AI News Archive — Page 10

Google AP2 + Coinbase x402, A2A stablecoin payments

Google's AP2 protocol and Coinbase's x402 standard enable AI agents to pay each other and model context protocols using A2A stablecoin payments.

Sep 18, 2025

Oracle Up 40% Because It's Building Stargate

Oracle's stock surged 40% after the company announced it would build Stargate, the massive AI infrastructure project backed by OpenAI, SoftBank, and MGX.

Sep 18, 2025

gpt-oss-safeguard-20b

OpenAI released gpt-oss-safeguard-20b, a 20 billion parameter safety reasoning model, on September 18, 2025 as open weights.

Sep 18, 2025

Upcoming AI Socratic Dinners and AI Engineer NYC

AI Socratic is hosting dinners in NYC on Sep 10 and SF on Sep 17, plus the AI Engineer conference Nov 20-22 in New York City, limited to 500 attendees.

Sep 9, 2025

Tempo: The Blockchain Designed for Payments

Stripe and Paradigm launched Tempo, a blockchain built for stablecoin payments including payroll, remittances, and agent-to-agent transactions.

Sep 9, 2025

Junior engineers down 23%, senior engineers up 14% (Harvard study)

Harvard study of 62 million workers across 285,000 firms found junior engineering positions fell 23% from 2015 to 2025, while senior roles rose 14%, as AI tools replace entry-level work and compress career ladders.

Sep 9, 2025

Are we in the first inning of an anti-AI movement?

Backlash against AI adoption is growing, with critics questioning whether the technology's promised benefits justify its costs and risks.

Sep 9, 2025

Apple AirPods Live Translation

Apple's AirPods now offer real-time translation that dims the other person's voice while playing the translation in your ear.

Sep 9, 2025

Cognition Raises $400M at $10.2B Valuation

Cognition, maker of the Devin AI coding agent, raised $400 million at a $10.2 billion valuation.

Sep 9, 2025

Mistral Raises $1.7B Series C Led by ASML

Mistral AI raised $1.7 billion in Series C funding led by chipmaking equipment maker ASML, backing the French AI startup's research into frontier models.

Sep 9, 2025

Graph Theory in State-Space (Klotski puzzle video)

A video uses the Klotski sliding block puzzle to visualize state spaces in computer science, showing how simple rules and empty spaces generate complex graph structures.

Sep 9, 2025

Dwarkesh: Two Bottlenecks to AGI

Dwarkesh argues AGI remains distant because current LLMs lack continuous learning and compute scaling alone won't overcome fundamental bottlenecks.

Sep 9, 2025

Kimi K2 0905

Moonshot AI released Kimi K2 0905 on September 3, 2025, a 1 trillion parameter mixture-of-experts model with 32 billion active parameters per forward pass, extending context length to 256k tokens and improving agentic coding and frontend generation…

Sep 3, 2025

Upcoming AI Dinners in New York and San Francisco

AI Socratic is hosting dinners in New York on Sep 10th at Solana Skyline and San Francisco on Sep 17th at Frontier Tower, featuring discussions, demos, and presentations on top AI news.

Aug 28, 2025

Benchmarks & Metrics: Models Increasingly Overfitted to Leaderboards

Most AI models are heavily overfitted to popular benchmarks, making leaderboard scores a poor proxy for real-world performance and more of a tracker for release velocity.

Aug 28, 2025

OpenAI Launches GPT-5 with Advanced Reasoning

OpenAI launched GPT-5 on August 7, 2025, claiming PhD-level reasoning and a 70% coding win rate over GPT-4, but the rollout sparked backlash over glitches, hallucinations, and a colder tone, prompting the company to restore legacy models for paid users.

Aug 28, 2025

Google Launches Gemini 2.5 Deep Think Reasoning Mode

Google launched Gemini 2.5 Deep Think, a reasoning mode for Gemini 2.5 Pro that uses parallel thinking and reinforcement learning, scoring 88% on AIME 2025 math problems and 86.4% on GPQA science benchmarks.

