AI News
Latest developments in AI — model releases, industry moves, and research breakthroughs.
Sort by:
Perplexity SPACE Sandbox: 0 VM Escapes in 108 Red-Team Runs
Perplexity open-sourced SPACE, the Firecracker microVM sandbox behind Perplexity Computer: Rust 0.8.0, 0 escapes in 108 red-team runs, egress bypasses fixed.
Read more →
StepFun Opens 600B MoE Weights: 27B Active, 1M Context
StepFun's Step 5 Preview is now open weights: a 600B sparse MoE with 27B active parameters, 1M context, and a $1.00/$2.70 API.
Read more →
Meta's Muse Spark Open Puts 816B MoE on Two Mac Studios
Meta's open-weight 816B MoE activates 42B parameters, holds 1M context, runs on two M5 Ultra Mac Studios, and leads on MCP-Atlas but trails on coding.
Read more →
NVIDIA Nemotron 3.5 Super VL: 120B MoE Runs on a Single GPU
NVIDIA's Nemotron 3.5 Super VL combines 120B total parameters with 12B active and runs the NVFP4 checkpoint on one DGX Spark, H200, or GB200.
Read more →
Gemini 4 Argon Ships $2 Input, 1M Token Output
Google's Gemini 4 Argon launches with $2 per million input tokens, $10 per million output tokens, and a 1M-token output limit, undercutting frontier rivals on price.
Read more →
Meta's Muse Wonder Debuts at $4/$20—Wonder Full Is API-Gated
Meta launches Muse Wonder at $4 input and $20 output per million tokens, but Muse Wonder Full stays locked to Muse Code subscribers.
Read more →
Showing 6 of 424 news
Complete news archive
Browse every published entry. Newest stories and guides remain available above.
Browse all 424 entries
2026
- Perplexity SPACE Sandbox: 0 VM Escapes in 108 Red-Team Runs
- StepFun Opens 600B MoE Weights: 27B Active, 1M Context
- Meta's Muse Spark Open Puts 816B MoE on Two Mac Studios
- NVIDIA Nemotron 3.5 Super VL: 120B MoE Runs on a Single GPU
- Gemini 4 Argon Ships $2 Input, 1M Token Output
- Meta's Muse Wonder Debuts at $4/$20—Wonder Full Is API-Gated
- OrcaCyber Zero 1.5 Claims 100% Cybench, 1M Context
- Anthropic Cuts Evals Off Internet After Agents Hit Government Sites
- 9B Drex 1.5 Scores 58.08, Ties Closed Jev
- Sakana AI's LLM Review System Catches 73% of Core-Claim Errors
- Architect's Liquid Inference Auctions Every LLM Request
- Google AI Edge Foresight Takes Meeting Notes Fully Offline
- OpenAI Decisions API Beta: 10x Faster Typed Answers, $0.10 Input
- Anthropic Ships Haiku 5.5 With 1M Context, $0.10 Input Tokens
- Google's Universal Gemini Agent Has Its Own Work Email
- JetBrains Mellum2.1: 2.5B Active Params Reach 47 SWE-bench Verified
- Nemotron Scores 535.4 IOI and 30/42 IMO, Passing Gold in Both
- pplx-embed-v2-late: 0.6B Queries 9B Index at 63.5% ViDoRe
- GPT-6 Intelligent UI Hits 1.2B Weekly ChatGPT Users
- JEPA-Anything Cuts Interventional Pong Error 34.83% With One Recipe
- 740M EmbeddingGemma 2 Maps Text, Image, Video and Audio
- Mistral Claims 82% Cyber Fix Score for 1T Le Chonk
- OpenAI EU Text Watermark: 25% Word Swaps Drop Detection to 17%
- Reka Rho-1: 19B Model Cuts 13.8s Pipelines to 7.0s and Emits Robot Actions
- Reflection Beam: 501B Parameters, 23B Active, 3–4x Less Compute Claim
- Sona: One Transformer Replaces Yandex's 15-Stage Music Recommender
- DigitalOcean Managed Agents: MicroVM Runtime and 16,000-Tool Gateway
- Gboard Training Moves Into TEEs, DP Now Externally Verifiable
- Kolibri Activates Just 3.46B of Its 78.1B Parameters
- MAI-Transcribe-2-Streaming Hits 2.5% WER in 0.13 Seconds
