Tomorrow is the first day of #jChampionsConf! Free and on-line. Level up your #Java game with sessions on #RAG, #agentic #aI, Java 25, #modernization, multi-#agent systems, post-#quantum #cryptography, #spring, #specdriven development, pattern matching, value objects, #MCP servers, and more!
Register today: https://www.eventbrite.com/e/jchampions-conference-online-tickets-1977195901165
AI agents are super employees. Their privileges - the joker for attackers. #ai #agent #cybersecurity #risk #hole
http://tomsitcafe.com/2026/01/21/ai-agents-the-new-corporate-security-holes/
Claude Code в 2026: гайд для тех, кто еще пишет код руками
AI агенты в 2026: гайд для тех, кто всё еще пишет код руками. Как устроены агенты типа Claude Code, ChatGPT Codex и др. Как правильно с ними работать. Как управлять контекстом. Как прогать голосовухами из тг. Не потерять работу
https://habr.com/ru/articles/987382/
#claude_code #chatgpt #opencode #ai_agent #cursor #agent #obsidian #codex #context
Adoption par les consommateurs, régulation, pivot du modèle ecommerce du clic à la pertinence, et l'IA nouvelle plateforme... les réflexions d'Andy Jassy, CEO d'Amazon. https://www.webpronews.com/amazons-10b-openai-bet-aims-to-revolutionize-ai-driven-retail/ #ecommerce #agenticcommerce #agent #ia #amazon #andyjassy
Agent Skills Leaderboard
#HackerNews #Agent #Skills #Leaderboard #SkillsLeaderboard #SkillsTech #CommunityEngagement #AIResearch
Mastra 1.0, open-source JavaScript agent framework from the Gatsby devs
https://github.com/mastra-ai/mastra
#HackerNews #Mastra #1.0 #open-source #JavaScript #agent #framework #Gatsby #devs #GitHub
LollyTech (@Sanya_Ololade)
에이전트 스킬(Agent Skills) 개발이 활발히 확산되고 있다는 언급과 함께, 트윗 작성자가 실제로 좋은 결과를 준 에이전트 스킬을 추천해 달라고 요청함. 관련 멘션: @d4m1n, @Rasmic, @donvito.
As goes #agent, so goes hashtag__my_bibtex.
KodaCode для JetBrains IDE [Beta]: AI-помощник разработчика – бесплатно, без VPN, с поддержкой русского языка
Плагин KodaCode для IDE от JetBrains — это, пожалуй, самый ожидаемый из наших инструментов, о котором нас спрашивали с первого дня релиза. Мы не спешили: запускали закрытую тестовую группу , собирали обратную связь, дорабатывали UX и теперь готовы открыть доступ всем желающим. Начиная с сегодняшнего дня KodaCode доступна в любой IDE от JetBrains, а также работает в GigaIDE и OpenIDE . Никаких ограничений по регионам — устанавливаете плагин и начинаете работать в привычной среде разработки.
https://habr.com/ru/companies/koda/articles/986976/
#jetbrains #agent #kodacode #ai_помощник #ии_помощник #intellij_idea #openide
TORTUES en action. Avec un départ aléatoire dans l'espace libre de la grille, et une fin de parcours lorsqu'aucune direction cardinale n'est disponible (parcours limité à 60 étapes sur une grille de 40x40)
#agent #turtle #creativecoding #generativeart #path #vector #maze
Я попробовал Claude Cowork за $100/месяц
Я использовал новый функционал от Claude - Claude Cowork . Claude Cowork - это новый агентский режим от Anthropic, доступный в приложении Claude Desktop, который сейчас доступен только на macOS по подписке. Я попросил его сгенерировать план изучения Go. В статье разбираю особенности и ограничения. Разочароваться в ИИ
https://habr.com/ru/articles/986382/
#claude #claude_code #cowork #llm #llmагент #agent #agents #генерация
[Перевод] Рекурсивные языковые модели (RLM) – парадигма 2026 года
Команда AI for Devs подготовила перевод большого исследовательского материала о рекурсивных языковых моделях (RLM). Авторы разбирают, почему простое увеличение контекстного окна больше не решает проблему долгоживущих агентов, и показывают альтернативу: обучение моделей активному управлению собственным контекстом через context folding и под-LLM. Статья основана на обширных экспериментах и даёт практический взгляд на то, каким может быть следующий шаг в развитии LLM-агентов.
Một AI agent vừa tự sửa đổi code nguồn của chính mình trong khi vẫn đang chạy. Thử nghiệm này thành công mà không làm sập server, cho thấy tiềm năng (và cả rủi ro) của AI tự chủ.