Aug 28, 2025

Genie 3: World Model Generates Interactive Environments

Genie 3 generates interactive 3D environments from text prompts that users can explore in real time using arrow keys, running at 24 fps and 720p resolution while remembering world changes.

Aug 28, 2025

Nano-Banana: Google's Gemini 2.5 Flash Image for Consistent Editing

Google's Gemini 2.5 Flash Image, codenamed Nano Banana, prioritizes consistency when editing visual elements like faces and objects across multiple iterations, addressing a persistent weakness in AI image tools.

Aug 28, 2025

Blog Post: The Second Half by Shunyu Yao

OpenAI researcher Shunyu Yao argues that AI's second half will shift from developing new training methods to defining the right problems and evaluation metrics, with pre-training priors mattering more than algorithms.

Aug 28, 2025

Mechanize: How to Fully Automate Software Engineering

Full automation of software engineering requires scaling beyond larger models to include training on real coding data and reinforcement learning in realistic environments, though building AI that can handle high-level planning and coordination remains…

Aug 28, 2025

Gartner: 2025 AI Hype Cycle Peaks with AI Agents

Gartner's 2025 hype cycle peaks with AI agents as the industry shifts focus from speculation to delivering stable business applications.

Aug 28, 2025

Nic Carter Tweet on AI Opportunities and Negative Effects

Nic Carter warns AI will widen wealth inequality like the post-1850 horse population collapse, predicting job disruptions and regulatory backlash, though the NVIDIA rally has shifted his investment outlook.

Aug 28, 2025

Video: Can Goodness Compete?

A philosophical exploration of whether humanity can coordinate around shared values in a post-AGI future.

Aug 28, 2025

Podcast: The Fractured Entangled Representation Hypothesis

A podcast explores how current diffusion models are brittle "castles in the sand" and proposes building images bottom-up by composing representations piece by piece, demonstrated with a skull example.

Aug 28, 2025

Demis Hassabis on Lex Fridman Podcast: Future of AI and Simulating Reality

Demis Hassabis discusses AI's ability to model complex natural patterns, the path to AGI, AlphaFold's breakthroughs, and why responsible development and shared benefits matter as technology accelerates.

Aug 28, 2025

Video: How AI Image Generation Draws from Physics (Welch Labs)

Welch Labs explains how diffusion models generate images by visualizing the geometry of gradient descent through physics principles.

Aug 28, 2025

MLST: DeepMind Genie 3 - Simulate The World

DeepMind's Genie 3 is a general-purpose world model that generates interactive environments in real time at 30fps and 720p, with emergent properties like parallax and the ability to maintain consistent details across several minutes of simulation.

Aug 28, 2025

Theory Of Everything: The Theory That Your Thoughts Were Never Your Own

A University of Toronto discussion explores whether language is an autonomous entity that colonizes human brains, and whether consciousness itself operates like a large language model rather than generating original thought.

Aug 28, 2025

DeepSeek V3.1

DeepSeek released V3.1, a 671-billion-parameter hybrid reasoning model with 37 billion active parameters that supports both thinking and non-thinking modes.

Aug 21, 2025

DeepSeek V3.1 Base

DeepSeek released DeepSeek V3.1 Base on August 19, 2025, a 671B parameter open-weights mixture-of-experts model with 37B active parameters, 128K context length, and training on 14.8T tokens.

Aug 19, 2025

gpt-oss-20b (free)

OpenAI released gpt-oss-20b, a 21 billion parameter open-weight model under the Apache 2.0 license.

Aug 4, 2025

Qwen3 Coder 30B A3B Instruct

Alibaba released Qwen3 Coder 30B A3B Instruct on July 31, 2025, a 30.5B parameter mixture-of-experts model with 256K token context designed for code generation, repository understanding, and tool use.

Jul 31, 2025

Grok 4 gets an AI companion

xAI launched Grok 4 with 100× more training than Grok 2, claiming state-of-the-art on public benchmarks, but independent testing shows mixed results compared to rivals.

Jul 17, 2025

Windsurf's Updates: Google asset deal and Cognition acquisition

Google acquired Windsurf's assets while Cognition Labs acquired the company itself, allowing employees to join Cognition after the deal structure left them out of Google's asset purchase.