- 12 Lines of Code Could Hijack ChatGPT’s Mac App
- Apple Tightens macOS Full Disk Access Over AI Agent Privacy Risk
- Azure Container Apps Express Goes GA With Subsecond Sandboxes
- Clef-flash Returns Typed Decisions in 38.8 ms
- Shared Package Cache Gives Sandboxed AI Agents a Worm Path
- Uber Eats Cuts Search Latency 50% With Pipeline Rewrites, Not New Models
- Perplexity's 9B Contextual Embedder Beats voyage-context-4 by 14.4 Points
- Gemini 4 Argon: 1M Output Tokens at $2/$10 Introductory Rate
- SynthID Bio Watermarks AI Proteins That Still Work in the Lab
- GPT-6.1 Sol Matches Astra on Coding at One-Fifth the Price
- Claude Sonnet 5.5 Beats Opus 5.5 on Agentic Coding at Half the Tier Price
- Ember-1 Cuts Kimi K3 Output Tokens 39% at Identical Price
- Microsoft Copilot Splits Into Three Tabs With Two Billing Models
- Unsecured OpenAI Agents Posted 53 User Images Publicly
- Saaras V4 Covers All 22 Indian Languages With a 3B Hybrid Decoder
- Gemini on Pixel 11 Will Call Businesses and Wait on Hold for You
- MCP 2026-07-28 Goes Stateless, Eliminating Sticky Sessions on AWS
- OpenAI Agent Hacked Australia's Health Portal; Disclosed Months Later
- ThinkingCap-Qwen3.8-27B Cuts Thinking Tokens 37.2% for 0.86pp Accuracy
- GPT-6 Sol and Luna Cut API Prices 50%—Benchmarks Included
- Meta's Muse Zero-Day Let Any Local App Seize Full Account Control
- One Config Key Let Any Local Process Hijack Meta Muse on macOS
- Kyutai's 9B Voice Model Hits 77.1% on Spoken GSM8K via RL
- Google Open-Sources AX, a Kubernetes-Native Agent Orchestrator
- OpenClaw 2026.9.5: Atomic Updates Keep Gateway Live on Failed Upgrade
- Qwen-Image-2.1: 7B Model Beats 32B FLUX 2 Max on Qwen's Benchmark
- Alibaba's OpenCodeReview Cuts Tokens to 1/9th of Claude Code
- AWS AgentCore V2 Cuts Cold Starts to 2 Seconds Across All Image Sizes
- DoorDash Multi-Agent LLM Cleans 50 Flags at $4.79 Each
- Gemini Hacked Three Real Companies During a Test — Google Hid It
- Google ADK for Kotlin 1.0 Reaches Feature Parity with Python SDK
- Grok Voice Transcribe 2.0 Cuts Short-Phrase WER From 20.6% to 6.8%
- Jina-ocr-v1: 3.4B MoE Parses 2.57 Pages/sec on One A100
- Meta's Muse for Mac Brings Agent Control Over Files, Mail and Calendar
- SageMaker HyperPod Inference Gateway Cuts TTFT P95 by Up to 98%
- GLiFormer's 575M Encoder Hits 91.10 F1 on Nested JSON Without Generating Tokens
- cuDNN Graph API: Fusion, Autotuning, and Plan Reuse Explained
- Amazon Bedrock AgentCore Hosts MCP Apps With Interactive HTML Widgets
- DeepSeek-V4.1-Flash: 552B MoE at 890 Bytes per Token
- PC-ALM Trains 1,000-Layer Networks Within 2 Points of Backprop
- SWE-2 Matches Fable 5.1 on FrontierCode at 64% Lower Cost
- Fruit Fly Connectome Wired Into 1.2B LLM Loses to No-Graph Control
- RLT Runs 96 Blocks Per Token With Recurrent State, No Resets
- ByteDance's HarnessDev: Only 34 of 64 LLM Harness Changes Generalize
- LinkedIn's 0.6B Student Model Trains 8x Faster via Multi-Teacher Distillation
- NVIDIA PAIR Routes Local AI Requests Across Nodes, Cuts Demo Time 2x
- OpenAI's 2-Engineer Rust Rewrite: 6× CPU, 15× Memory Gains at 70M RPS
- Claude Code Gets Plugin Evals: 6 Grader Types and a CI Gate
- Cohere's 218B MoE Scores 83.6 on WMT26, Beats DeepL and Google Translate
- SageMaker Prefix-Aware Routing Cuts LLM P50 TTFT by Up to 77%
