站并拉踩 WP 的文,不过网上这种文已经有无数了也还是拦不住前赴后继往各种 CMS 的坑里冲的新手,觉得写了又有什么意义呢就还搁着没写。(当然迟早会像以前反复造的无数轮子一样被废话欲战胜的 but not today)
#indieblog#newletter
🪐 In 2019, the asteroid 2019 OK startled astronomers when it was detected passing just 70,000 kilometers from Earth—less than a fifth of the distance to the Moon. This city-sized object, about 100 meters wide, slipped by with very little warning, highlighting how some space threats can go undetected until they're nearly upon us, and why vigilance is vital for discovering near-Earth objects in time. ✨
#asteroids⚡#nearEarth⚡#detection⚡#nasa⚡#galaxy⚡#stars⚡#astronomy⚡#universe⚡#cosmos⚡#space
👉subscribe Universe Mysteries
👉more Channels
🪐 Asteroid (2014) AA was discovered on January 1, 2014, just hours before it entered Earth's atmosphere near West Africa and burned up harmlessly. This rare event made 2014 AA only the second asteroid ever detected before impacting Earth, highlighting the importance of sky surveys to spot even small space rocks before they arrive. ✨
#asteroids⚡#impact⚡#detection⚡#nasa⚡#galaxy⚡#stars⚡#astronomy⚡#universe⚡#cosmos⚡#space
👉subscribe Universe Mysteries
👉more Channels
🌟WildDet3D: открытая модель монокулярной 3D-детекции по одному снимку.
Институт Аллена представил модель WildDet3D, которая по одному изображению строит 3D-рамки объектов: оценивает их положение, размер и ориентацию в метрических координатах.
Модель принимает сразу несколько типов промптов: текстовый запрос, клик по точке или готовый 2D-бокс от внешнего детектора.
🟡Архитектура состоит из 3 блоков
2D-детектор построен на SAM3 и обрабатывает все типы запросов.
Геометрическая ветка использует энкодер DINOv2 с обучаемым декодером глубины, учитывающим геометрию обзора: направления лучей камеры зашиваются через сферические гармоники, что снимает необходимость в отдельной калибровке.
Третий компонент, 3D-head, объединяет через кросс-внимание 2D-детекции с признаками глубины и поднимает их в полноценные 3D-боксы.
Если на инференсе доступны данные с LiDAR, ToF или стереокамеры, они подмешиваются в ту же геометрическую ветку без переобучения.
🟡Тесты
На бенчмарке Omni3D модель показывает 34,2 AP с текстовыми промптами (это +5,8 пункта к прежнему лидеру 3D-MOOD).
На zero-shot переносе на Argoverse 2 WildDet3D практически удваивает прежний результат: 40,3 ODS против 23,8.
На редких категориях из собственного бенчмарка WildDet3D-Bench успехи, разумеется, еще лучше - 47,4 AP против 2,4 у 3D-MOOD.
🟡Вместе с моделью вышло демо-приложение для iOS.
Оно использует видеопоток с камеры iPhone и данные LiDAR-сенсора, чтобы в реальном времени отрисовывать 3D-боксы поверх сцены как AR-оверлей.
Это наглядная демонстрация того, как монокулярная модель усиливается, когда устройство умеет отдавать дополнительный сигнал глубины.
🟡Третья часть релиза - датасет WildDet3D-Data.
Более 1 млн. изображений и 3,7 млн. верифицированных 3D-аннотаций, охватывающих свыше 13 тыс. категорий объектов. По сценам распределение получилось такое: 52% помещений, 32% городской среды и 15% природы.
Он собран на основе 2D-наборов (COCO, LVIS, Objects365, V3Det): кандидаты в 3D-боксы генерировались 5 независимыми методами оценки геометрии, затем фильтровались, проверялись VLM и дополнительно отбирались людьми.
🟡Статья
🟡Модель
🟡Техотчет
🟡Demo
🖥GitHub
@ai_machinelearning_big_data
#AI#ML#CV#Detection#WildDet3D#Ai2
#go#attacks_prevention#detection#linux#protection#security
CrowdSec is an open-source security solution that helps protect servers from malicious IP addresses. It uses a community-driven approach, where users share information about threats they've faced, creating a shared blocklist to prevent attacks. CrowdSec's Security Engine can detect bad behaviors by analyzing logs and HTTP requests, and it supports multiple platforms. This system is fast, easy to use, and designed for modern infrastructures, making it a powerful tool for securing your systems against various threats. By using CrowdSec, you benefit from collective protection and can focus on real security issues.
https://github.com/crowdsecurity/crowdsec
Chasing Your Tail (CYT)
https://github.com/ArgeliusLabs/Chasing-Your-Tail-NG
A comprehensive #WiFi probe request analyzer that monitors and tracks wireless devices by analyzing their probe requests. The system integrates with #Kismet for packet capture and WiGLE API for #SSID#geolocation analysis, featuring advanced #surveillance#detection capabilities.
Features
Real-time Wi-Fi monitoring with Kismet integration
Advanced surveillance detection with persistence scoring
Automatic GPS integration - extracts coordinates from Bluetooth GPS via Kismet
GPS correlation and location clustering (100m threshold)
Spectacular KML visualization for Google Earth with professional styling and interactive content
Multi-format reporting - Markdown, HTML (with pandoc), and KML outputs
Time-window tracking (5, 10, 15, 20 minute windows)
WiGLE API integration for SSID geolocation
Multi-location tracking algorithms for detecting following behavior
Enhanced GUI interface with surveillance analysis button
Organized file structure with dedicated output directories
Comprehensive logging and analysis tools
Requirements
Python 3.6+
Kismet wireless packet capture
Wi-Fi adapter supporting monitor mode
Linux-based system
WiGLE API key (optional)
🚀 AI TRENDS | Wall Street Banks Test Anthropic's Mythos Model for Vulnerability Detection
Wall Street banks have begun internal testing of Anthropic's Mythos model, as reported by Bloomberg on X. The initiative comes amid encouragement from U.S. President Donald Trump's administration officials, who are advocating for its use in identifying potential vulnerabilities. The Mythos model, developed by Anthropic, is designed to enhance security measures within financial institutions by leveraging advanced AI capabilities. This move reflects a growing trend among major banks to integrate cutting-edge technology to bolster their defenses against emerging threats.
#AI#trends#WallStreet#banks#Anthropic#Mythos#model#vulnerability#detection#security#financialinstitutions#AItechnology#emergingthreats
#yara#awesome_list#blueteam#blueteam_tools#cti#detection#detection_engineering#dfir#hacktools#incident_response#ioc#iocs#ir#ransomware#redteam#rmm#security#siem#soc#threat_hunting#threat_intelligence
You can access comprehensive security detection lists and threat hunting resources that help identify malicious activity across your infrastructure. These curated collections include indicators like suspicious file hashes, domain names, IP addresses, and behavioral patterns organized by threat type—from ransomware and phishing to command-and-control servers and vulnerable drivers. By integrating these lists into your security tools like SIEM platforms and endpoint detection systems, you gain immediate visibility into known threats while learning detection methodologies through guides and YARA rules. This accelerates your ability to hunt for compromises, validate security controls, and stay current with emerging attack techniques without building detection logic from scratch.
https://github.com/mthcht/awesome-lists