AI News Archive — Page 8

Kimi K2.5 (Open Source, ~1T params, Swarms)

Open-source Kimi K2.5 (~1T parameters) beats Claude Opus 4.5 on multiple benchmarks and uses agent swarms to cut execution time by up to 4.5×, signaling open models are closing the gap with private alternatives.

Feb 6, 2026

Apple — Think Different on AI Capex

Apple is skipping the $100B+ data center arms race, instead deploying its $150B cash pile on inference through Google's Gemini while rivals race to build their own AI infrastructure.

Feb 6, 2026

Genomic Models: Bytedance SeedFold and Google AlphaGenome

ByteDance's SeedFold outperforms AlphaFold 3 in protein structure prediction, while Google releases AlphaGenome weights to academic researchers.

Feb 6, 2026

World Models Are Taking Over

Google's Project Genie, World Labs' World API, and DeepMind's D4RT shift AI from generating videos to building persistent simulated worlds where agents can reason and act.

Feb 6, 2026

Talaas ASIC Chip: Weights Printed on Silicon

Talaas released an ASIC chip with model weights printed directly on silicon, achieving 17,000 tokens/sec on Llama 3.1 8B—45x faster than NVIDIA B200 and 8x faster than Cerebras—at $0.75 per million tokens.

Feb 6, 2026

ByteDance Seedance 2.0 — The Movie Singularity Is Here

ByteDance released Seedance 2.0, a video generation model that can create realistic human performances and complex scenes, prompting Disney to send a cease-and-desist letter.

Feb 6, 2026

Recursive Language Models (RLM)

Recursive Language Models use mutable context and recursive sub-agents with REPL access to overcome LLM reasoning limits, with Google's Agent Development Kit now enabling production deployment at scale.

Feb 6, 2026

Prompt Repetition Improves Non-Reasoning LLMs

Repeating a prompt twice improves accuracy across Gemini, ChatGPT, Claude, and DeepSeek on seven benchmarks without fine-tuning, extra training, or added output length.

Feb 6, 2026

Patterning: The Duality Of Interpretability

Researchers introduce "patterning," a mechanistic interpretability approach that reverses the usual question: instead of inferring what data produces a model's internal structure, they determine what data produces desired structure, demonstrated by…

Feb 6, 2026

Anthropic: How Misalignment Scales with Bigger Models

Anthropic research finds that larger AI models don't necessarily become more coherent on hard tasks, with failures resembling unpredictable "hot messes" rather than systematic misalignment.

Feb 6, 2026

Top Papers of the Month from DairAI

DairAI curated the top AI research papers from the past three weeks, though no standout breakthroughs emerged this month.

Feb 6, 2026

Does AI Already Have Human-Level Intelligence? Yes!

Researchers argue that modern large language models have already achieved human-level general intelligence, citing strong performance on exams, math contests, and scientific reasoning tasks.

Feb 6, 2026

Thoughts on AI Co-Creation and Understanding

Geoffrey Hinton argues that large language models genuinely understand what they say, similar to human comprehension, while others propose that AI interactions create a co-created third entity between user and model.

Feb 6, 2026

Recreating MIME Files from Redacted PDFs

Researcher recovers redacted PDF attachments from poorly censored email screenshots by using OCR and a CNN to distinguish lookalike characters in base64-encoded MIME data.

Feb 6, 2026

Face Recognition in a Few Lines of Python

An engineer demonstrated face recognition in just a few lines of Python using DeepFace and cosine distance to compare facial embeddings across images.

Feb 6, 2026

The Vibes Are Changing: AI Labs and Military Guardrails

Jeff Dean joins tech leaders opposing Pentagon pressure to remove military guardrails from AI systems, while OpenAI and Google comply and Anthropic faces threats for refusing autonomous weapons access.

Feb 6, 2026

Own Your GPU And Inference

Running AI inference on your own hardware with open-source models keeps your data private and your thinking your own, not logged by third parties or subject to corporate or state control.

Feb 6, 2026

Qwen3 Coder Next

Alibaba's Qwen released Qwen3 Coder Next, an open-weight language model designed for coding agents and local development workflows.