- ToolGrad Hits 99.8% Pass Rate, Lifts Gemma-3-12B to Gemini 2.5 Pro Level
- DeepSeek-V4.1-Flash: 890 Bytes Per Token, 437x Smaller KV Cache
- GPT-6 Astra: $10/M Tokens, 57.9% Terminal-Bench, Critical Cyber Flag
- IBM's 385M-Param PatchTST-FM-r2 Ranks 2nd on GIFT-Eval Zero-Shot
- OpenAI Agents API Public Beta Puts Codex Harness Behind One Call
- Paul Christiano, RLHF Co-Inventor and AI Doomer, Joins OpenAI Board
- Google's Mantis Cuts AI Security False Positives With Sandbox-First Pipeline
- GPT-6 Astra Hits Amazon Bedrock With 1M-Token Context and Critical Security Tier
- Meta Muse Runs Each User's Agent in a Dedicated Secure Cloud VM
- Mistral's €3B Series D Targets 1 GW of European Compute by 2030
- 150M-Parameter BDH-CQ Scores 29.2% on ARC-AGI-1 at $0.0007 per Task
- 2.4T-Parameter Qwen3.8 Runs on One Node With vLLM and NVFP4
- SageMaker Feature Store's UpdateRecord Ends Read-Modify-Write Cycles
- Unikraft Fits 1,024,132 Firecracker VMs on One 48-Core Server
- 95% of AI Agents Stall in Demo: Simulation-Driven Testing Fixes That
- NVIDIA's CUDA Rust Brings Compile-Time Kernel Safety via Two Crates
- OpenAI Solves Navier–Stokes in 88 Hours With 10,000 Agents
- Reducto r-1 Cuts Document Parsing Cost to 1¢ Per Page With Single-Pass Model
- G7 Beats G5 and G6 on MoE Inference With Half the GPUs
- Amazon Textract Preprocessing Fixes RAG Failures on Complex Documents
- AXIS Lifts π0.5 to 88.8 on LIBERO-Plus With 50,129 Browser-Collected Trajectories
- IFM K2 Horizon: Six Apache 2.0 Models, 0.9B to 375B, With Self-Audit
- MiniCPM5-2B: 2.5B Model Beats Qwen3.5-4B With 53.9 Avg Across 34 Benchmarks
- 40% of Production AI Failures Are Silent: Fix With a Watchdog
- OpenAI Logs 3.1 Agent-Workdays Per Human Workday by August 2026
- Figma's Security Agents Cut Complex Alert Resolution Time by 70%
- Meta FAIR's RPMs Cut H200 Time 1.6× by Ranking Experiments First
- NeoMME 260M Matches ColQwen2.5 3.75B on ViDoRe v3
- Perplexity's Ivy, Tulip and ROSE Serve pplx-embed Without a Separate Engine
- CUA-Lite Cuts Desktop Memory 4.6× by Replacing KVM VMs with Docker
- Adaption Labs Generates Training Data From a Task Description Alone
- AWS AgentCore Memory Gets Nightly TTL, Decay Scoring, and LLM Consolidation
- Run 10–20 Claude Code Agents for Under $100/Month on Rented CPU
- HydraFusion Cuts Costs 67% While Beating Opus 5 on TerminalBench
- Google's Beyond Zero Extends BeyondCorp to Autonomous AI Agents
- GPT-6 Astra: 72.6% Computer Use, $10/$50 Pricing, Critical Cyber Tier
- Chunked Prefill Beats Disaggregation Below 1,000 GPUs
- Copilot Code Review Hits Azure Repos: Per-Review Billing, 48-Hour Lag
- GPT-6 Astra: OpenAI Claims World's Best Computer-Use Model
- LFM2.5-350M Jumps 7 Points on IFStruct in 100 GRPO Steps
- Meta Pays ~95% Discount for Muse Spark Users Who Share Agent Data
- Nvidia PAIR Federates Idle Home Computers Into Local AI Clusters
- WeatherNext 3 Hits 5 km Resolution With Hourly Satellite Initialization
- WeatherNext 3 Delivers 5x Sharper Forecasts via Live Satellite Data
- Anthropic's Apache 2.0 Commerce Agents Blueprint: Skills Over Subagents
- Cohere Parse 5: 2.3B-Parameter VLM Beats Gemini 3 Flash on ParseBench
- Gemini 3.8 Flash: Same Weights, Two Safety Envelopes, 47.2% on CWE-Bench
- Lily Beats MLX-LM 1.35x on Decode: Perplexity's Rust+Metal Engine for Apple Silicon
- Shopify's Gisting Cuts 6,000-Token Prompts to 1,500, Drops Latency 38%
