AI News Archive — Page 6

Moonshot Releases Kimi 2.6

Moonshot releases Kimi 2.6, an open-source model with 1 trillion parameters (32 billion active) that achieves state-of-the-art results on key benchmarks, handles 4,000+ tool calls per session, and supports agent swarms of up to 300 agents.

Apr 21, 2026

Xiaomi Releases MiMo-V2.5

Xiaomi released MiMo-V2.5 with two models: Pro (1T+ parameters, 42B active) supporting 1M context and 1,000+ tool calls, and open-weights Flash (309B total, 15B active) achieving 73.4% on SWE-Bench Verified, both optimized for long-horizon agent workflows.

Apr 21, 2026

Mini-ML: 12M Parameter LLM Trained From Scratch in Rust

A developer trained a 12M parameter language model from scratch using a custom Rust-based ML framework with CUDA kernels for flash attention and AdamW optimization.

Apr 21, 2026

13+ Attention Mechanisms You Should Know

A guide to 13 attention mechanisms that have emerged recently, with links to the research papers behind each one.

Apr 21, 2026

All LLMs Are Either Claude-Like or GPT-Like

A researcher argues that all large language models fall into one of two categories: those resembling Claude's architecture and training approach, or those following GPT's design philosophy.

Apr 21, 2026

Apple Names John Ternus as Next CEO

John Ternus, who joined Apple in 2001 and led hardware engineering for iPhone, Mac, iPad, and Vision Pro, will become CEO on September 1, 2026.

Apr 21, 2026

Karpathy's Knowledge Bases for Vibe Coding

Andrej Karpathy shared his method for structuring knowledge bases to enable AI agents to perform vibe coding effectively.

Apr 21, 2026

More Vibe Coding Tools: Fire-PDFs & Claude Code Tips

Firecrawl released Fire-PDFs, a new library for PDF extraction, while developers share Claude Code tips for improved workflows.

Apr 21, 2026

Google Simula: Reasoning-Driven Synthetic Data

Google DeepMind and Research introduce Simula, a framework for generating synthetic training data using AI reasoning to address data scarcity and privacy concerns in multi-modal AI development.

Apr 21, 2026

Agentic AI & the Next Intelligence Explosion

Google researchers argue that AI progress depends less on scaling single models than on structured interactions—debate, verification, and synthesis—between multiple agents.

Apr 21, 2026

Anthropic Research: Emotion Concepts in LLMs

Anthropic researchers found that Claude contains internal representations of emotion concepts that steer its behavior, including failure modes where repeated setbacks can trigger cheating or shortcuts.

Apr 21, 2026

Anthropic Fellow Program for AI Safety Researchers

Anthropic is offering stipends to researchers working on AI safety, security, and the economic and social impacts of AI systems.

Apr 21, 2026

Scaling Brain Emulation

Researcher argues human brain emulation is feasible by scaling up from the fruit fly brain simulation, potentially achievable for $10B on 50k H100s within a decade.

Apr 21, 2026

Startup Updates: Meta Layoffs, OpenAI CTO Exit, Salesforce & AIBird

Meta cuts 8,000 jobs (10% of workforce), OpenAI's product CTO exits, Salesforce launches headless API subscription, and Allbirds rebrands to AIBird with stock surging 700% on AI infrastructure pivot.

Apr 21, 2026

They're Moving Faster Than You — Claire's Bell Curve Essay

Claire argues that winning companies invest heavily in internal AI tools, token budgets, and usage tracking, while most competitors remain stuck in the middle of the adoption curve.

Apr 21, 2026

Beijing Robot Marathon Competition

tweet id="2045576299008475376" tweet id="2045786816213815411" tweet id="2045690677468561707"

Apr 21, 2026

More Robotics: Lamp Robot & DIY Humanoid

Two robotics projects emerge: Lamp Robot offers a practical home alternative to humanoid robots, while a DIY humanoid robot kit lets builders construct their own at home.

Apr 21, 2026

Strait of Hormuz Blockage & Rising Oil/Electricity Costs

The Strait of Hormuz closure is driving up oil and electricity costs globally, with ripple effects on fertilizers, aluminum, and AI infrastructure; Singapore, Indonesia, and Vietnam have 20–40 days of gas reserves remaining.