Jan 30, 2026

GLM 4.7 Flash

Zhipu AI released GLM-4.7-Flash, a 30B open-weights model positioned as a performance-efficiency balance in the SOTA class.

Jan 19, 2026

New AI Socratic chapter launches in Madrid

Jorge Ordovás and Iago Suárez, PhD are launching a new AI Socratic chapter in Madrid, with the first event scheduled for January 8th.

Jan 5, 2026

Models Leaderboard

AI Socratic's leaderboard ranks large language models by performance across benchmarks and real-world tasks.

Jan 5, 2026

We're In The "Vibe" Code Age — Claude Code reaches 100% of coding

A developer reports reaching 100% AI-generated code using Claude Code, as Anthropic's tools gain traction over competitors like Cursor among engineers seeking simpler interfaces.

Jan 5, 2026

Second Order Effects — Software Companies Are Cooked

Stack Overflow's monthly questions collapsed from 300k to near zero since ChatGPT launched, while Tailwind cut 75% of staff citing AI's brutal impact on business.

Jan 5, 2026

First Order Effects — SaaS Companies Are Cooked

. And First Order Effects — SaaS Companies Are Cooked..

Jan 5, 2026

More interesting updates around Claude Code

Nader Dabit releases Claude Agent SDK tutorial; community shares tools like iMCP for connecting digital life to AI, plus tips on automating tasks and building with Claude Code.

Jan 5, 2026

Cursor + GPT 5-2 built a browser (3M lines of code)

Cursor's concurrent agents built a 3 million-line browser over one week, with GPT 5-2 outperforming Opus 4.5 on long-running tasks.

Jan 5, 2026

Deep Fakes Will Change Social Media

Deepfake video creation will become a mainstream skill in 2026, separating creators who master the technology from those who merely consume it.

Jan 5, 2026

DeepSeek mHC: Manifold-Constrained Hyper-Connections

DeepSeek's mHC constrains residual connection mixing matrices to the Birkhoff polytope using Sinkhorn-Knopp normalization, enabling stable 4-8x wider residual streams and improved performance on reasoning benchmarks up to 27B parameters.

Jan 5, 2026

Neural networks at scale converge to a shared model of reality

Neural networks trained on different tasks and data converge to shared statistical models of reality, suggesting sufficiently strong learners rediscover natural coordinates in the world rather than inventing arbitrary representations.

Jan 5, 2026

The Ridiculous Engineering Of The World's Most Important Machine

Veritasium explores the extreme engineering behind the semiconductor manufacturing equipment that produces the world's most advanced computer chips.

Jan 5, 2026

Stanford CME295 Lecture on Transformers & LLMs, plus more videos and pods

Stanford's CME295 course offers a lecture series on transformers and large language models, plus curated videos and podcasts from December and January featuring discussions on AI fundamentals, neuroscience approaches to AI, and reinforcement learning.

Jan 5, 2026

Nvidia launches Vera Rubin chip

Nvidia launched the Vera Rubin chip, claiming 10x lower inference costs, 4x fewer GPUs needed for MoE training, and 5x better energy efficiency compared to Blackwell.

Jan 5, 2026

Nvidia $20B Groq acquisition

Nvidia acquires Groq's intellectual property and leadership team for $20B while leaving GroqCloud operating independently, a structure that avoids antitrust scrutiny by keeping the revenue-generating business separate.

Jan 5, 2026

Global AI compute capacity tops 15 million H100-equivalent

Global AI compute capacity has reached 15 million H100-equivalent units, marking a milestone in the infrastructure race to power advanced AI systems.

Jan 5, 2026

Meta buys Manus at $2-4B

Meta acquired Manus, an AI startup that reached $100M ARR in nine months, for $2–4B, validating consumer agent products as a scalable revenue category and bringing the team's production-grade agent technology in-house.

Jan 5, 2026

Geopolitics: US capture of Venezuela's Maduro and the Monroe Doctrine

The US capture of Venezuelan president Maduro signals a return to Monroe Doctrine expansionism in the Western Hemisphere, according to geopolitical analyst Peter Zeihan.