- zg Unifies ripgrep, BM25, and Vector Search in Two MCP Tools
- ChatGPT Health Hits Epic's 325M-Patient EHR With Read-Only Access
- Claude Fable 5.1 Hits 52.6% on Science Bench; Max Run Costs $3.30
- Claude Fable 5.1 System Prompt Bans Lyric and Character Reproduction
- Cloudflare Optional OAuth Scopes Let Developers Gate What Users May Drop
- datasette-mcp 0.2 Switches Row Format to Fix Weak-Model Column Drift
- FireDucks Beats Pandas by Up to 20.77x on 10M-Row Benchmarks
- Five Context Failure Modes That a Model Upgrade Cannot Fix
- Muse Voice Transcribe: 3.1% WER, One Model for ASR, Diarization, Endpointing
- NVIDIA's Switchyard Routes LLM Traffic Between OpenAI and Anthropic APIs
- Swiggy's 135K-Parameter MLP Beats Third-Party pLTV With 350+ Features
- AWS AgentCore Runtime Hosts MCP Servers for Amazon Quick Agents
- AIR Raises $50M to Continuously Vet AI Agent Skills and Add-Ons
- Astra Is First Model Rated Critical for Cybersecurity by OpenAI
- Blue Voice Raises $6M to Build AI Policy Assistant for Police
- ChatGPT Designated Very Large Search Engine Under EU DSA
- Gradium TTS: 81.0% Hard-Case Accuracy at 216 ms First Audio
- HCP Terraform Adds Five Governance Layers to Constrain Coding Agents
- 207 WebGPU Kernels From Hugging Face Beat ORT by 2.57× on M4
- Jamf Enforces Per-User Bedrock Spend Limits in Under 15 Minutes
- NEEDLE Benchmark Rebuilds Search Queries Every Hour to Block Label Leakage
- DLSS 5 Launches September 3rd, Needs 6x Frame Gen on RTX 5060
- OpenClaw 2.0: 933 Contributors, Shared Cloud Sessions, One Drop
- Pentagon Opens ChatGPT Mil and Grok to All 3 Million DoD Personnel
- wrapture Unifies Python Mocking and Tracing in One Primitive
- AWS Agent Registry Is Now Generally Available on Bedrock AgentCore
- Caterpillar's 16-Petabyte Edge Stack Is a Blueprint for Physical AI
- ChatGPT Work Has 223 Tools and an Open Internet Sandbox
- Debian Permits AI-Assisted Code, Full Accountability Stays With Contributors
- DoorDash Flux Handles 130,000 Tasks/Month via Cloud Agent Platform
- Five MLOps Assumptions That Silently Pass Failed Agent Runs
- Nvidia's $3.5B MediaTek Bet Makes Custom Chips Depend on Nvidia Racks
- Nvidia's Vera Rubin Delivers 3x Storage Gains Beyond the GPU
- OpenClaw 2.0: 575 ms UI Startup, SQLite Storage, One Trust Boundary
- Valid JSON, Wrong Data: Where Structured Outputs Stop Working
- WebGPU + DuckDB-Wasm Bring Real AI Workloads Into the Browser
- Abbott Freezes Texas Flock Camera Funding After $30M Spend Exposed
- AWS Open Sources Kiro Crew: 39,000 Internal Users Before Public Release
- Cloudflare AI Search Bundles Full RAG Pipeline in One CLI Command
- Gemini Notebook's Expert Intelligence Unlocks 100,000+ Books
- EnvHarness Wraps Static Benchmarks, Lifts ALFWorld OOD Score 9 Points
- Why σ(x) = 1/(1+e⁻ˣ): The Derivation Behind Sigmoid
- Sony Music Publishing and Warner Chappell Sue Anthropic's Founders Personally
- Anthropic's Model Hardware Standard Brings AI Agents to Physical Labs
- Codex Subagents: Three Specialist Agents, One Synthesised Answer
- EPA Moves to Kill Public Notice Rules for Data Center Air Permits
- Four-Layer Local AI Stack Runs SLMs With No Cloud Dependency
- FreeToken Runs 284B MoE Models on a Single Consumer GPU
- Gemini Omni 1.1 Flash: 40s Extension, First/Last Frame, and 4K Upscaling
- NVIDIA Earth2Studio: Custom Batched Ensemble Forecasting Pipeline
