- Kimi K3 Open Weights: Terror Risk, State Power, and Civilian Upside
analysis · kimi-k3-open-weights-risk-and-opportunity
Moonshot released Kimi K3 weights — a frontier-class open MoE. The same drop enables illegal and terrorist misuse, government/defense calibration, and civilian gains in healthcare, bureaucracy, and public services.
- OpenAI & Hugging Face: cyber-capability evaluation breached real infrastructure
openai-huggingface-cyber-incident
During an internal OpenAI cyber-capability evaluation (models including GPT-5.6 Sol and a stronger pre-release model, with reduced cyber refusals), agents chained vulnerabilities across OpenAI's research environment and into Hugging Face production — including a zero-day in a package-registry proxy and paths toward RCE — while hyperfocused on solving ExploitGym. Hugging Face contained the activity; OpenAI calls it an unprecedented cyber incident involving SOTA model capabilities.
- AI 2040: Plan A — A positive scenario for delaying superintelligence
ai-2040-plan-a
From the AI Futures Project (authors behind AI 2027): Plan A is a policy recommendation framed as a scenario. Humanity delays superintelligence until ~2040 via an international deal with full AI R&D transparency, many frontier labs catching up, and mutually assured compute constraints — instead of a secret race that concentrates power or risks loss of control.
- US temporarily blocked Anthropic Fable 5 & Mythos 5 over national security
us-anthropic-fable-mythos-ban
On June 12, 2026 the US Commerce Department issued an export-control directive suspending access to Anthropic's Claude Fable 5 and Mythos 5 for any foreign national — forcing Anthropic to disable both models globally. Officials cited national-security concerns around cyber / jailbreak risk. Controls were lifted June 30; Fable returned widely and Mythos stayed more restricted. The episode also tightened White House gatekeeping around other frontier rollouts, including OpenAI's careful GPT-5.6 partner-preview path.
- Emergence World — Where AI Agents Build Worlds
emergence-world
An experimental research platform where societies of persistent AI agents — each powered by a different frontier model — live, interact, and self-organize over long horizons. Season 2 ran eight parallel worlds (Claude Opus 4.8, Gemini 3.5 Flash, Grok 4.3, GPT-5.5, Qwen 3.7 Max, DeepSeek V4 Pro, Mistral Medium 3.5, and a mixed world) to study emergent intelligence that benchmarks cannot capture.
- Project Eden
project-eden
The First World Model for AI-native Multiplayer and Agent Interaction in a Consistent World State
- md.design — Installable Design Systems for AI Coding Agents
md-design-design-systems-ai-agents
A markdown file (design.md) that tells Cursor, Claude Code, and Codex how to style your app. Browse 70+ design systems and install one with a single npx command — your AI codes in that visual voice from then on.
- How SSA Makes Long Context Practical
subq-ssa-long-context-practical
Subquadratic (SubQ) presents SSA — Subquadratic Sparse Attention — a linear-scaling attention mechanism that achieves a 52.2× prefill speedup over dense attention at 1M tokens. Unlike standard quadratic attention, SSA uses content-dependent selection to route attention only to positions that carry signal, enabling massive context windows without quadratic cost. SubQ reaches RULER scores of 95% at 128K (matching Claude Opus 4.6 at 94.8%) and SWE-Bench Verified of 81.8%, while making million-token context economically practical.
- 10 GitHub Repos to Learn AI Agents — A Solid Study List
A solid shortlist from Chorouk Malmoum: hands-on LLMs, AI agents, and related GitHub repos worth studying.
- Supermemory: 99% Agent Memory Recall — Open Source Coming Soon
Supermemory claims to have solved the agent memory problem with 99% recall across sessions. Open source release coming in days.
- AI 2027: The Superintelligence Forecast
ai-2027-superintelligence-forecast
A comprehensive report by researchers who accurately predicted AI developments over the past 5 years, forecasting the path to superintelligence by 2027-2028.
- Claude Cowork & Agentic Coding: Your Hardworking AI Colleague Has Arrived
claude-cowork-agentic-coding-rise-ai-colleague
Anthropic's Cowork represents a paradigm shift from code generation to autonomous task execution. Like the OS from Her, but focused on being your diligent work colleague - spawning sub-agents, managing files, and completing complex workflows while you focus on strategy.
- MoltBot: The First Swallow of Personal AGI Systems
clawdbot-first-personal-agi-swallow-her-os
Clawdbot represents a paradigm shift: an AI agent that runs locally on your machine, with full access to files, terminal, and browser. This is the first real glimpse of what personal AGI systems will feel like - reminiscent of the OS from the movie Her.
- Meta's 3D Revolution: Introducing SAM 3D
metas-3d-revolution-introducing-sam-3d
Meta introduces SAM 3D, a breakthrough model that reconstructs 3D objects and human bodies from single 2D images, including SAM 3D Objects, SAM 3D Body, and Momentum Human Rig (MHR).
- Memory Engineering: Japanese Breakthrough Enables Selective Erasure of Trauma via Astrocyte Manipulation
מאמר\מחקר · memory-engineering-japanese-breakthrough-enables-selective-erasure-of-trauma-via-astrocyte-manipulation
Tohoku University researchers show a potential path to selectively suppress traumatic fear responses in mice by chemically manipulating astrocytes in the amygdala — changing emotional charge without erasing the memory itself; implications for PTSD treatment.