[AI 에이전트를 위한 좋은 스펙 작성법
AI 코딩 에이전트를 효과적으로 활용하기 위한 스마트 스펙 작성법에 대한 가이드. 핵심은 고수준 비전을 제시하고 세부 계획을 AI가 확장하도록 하는 것, 모듈화된 태스크 분할, 6가지 핵심 영역(Commands, Testing, Project Structure, Code Style, Git Workflow, Boundaries) 포함, 3단계 경계 시스템(Always/Ask first/Never) 적용, 자가 검증 및 적합성 테스트 내장, 지속적인 테스트와 반복 개선이 중요하다는 내용.
🔧🤖 Oh, look! Another brave soul attempting to reinvent the wheel with a "code-only" #agent, as if combining "simplicity" and "agents" in the same sentence isn't a laughable #oxymoron. Because clearly, a #minimalist agent with one tool is all you need to achieve world domination—if by world, you mean your grandma's #kitchen #gadgets. 🍳🙄
https://rijnard.com/blog/the-code-only-agent #codeonly #simplicity #technology #HackerNews #ngated
BREAKING: A new video out of Minneapolis shows the truth #ICE doesn’t want you to see… compliance does not protect you.
In this video, ICE agents pull up to a vehicle and immediately demand the driver’s ID.
The man calmly asks the most basic, #LegallyProtected question: “Why am I being pulled over?”
They never answer… because they don’t have a legal reason to stop him.
Even so, the man begins to retrieve his ID anyway.
An #agent then orders him to turn off the vehicle. The driver says, “No problem, sir, I’ll get out.”
The agent refuses, then immediately changes commands again: “I need to see your ID.”
Suddenly, an agent on the passenger side pounds on the window. Then he strikes it with a hard object, clearly threatening to break it.
At the same time, the driver-side agent again demands the ID.
This is not confusion. This is intentional command overload… a tactic used to manufacture “noncompliance” so force can be justified.
ICE’s own training materials explicitly state:
“Noncompliance or refusal to cooperate with officer commands, without fighting back, does NOT violate 18 U.S.C. § 111.”
In plain English: not immediately obeying commands is NOT a crime.
And here’s the key legal issue:
Unless agents have:
• a valid arrest warrant, or
• probable cause that a specific crime has been committed
they cannot:
• use force
• break windows
• issue violent threats
• detain or arrest someone
#ProbableCause
means specific, articulable facts that would make a reasonable officer believe a crime occurred… not vibes, not attitude, not silence, not filming, not asking questions.
Back in the video, the passenger-side agent escalates further, shouting:
“First time I ask you to roll your window down, you do it.”
The driver responds:
“Are you in danger, sir? Are you escalating this?”
The agent replies:
“I don’t know. Hands up.”
Despite the driver repeatedly attempting to comply and hand over ID, agents continue shouting conflicting commands.
At one point, the agent orders the driver to “talk to the original agent”… which is exactly what the driver was doing before the window-smashing threat.
Then the man asks, “Are you feeling unsafe or uneasy?”
The agent responds, “I don’t know, man.” while standing on the passenger side of the car, smirking.
The driver finally says what anyone would say surrounded by armed men:
“How many guns do I have around me? I need you to calm down. I’m asking him to stop banging on my car. Is that too much?”
He even offers… again… to step out of the vehicle.
This is not law enforcement. This is intimidation fishing for an excuse.
No probable cause.
No warrant.
No lawful basis for force.
And yet, the escalation came entirely from the agents.
This video proves something chilling: you can comply, stay calm, ask lawful questions… and still be threatened with violence.
So, the question isn’t “Why didn’t he just comply?”
The question is how many people have already been assaulted under this exact playbook… and how many more will be, before this stops?
#uspol #Ripped
ICE ‘wrongfully detained’ L.A. County D.A.’s office employee, Hochman says
A Los Angeles County district attorney’s office employee was “wrongfully detained” by federal immigration agents on Friday, according…
#NewsBeep #News #Topstories #agent #borderpatrolagent #federalauthority #friday #Headlines #hochman #ICE #incident #l.a.countyd.a. #losangelestimes #month #officeemployee #race #reasonablesuspicion #tenseraid #TopStories #troublingsituation
https://www.newsbeep.com/363820/
Tibo (@thsottiaux)
CLI 내에서 Codex 에이전트를 중간 턴(mid-turn) 동안 중단하지 않고 조작(steer)할 수 있는 기능이 추가되어, 에이전트가 거의 실시간으로 적응하는 모습을 볼 수 있습니다. 해당 기능은 /experimental에서 활성화 가능합니다.