Jul 17, 2025

Kimi 2 👑 — New open source 1T LLMs

Moonshot released Kimi 2, an open source 1T parameter model using a DeepSeek V3-inspired architecture that achieves state-of-the-art performance on several benchmarks while training faster and cheaper than competitors, powered by a custom Muon optimizer…

Jul 17, 2025

Let's enter the AI Code CLI war

A comparison of six AI code CLI tools—claude-code, anon-kode, codex, opencode, ampcode, and gemini—as Google and AWS enter the command-line coding market.

Jul 17, 2025

U-Net, a new recursive tokenizer

U-Net tokenizer processes raw bytes directly with Autoregressive U-Nets instead of predefined vocabularies, eliminating embedding table memory overhead and enabling infinite vocab size.

Jul 17, 2025

A comment on "The Illusion of Thinking"

Pfizer researchers argue AI's reasoning failures are actually an agentic gap: models think fine but fail at action, then succeed when given tools to execute their thoughts.

Jul 17, 2025

Potemkins Understanding in LLM

Researchers identify "Potemkins"—reasoning inconsistencies where LLMs including o3 contradict their own assertions, raising questions about whether systems with such flaws can achieve AGI.

Jul 17, 2025

Videos And Podcasts

https://x.com/karpathy/status/1935518272667217925 https://x.com/dwarkesh\sp/status/1938271893406310818

Jul 17, 2025

AI Dinner 11.0 at Arize.ai Office

AI Dinner 11.0 takes place June 18th at the Arize.ai office, featuring presentations on LLM chess training via reinforcement learning and parallel-sample reasoning techniques.

Jun 16, 2025

AI.Engineer SF Conference: Takeaways and Recordings

AI.Engineer SF conference featured organizational hiccups but delivered high-level insights: competition hinges on clear ideas over code, engineers must master AI tools to stay relevant, and the field is standardizing on MCP while costs plummet and model…

Jun 16, 2025

Anthropic: How We Built Our Multi-Agent Research System

Anthropic describes its multi-agent Research system architecture where a lead Claude agent orchestrates parallel subagents to dynamically explore complex queries, achieving 90% speedup through parallelism but at 15× the token cost of standard chats.

Jun 16, 2025

Google Keeps on Shipping: Gemini 2.5 Pro and Veo3

Google launched Gemini 2.5 Pro, now second only to OpenAI's o3-pro, and Veo3, ranked #2 in video generation models, as the company continues shipping new AI products.

Jun 16, 2025

OpenAI Leads with o3-pro at 80% Lower Cost

OpenAI's o3-pro matches o3 performance on ARC1 and ARC2 benchmarks while cutting costs by 80%, pricing it near 4o-mini levels.

Jun 16, 2025

Meta Takes 49% Stake in Scale AI for $14.3 Billion

Meta acquires 49% stake in Scale AI for $14.3 billion and appoints founder Alexandr Wang as head of AI, though critics question his operational abilities despite fundraising success.

Jun 16, 2025

Mistral Releases Magistral RL Reasoning Model

Mistral released Magistral, a reasoning model trained with modified GRPO that removes KL divergence, normalizes by total length, and adds minibatch advantage normalization for domain-specific and multilingual reasoning tasks.

Jun 16, 2025

Philosophy: Defining Intelligence and Life

Philosophers and researchers grapple with fundamental definitions of intelligence and life, questioning what separates the two concepts.

Jun 16, 2025

Sam Altman: The Gentle Singularity

Sam Altman argues we've passed the event horizon of AI advancement and proposes a "gentle singularity" where AI development continues smoothly rather than catastrophically, noting that ChatGPT queries use 0.34 watt-hours and a teaspoon of water.

Jun 16, 2025

Google AP2 + Coinbase x402, A2A stablecoin payments

Google's AP2 protocol and Coinbase's x402 standard enable AI agents to pay each other and model context protocols using A2A stablecoin payments.

Sep 18, 2025

We value your privacy

All news

Google AP2 + Coinbase x402, A2A stablecoin payments