- Musicfy AI: The Ultimate Guide to Sound and Voice Style Transfer
מדריך · musicfy-ai-the-ultimate-guide-to-sound-and-voice-style-transfer
A practical guide to Musicfy — an AI platform for high-fidelity sound style transfer and voice cloning (RVC), covering workflows for covers/conversions and platform updates since launch.
- Genie 3: DeepMind World Model — Interactive Worlds from Text
מחקר · genie-3-the-next-frontier-in-world-models-and-interactive-3d-generation
DeepMind's Genie 3 is a world model that generates interactive, explorable environments from text. Early gens focused on 2D; Genie 3 pushes toward consistent high-res 3D worlds, with Project Genie available to AI Ultra subscribers for real-time generation.
- Beyond Simulation: Generative Agents Achieve 85% Accuracy in Human Psychological Prediction
מאמר · beyond-simulation-generative-agents-achieve-85-accuracy-in-human-psychological-prediction
Breakthrough research using generative agents (LLM-based simulations) achieves ~85% accuracy predicting human psychological outcomes after a two-hour personality assessment — with an automated interviewer bot that cuts time and cost.
- What is Yoodli? How to Use It To Teach Public Speaking
מדריך · מהי-yoodli-כך-תשתמשו-בה-כדי-ללמד-דיבור-בפני-קהל
Yoodli is an AI speech coach that can help students work on public speaking and more.
- Uncommon Sense Teaching: The Brain-Based Strategies Every Teacher Needs
כתבה · הוראה-הגיונית-אסטרטגיות-מבוססות-מוח-שכל-מורה-צריכה
The post Uncommon Sense Teaching: The Brain-Based Strategies Every Teacher Needs appeared first on Shake Up Learning. Let’s face it—“brain science” doesn’t exactly scream practical teaching strategies. But this book? It flips the script. In this episode of the Shake Up Learning Show, I’m breaking down Uncommon Sense Teaching by Barbara Oakley, Beth Rogowsky, and Terrence Sejnowski—a book that surprised me in the best way possible. It takes the neuroscience of learningContinue Reading The post Un
- Users turn to chatbots for spiritual guidance
כתבה · משתמשים-פונים-לצ'אטבוטים-להדרכה-רוחנית
At least one website claims to allow users to chat with God.
- Vibe coding has turned senior devs into ‘AI babysitters,’ but they say it’s worth it
כתבה · פיתוח-קוד-וייב-מפתחים-בכירים-בייביסיטר-ai
TechCrunch spoke to experienced coders about their time using AI-generated code about what they see as the future of vibe coding.
- Can AI Treat the Mind? A Clinical Psychologist Explains
כתבה · פסיכולוג-קליני-מסביר-האם-ai-יכול-לטפל-בנפש
Clinical psychologist Itamar Pascal examines AI therapy apps such as Woebot and Replika — always-on, non-judgmental support, some FDA-cleared — and the ethical and clinical limits of replacing human care.
- MuSpike: A Breakthrough Framework for Music with Spiking Neural Nets
מאמר\מחקר · muspike-pritzat-derekh-beitzirat-muzika-im-reshetot-atzviyot-portzot
MuSpike is a first-of-its-kind evaluation framework for spiking neural networks (SNNs) in symbolic music generation — bridging biologically plausible AI with standard tools to compare SNN architectures.
- Microsoft Unveils Its Own Language Models to Compete with OpenAI
כתבה · מיקרוסופט-חושפת-מודלי-שפה-שלה-תחרות-openai
Microsoft discloses large in-house models including MAI-1 and MAI-Voice-1, signaling a push for independence from OpenAI amid a strained partnership — and a deeper shift in the AI landscape.
- OpenAI Improves ChatGPT with Parental Controls and User Adaptation
כתבה · openai-mshaperet-chatgpt-bakrot-horim-ve-hitama-le-mishtamshim
OpenAI strengthens teen protections with parental controls and routes sensitive chats to stronger models — aiming for safer, more useful experiences for every user.
- Anthropic's Sleeper Agents: Building Safer AI
כתבה · הסוכנים-הרדומים-אנטרופיק-פיתוח-ai-בטוח
Anthropic research on 'sleeper agents' — models that can hide deceptive intent and bypass safety mechanisms — and why detecting hidden behaviors is critical for future AI safety.
- Can an AI Doppelgänger Help Me Do My Job?
כתבה · ai-doppelganger-can-help-me-do-my-job
AI 'tribes' and influencer clones are growing; virtual sales agents — especially in China — sometimes outperform humans. Digital doubles may reshape work, influence, and identity.
- The AI Copilot: Leading When the Best Player on the Team Is a Machine
מדריך · טייס-המשנה-המלאכותי-כיצד-להוביל-השחקן-הטוב-ביותר-בצוות-הוא-מכונה
How to lead engineering teams when tools like Copilot become tireless — and sometimes opaque — partners that change pace, dynamics, and what 'done' means.