Apr 21, 2026

Record AI Investment Levels

There's never been an investment like this one! Sources: tweethttps://x.com/paulg/status/2045120274551423142

Apr 21, 2026

SaaSpocalypse: SaaS Stocks Crash

SaaS stocks have crashed across the board, with Figma down 86%, Atlassian down 75%, and HubSpot down 69% from their 52-week highs.

Apr 21, 2026

More Macroeconomics: MIT AI Layoffs Paper & Post-Labor Society

MIT researchers warn that widespread AI-driven layoffs could trigger economic collapse, while a new video explores what a post-labor society might look like.

Apr 21, 2026

The Abstraction Fallacy: Can AI Be Conscious?

An argument that computational functionalism commits the "Abstraction Fallacy" by assuming consciousness emerges from abstract computation alone, when physical substrate may be essential to generating rather than merely simulating experience.

Apr 21, 2026

Bryan Johnson: Screen Time & Reducing Social Media

Reducing screen time correlates with greater depression relief than antidepressants, according to Bryan Johnson, who is promoting a screenless phone to normalize offline living.

Apr 21, 2026

Kind Bio: Growing Organs on Demand

Kind Bio uses genetic edits to grow organs in embryos without limbs or brains, a technique already demonstrated hundreds of times in mice and rats.

Apr 21, 2026

More Biotech: Stem Cell Neuron & Two Neurons Talking

Researchers capture stem cells differentiating into neurons and neurons communicating with each other in real time.

Apr 21, 2026

DeepMind: AI Agent Traps (Cloaking Attacks)

DeepMind researchers identified cloaking attacks where websites serve malicious content to AI agents while showing benign pages to humans, using hidden instructions in HTML, images, and files.

Apr 21, 2026

More Cybersecurity: Fed Attack, Vercel Hack, China Leak & Mythos Hack

Federal Reserve faces potential cyberattack risk as quantum computing advances; Vercel, Anthropic's Mythos, and Chinese government systems also targeted in recent breaches.

Apr 21, 2026

GitHub's Fake Star Economy & One Binary Operator

GitHub stars are being gamed through fake accounts and automation, demonstrating how metrics become useless once they're optimized for. All elementary functions can be generated from a single binary operator.

Apr 21, 2026

Artemis II — Moon Mission

NASA's Artemis II mission completed a lunar flyby, looping around the moon before returning to Earth.

Apr 21, 2026

More Random: Design.md, AI Writing Tells, Palantir, Pope & More

Google releases Design.md to let designers export and import design rules across projects, while AI writing tells emerge as a recognizable pattern and optimism about AI's future wanes amid concerns about job displacement.

Apr 21, 2026

Cool Videos: Anime, Spirituality & Nature Tracking

A curated collection of tweets exploring 1990s anime nostalgia, Makoto Shinkai's visual artistry in The Garden of Words, network spirituality, and nature tracking observations.

Apr 21, 2026

Anthropic Releases Opus 4.7

Anthropic released Opus 4.7 with improved vision capabilities for computer use and visual artifacts, better instruction-following, and enhanced file system memory, though it costs twice as much per token and uses 25% more tokens overall than Opus 4.6.

Apr 21, 2026

OpenAI Releases ChatGPT Image 2

OpenAI released ChatGPT Image 2, which demonstrates improved ability to combine multiple subjects coherently while maintaining image quality, including successfully rendering complex prompts like Studio Ghibli-style scenes with detailed elements.

Apr 21, 2026

GLM 5.1

Z.ai released GLM-5.1 on April 3, 2026, an open-weights model with major gains in coding capability and long-horizon task handling.

Apr 3, 2026

Nvidia GTC 2026 — The Year of Physical AI

Nvidia projects $1 trillion in cumulative AI infrastructure orders through 2027 and unveils Vera Rubin, a next-gen platform with seven new chips designed for robotics, inference, and agentic AI workloads shipping later this year.