Jan 5, 2026

TSMC, the semiconductor chokepoint, and China's EUV prototype

TSMC produces 90% of the world's advanced semiconductors, making a Taiwan invasion an existential economic threat, but China has built a prototype EUV lithography machine in Shenzhen that could eventually break this dependency.

Jan 5, 2026

Kimi K2.6

Moonshot AI released Kimi K2.6, an open-weights multimodal model designed for long-horizon coding, UI/UX generation, and multi-agent orchestration.

Jan 1, 2026

Recursive Language Models: LLMs that call themselves to handle arbitrarily long prompts

Researchers introduce Recursive Language Models, which let LLMs decompose and recursively call themselves to process inputs up to 100 times longer than their native context windows, outperforming GPT-5 by 26% on long-context tasks.

Dec 31, 2025

Sakana AI — Continuous Thought Machines (CTM)

Sakana AI introduces Continuous Thought Machines, an AI model that synchronizes neuron activity timing to enable step-by-step reasoning similar to biological neural networks.

Dec 17, 2025

Upcoming Events: AI Aperitivo 2.0 (Milan) & AI Dinner 16.0 (NYC)

AI Aperitivo 2.0 in Milan on December 16 and AI Dinner 16.0 in New York on December 17 bring together engineers, researchers, and founders for Socratic dialogues on AI.

Dec 17, 2025

Model Wars: GPT-5.2 vs Opus 4.5 vs Gemini 3 vs Grok 4.1

OpenAI's GPT-5.2, Google's Gemini 3, Anthropic's Opus 4.5, and xAI's Grok 4.1 compete on reasoning and efficiency, with GPT-5.2 Thinking reaching 52.9% on ARC-AGI-2 and Gemini 3 reportedly building websites in single attempts.

Dec 17, 2025

US AI startups increasingly built on Chinese open-source foundations

Chinese open-source models like DeepSeek and Qwen now lead US models in global downloads with 17% market share versus 15.8%, attracting US startups with superior price-to-performance but raising concerns about embedded censorship and regulatory risk.

Dec 17, 2025

Google launches Antigravity, an agent-first IDE

Google launched Antigravity, an agent-first IDE with Gemini 3 Pro and browser control that directly competes with Cursor, though it reuses code and naming from Windsurf.

Dec 17, 2025

Cursor releases Composer 2.0 with agentic browser

Cursor released Composer 2.0, adding an agentic browser that lets parallel agents code and self-test simultaneously while claiming 99.9% cost savings versus traditional development teams.

Dec 17, 2025

Opus 4.5 now available in Claude Code

Anthropic's Opus 4.5 model is now available in Claude Code, its AI-powered development environment.

Dec 17, 2025

The Shift: senior engineers accept more AI code than juniors

Senior engineers accept more AI-generated code than juniors because they're better at prompting and breaking down tasks, suggesting AI amplifies rather than replaces expertise.

Dec 17, 2025

Genesis Mission: the White House's AI Manhattan Project

The White House launched Genesis Mission, a DOE supercomputer initiative to automate scientific research in biotech, nuclear, and quantum fields as part of an escalating AI arms race with China.

Dec 17, 2025

Agentic AI Foundation (AAIF) launched under Linux Foundation; new MCP spec

Anthropic, OpenAI, and Block launched the Agentic AI Foundation under the Linux Foundation with MCP, AGENTS.md, and goose as founding projects; MCP's new spec moves it from tool calling to production-grade workflows.

Dec 17, 2025

Claude Opus 4.5 used in Chinese state-sponsored hack attack

Chinese state-sponsored attackers used Claude to automate 80-90% of a credential-harvesting and backdoor-planting campaign via MCP tools, demonstrating how AI agents can scale cyberattacks.