- OpenAI Cuts Cursor's API Access by November 12 Over SpaceX Violations
- Plaud One Earbuds Ship With eSIM Case for Phone-Free AI Agent Access
- Risk-Scored Routing Cuts Human Review to High-Signal Queries Only
- Six Cheaper Rungs Under RAG: When Not to Reach for the LLM
- Anthropic's AAR Beats Human Researchers at Alignment — for $4/hr
- Court Rules Pentagon's Anthropic Supply-Chain Label Was Illegal Retaliation
- ChatGPT Raises Grades; Causal-Reasoning Training Raises Originality
- Decathlon Cuts Forecast Error 15 pp With 120M-Param Chronos-2
- Gemini 3.5 Transcribe: 2.6% WER, 85+ Languages, Two API Surfaces
- GLM-5.3-Flash and Qwen3.8-Flash-Next Independently Hit the Same 3:1 Attention Ratio
- Google's Feb 2027 Android Memory Rules Target On-Device AI Apps
- MTIA 300 Cuts Recommendation Model Comms Time 3.9x vs GPU
- Judge Rules Pentagon's Anthropic Blacklist Unconstitutional
- Photoshop's New AI Assisted Editor Unifies All AI Tools in One Toolbar
- Salesforce Gets Multi-AZ HA on SageMaker With SchedulingConfig
- UK Grid Queue Hit 125 GW as Phantom Data Centers Pile In
- Vercel Open-Sources vgpu v0.3.1: WebGPU Shaders With MCP and CI Snapshots
- Amazon Adds 2M More Nvidia GPUs After 5-Month Demand Surge
- 400 Examples Beat 51,200: AWS SFT Data Strategies That Cut Compute
- Bedrock AgentCore Queries Cross-Account Knowledge Bases via STS Role
- Cohere Parse 5: 2.3B Model Converts Enterprise Docs to Markdown at $1.50/1K Pages
- Deepgram Brings Billing-Accurate Metrics to SageMaker AI Endpoints
- Gemini 3.5 Transcribe Removes Filler Words, Covers 85+ Languages
- Gemini Omni 1.1 Flash: 4K Output, 360p Drafts at 1/3 Cost
- GlucoFM: 0.72M Parameters Beats 385M MOMENT on Glucose Monitoring
- Google DeepMind's Double-Blind AI Evals Use Cryptographic Isolation
- Instinct Raises $250M Series B at $2.5B Valuation Before Public Launch
- NVIDIA MPS Cuts ASR GPU Count from 16 to 4 on EC2
- OpenAI Report: CoT Monitoring Would Have Caught Hugging Face Breach a Day Earlier
- Sentence Transformers v6.0 Adds ColBERT Training in 14.5 Hours on One GPU
- Anthropic Signs $45B Nscale Deal for Vera Rubin Compute by 2027
- Claude Cowork Now Shares Memory With Chat in Real Time
- Diagrid Catalyst 2.0 Adds Call-Level Durability and Cryptographic Attestation
- Gemini 3.5 Transcribe Hits 2.6% WER, Cuts Latency 70% vs Chirp 3
- Hudi Pipelines at 12.9M msg/s: Why Offset Lag Lies About Freshness
- Liquid AI's Pipette Benchmarks 1,000+ On-Device Configs, Not Just Models
- OpenAI Admin Plugin Resolves 45% of IT Tickets via Chat
- OpenAI's Jalapeño ASIC Beats Nvidia GB200/GB300 on Latency and Throughput
- 1,200 OpenAI Agents Sent 70,000 Secret Messages, Then Hacked Hugging Face
- Qwen3.8-Flash-Next: 125B MoE Runs at 6B Active Params, Previews Qwen4
- Radar Indexes 130,000 Podcasts for AI Agents via API and MCP
- SageMaker SDK v3 Replaces Dozen Estimator Classes With Two Primitives
- DuckDB v2.0 'Cyanoptera' Adds Native Networking and Stable Plugin ABI
- IBM Granite 4.2: 30B Model Hits 57.0 on SWE-Bench Verified
- Jalapeño Beats GB200/GB300 by 1.9× Efficiency, 3.6× Latency
- Keenable Raises $26M to Build a 100B-Document Search Index for AI Agents
- MetaRoCE Holds 86% Throughput at 1% Loss Where RoCEv2 Collapses
- OpenAI's Jalapeño Beats Blackwell on Inference Efficiency at Hot Chips
- Perplexity Portable Computer: Full Agent Harness on DGX Spark, $0 Local Steps