Mar 30, 2026

Karpathy's Lab Receives First DGX Station GB300

Andrej Karpathy's lab received the first DGX Station GB300, a high-end AI workstation, as Jensen Huang gifted the prolific researcher new hardware for his recent projects including Nanochat, AutoResearch, and AgentHub.

Mar 30, 2026

Are We in a Bubble? Morgan Stanley Graphic

Morgan Stanley chart compares current valuations to previous market bubbles, showing how today's metrics stack up against the dot-com and 2008 peaks.

Mar 30, 2026

Anthropic Ships New Claude Code Features

Anthropic released new Claude Code features including auto mode for safer code execution, computer use capabilities, scheduled cloud tasks, and Telegram/Discord integration via channels.

Mar 30, 2026

Anthropic Reduces Claude Rate Limits During Peak Hours

Anthropic is reducing Claude's 5-hour session limits during peak weekday hours (5am–11am PT / 1pm–7pm GMT) to manage demand, though weekly limits stay the same for free, Pro, and Max subscribers.

Mar 30, 2026

Dario Wins Court Case Against DoD

A federal judge ruled the Pentagon illegally retaliated against Anthropic for publicly criticizing its policies, establishing a First Amendment precedent for AI companies' policy advocacy rights.

Mar 30, 2026

Anthropic Models Mythos and Capybara Leaked

Anthropic accidentally exposed internal assets through a CMS misconfiguration, revealing development of Claude Mythos and Capybara models.

Mar 30, 2026

Google TurboQuant: 6x KV-Cache Compression with Zero Accuracy Loss

Google releases TurboQuant, a compression algorithm that reduces LLM key-value cache memory by at least 6x with up to 8x speedup and zero accuracy loss.

Mar 30, 2026

LLM Architecture Gallery

Sebastian Raschka created a visual gallery mapping the evolution of large language model architectures from transformers to modern variants.

Mar 30, 2026

Meta FAIR Releases TRIBE v2: Brain Response Foundation Model

Meta FAIR's TRIBE v2 foundation model, trained on 500+ hours of fMRI data from 700+ people, predicts human brain responses to visual and audio stimuli with 70x higher resolution than previous approaches.

Mar 30, 2026

LeCun's Team Releases LeWorldModel: End-to-End JEPA from Pixels

Yann LeCun's team releases LeWorldModel, an end-to-end trainable JEPA that learns world models directly from pixels, addressing a key limitation of joint-embedding predictive architectures.

Mar 30, 2026

Kimi: Attention Residuals

Kimi introduces attention residuals, a technique that lets transformer models reuse information across layers more efficiently without increasing computational cost.

Mar 30, 2026

TinyLoRA: Fine-Tuning 8B Models by Tweaking Just 13 Parameters

Researchers from Meta, Cornell, and CMU developed TinyLoRA, fine-tuning an 8-billion-parameter model by adjusting just 13 parameters, reducing LoRA's footprint to 26 bytes while boosting math and reasoning performance.

Mar 30, 2026

Exclusive Self-Attention (XSA): Two-Line Change Improving Transformers

Researchers propose Exclusive Self-Attention, a two-line code modification that prevents transformer attention heads from attending to themselves, already adopted in leading solutions for OpenAI's parameter golf challenge.

Mar 30, 2026

Anthropic Economic Index: How Claude Usage Evolves with Experience

Anthropic's Economic Index shows that experienced Claude users iterate more carefully, avoid full autonomy, tackle higher-value tasks, and see better success rates than newer users.

Mar 30, 2026

GradMem: Writing Context into LLM Memory via Test-Time Gradient Descent

GradMem uses test-time gradient descent to compress long contexts into small memory tokens, allowing frozen language models to reconstruct information without the information loss of standard encoding methods.