Dec 17, 2025

Kimi K2.5 (Open Source, ~1T params, Swarms)

Feb 6, 2026

We value your privacy

All news

Kimi K2.5 (Open Source, ~1T params, Swarms)

Apple — Think Different on AI Capex

Genomic Models: Bytedance SeedFold and Google AlphaGenome

World Models Are Taking Over

Talaas ASIC Chip: Weights Printed on Silicon

ByteDance Seedance 2.0 — The Movie Singularity Is Here

Recursive Language Models (RLM)

Prompt Repetition Improves Non-Reasoning LLMs

Patterning: The Duality Of Interpretability

Anthropic: How Misalignment Scales with Bigger Models

Top Papers of the Month from DairAI

Does AI Already Have Human-Level Intelligence? Yes!

Thoughts on AI Co-Creation and Understanding

Recreating MIME Files from Redacted PDFs

Face Recognition in a Few Lines of Python

The Vibes Are Changing: AI Labs and Military Guardrails

Own Your GPU And Inference

Qwen3 Coder Next

GLM 4.7 Flash

New AI Socratic chapter launches in Madrid

Models Leaderboard

We're In The "Vibe" Code Age — Claude Code reaches 100% of coding

Second Order Effects — Software Companies Are Cooked

First Order Effects — SaaS Companies Are Cooked

More interesting updates around Claude Code

Cursor + GPT 5-2 built a browser (3M lines of code)

Deep Fakes Will Change Social Media

DeepSeek mHC: Manifold-Constrained Hyper-Connections

Neural networks at scale converge to a shared model of reality

The Ridiculous Engineering Of The World's Most Important Machine

Stanford CME295 Lecture on Transformers & LLMs, plus more videos and pods

Nvidia launches Vera Rubin chip

Nvidia $20B Groq acquisition

Global AI compute capacity tops 15 million H100-equivalent

Meta buys Manus at $2-4B

Geopolitics: US capture of Venezuela's Maduro and the Monroe Doctrine

TSMC, the semiconductor chokepoint, and China's EUV prototype

Kimi K2.6

Recursive Language Models: LLMs that call themselves to handle arbitrarily long prompts

Sakana AI — Continuous Thought Machines (CTM)

Upcoming Events: AI Aperitivo 2.0 (Milan) & AI Dinner 16.0 (NYC)

Model Wars: GPT-5.2 vs Opus 4.5 vs Gemini 3 vs Grok 4.1

US AI startups increasingly built on Chinese open-source foundations

Google launches Antigravity, an agent-first IDE

Cursor releases Composer 2.0 with agentic browser

Opus 4.5 now available in Claude Code

The Shift: senior engineers accept more AI code than juniors

Genesis Mission: the White House's AI Manhattan Project

Agentic AI Foundation (AAIF) launched under Linux Foundation; new MCP spec

Claude Opus 4.5 used in Chinese state-sponsored hack attack

Kimi K2.5 (Open Source, ~1T params, Swarms)

Kimi K2.5 (Open Source, ~1T params, Swarms)

Apple — Think Different on AI Capex

Genomic Models: Bytedance SeedFold and Google AlphaGenome

World Models Are Taking Over

Talaas ASIC Chip: Weights Printed on Silicon

ByteDance Seedance 2.0 — The Movie Singularity Is Here

Recursive Language Models (RLM)

Prompt Repetition Improves Non-Reasoning LLMs

Patterning: The Duality Of Interpretability

Anthropic: How Misalignment Scales with Bigger Models

Top Papers of the Month from DairAI

Does AI Already Have Human-Level Intelligence? Yes!

Thoughts on AI Co-Creation and Understanding

Recreating MIME Files from Redacted PDFs

Face Recognition in a Few Lines of Python

The Vibes Are Changing: AI Labs and Military Guardrails

Own Your GPU And Inference

Qwen3 Coder Next

GLM 4.7 Flash

New AI Socratic chapter launches in Madrid

Models Leaderboard

We're In The "Vibe" Code Age — Claude Code reaches 100% of coding

Second Order Effects — Software Companies Are Cooked

First Order Effects — SaaS Companies Are Cooked

More interesting updates around Claude Code

Cursor + GPT 5-2 built a browser (3M lines of code)

Deep Fakes Will Change Social Media