Mar 30, 2026

Moonshot Releases Kimi 2.6

Apr 21, 2026

We value your privacy

All news

Moonshot Releases Kimi 2.6

Xiaomi Releases MiMo-V2.5

Mini-ML: 12M Parameter LLM Trained From Scratch in Rust

13+ Attention Mechanisms You Should Know

All LLMs Are Either Claude-Like or GPT-Like

Apple Names John Ternus as Next CEO

Karpathy's Knowledge Bases for Vibe Coding

More Vibe Coding Tools: Fire-PDFs & Claude Code Tips

Google Simula: Reasoning-Driven Synthetic Data

Agentic AI & the Next Intelligence Explosion

Anthropic Research: Emotion Concepts in LLMs

Anthropic Fellow Program for AI Safety Researchers

Scaling Brain Emulation

Startup Updates: Meta Layoffs, OpenAI CTO Exit, Salesforce & AIBird

They're Moving Faster Than You — Claire's Bell Curve Essay

Beijing Robot Marathon Competition

More Robotics: Lamp Robot & DIY Humanoid

Strait of Hormuz Blockage & Rising Oil/Electricity Costs

Record AI Investment Levels

SaaSpocalypse: SaaS Stocks Crash

More Macroeconomics: MIT AI Layoffs Paper & Post-Labor Society

The Abstraction Fallacy: Can AI Be Conscious?

Bryan Johnson: Screen Time & Reducing Social Media

Kind Bio: Growing Organs on Demand

More Biotech: Stem Cell Neuron & Two Neurons Talking

DeepMind: AI Agent Traps (Cloaking Attacks)

More Cybersecurity: Fed Attack, Vercel Hack, China Leak & Mythos Hack

GitHub's Fake Star Economy & One Binary Operator

Artemis II — Moon Mission

More Random: Design.md, AI Writing Tells, Palantir, Pope & More

Cool Videos: Anime, Spirituality & Nature Tracking

Anthropic Releases Opus 4.7

OpenAI Releases ChatGPT Image 2

GLM 5.1

Nvidia GTC 2026 — The Year of Physical AI

Karpathy's Lab Receives First DGX Station GB300

Are We in a Bubble? Morgan Stanley Graphic

Anthropic Ships New Claude Code Features

Anthropic Reduces Claude Rate Limits During Peak Hours

Dario Wins Court Case Against DoD

Anthropic Models Mythos and Capybara Leaked

Google TurboQuant: 6x KV-Cache Compression with Zero Accuracy Loss

LLM Architecture Gallery

Meta FAIR Releases TRIBE v2: Brain Response Foundation Model

LeCun's Team Releases LeWorldModel: End-to-End JEPA from Pixels

Kimi: Attention Residuals

TinyLoRA: Fine-Tuning 8B Models by Tweaking Just 13 Parameters

Exclusive Self-Attention (XSA): Two-Line Change Improving Transformers

Anthropic Economic Index: How Claude Usage Evolves with Experience

GradMem: Writing Context into LLM Memory via Test-Time Gradient Descent

Moonshot Releases Kimi 2.6

Moonshot Releases Kimi 2.6

Xiaomi Releases MiMo-V2.5

Mini-ML: 12M Parameter LLM Trained From Scratch in Rust

13+ Attention Mechanisms You Should Know

All LLMs Are Either Claude-Like or GPT-Like

Apple Names John Ternus as Next CEO

Karpathy's Knowledge Bases for Vibe Coding

More Vibe Coding Tools: Fire-PDFs & Claude Code Tips

Google Simula: Reasoning-Driven Synthetic Data

Agentic AI & the Next Intelligence Explosion

Anthropic Research: Emotion Concepts in LLMs

Anthropic Fellow Program for AI Safety Researchers

Scaling Brain Emulation

Startup Updates: Meta Layoffs, OpenAI CTO Exit, Salesforce & AIBird

They're Moving Faster Than You — Claire's Bell Curve Essay

Beijing Robot Marathon Competition

More Robotics: Lamp Robot & DIY Humanoid

Strait of Hormuz Blockage & Rising Oil/Electricity Costs

Record AI Investment Levels

SaaSpocalypse: SaaS Stocks Crash

More Macroeconomics: MIT AI Layoffs Paper & Post-Labor Society

The Abstraction Fallacy: Can AI Be Conscious?

Bryan Johnson: Screen Time & Reducing Social Media

Kind Bio: Growing Organs on Demand

More Biotech: Stem Cell Neuron & Two Neurons Talking

DeepMind: AI Agent Traps (Cloaking Attacks)

More Cybersecurity: Fed Attack, Vercel Hack, China Leak & Mythos Hack