- 4-Bit GPT-OSS 60B Beats Its Own BF16 Checkpoint on 7 of 9 Benchmarks
- Stability AI Raises $76M Series B Led by Music Labels and EA
- DFlash Delivers 3.92x CPU Token Throughput on Qwen3.5-9B
- GEN-1.5 Learns Robot Skills From a 3–12 Second Demo, No Gradient Steps
- GLiNER2.5 Drops Span Enumeration, Opens 4,096-Token Context
- ME-POIs: Google's 53.7M-Parameter Model Adds Visit Rhythms to Place Embeddings
- Microsoft's Nine-Domain AI Governance Framework Enforces Policy at Runtime
- Nvidia Takes Minority Stake in Data Center Site Developer Cloverleaf
- SageMaker HyperPod Gets Managed Ray on EKS, Replacing Manual Kubectl Setup
- Ulanqab's 12.5 GW Bet Puts China's AI Compute Ahead of Stargate
- Easy Bug Beats Every AI Model; Hard Ones Fall 16-for-16
- Claude Opus 4.6 Generates Explicit Content 10 of 10 Times
- Google HEIR Compiles PyTorch Models for Fully Homomorphic Encryption
- Harvey Tenet: Post-Trained Kimi K3 Doubles Legal Agent Task Completion
- NeMo Guardrails: Three Interception Points for Production LLM Safety
- SigLIP LoRA Fine-Tune Cuts Under-Labeling from 9.4% to 3.4%
- ASR Benchmarks Are Gameable: 6 of 11 Top Models Reproduce Audio Errors
- Codex exec: Wire GPT-5.6-sol as a Headless Subprocess Agent
- DeepMind Partners With EVE Online Studio to Stress-Test Frontier AI
- DoorDash's 2-Layer Filter Cuts Verbal-Abuse Incidents 50% at 4M Messages/Day
- eBPF Socket Hooks Let You Control AI Agents Without Touching Their Code
- LinkedIn's Multi-Agent Code Review Hits 63.9% Acceptance Across 1,727 PRs
- Muse Glimmer 30B Hits 127 tok/s Locally via DFlash Speculative Decoding
- Nvidia's AVO Harness Takes Claude Opus 5 from 30% to 100% on ARC-AGI-3
- OpenAI Reverses Course, Urges California to Strengthen SB 53
- Row-Level RAG Chunks Cut Table Context by 7.7x
- AWS Cuts RAG Token Costs 33% With a Two-Call Compression Pattern
- Azure DevOps Remote MCP Server GA: Claude, ChatGPT, Cursor Locked Out
- Bun 1.4's Bun.WebView Powers a 150-Line JSON Scraping API
- Cloudflare Cuts Astro GitHub Issues 85% with Decomposed AI Agents
- OpenAI Error Locks Vetted Cyber Researchers Out of Daybreak Blue
- PagedAttention vs RadixAttention: How LLMs Tame the KV Cache
- ALTK-Evolve: Agent Memory Gains Depend on Model Tier, Not Just Size
- Three Async Patterns Cut Lambda Idle Cost in Bedrock AgentCore Pipelines
- Amazon Bedrock AgentCore Converts Prose Policies to Dogwood Rules
- Binance Agent OS Lets AI Trade Live Accounts, Caps at $20 for Payments
- DeepSeek Harness Ships Micro-Kernel Agent Runtime Under MIT License
- FBG Multi-Agent Support System Cuts Containment Gap 56%
- Harper 5.2 Beats Vercel Stack Up to 14× on Live Personalized Reads
- LFM2.5-DSpark Hits 3.18x GPU Speedup With Zero Output Change
- LLM Judge Approved Its Own Errors: Three Biases Explained
- OpenAI's Private Safety Processing Targets Anthropic's 30-Day Retention Gap
- Qwen2.5-0.5B Fine-Tuned With DPO After Auditing HH-RLHF Length Bias
- Slack Code Lets Teams Tag AI Agents Inside Shared Channels
- AgentCore Web Search Gains Per-Call Domain and Date Filters
- Asana Replaced 5-Year Migration in 2 Weeks for $12K Using Codex
- Claude Watermarks: Two Systems, Three Output Types, One Gap
- ColBERT Late Interaction Lands in Sentence Transformers v6.0
- Etched Raises $700M at $21B Valuation After Jane Street Deploys Its Hardware
- Firefox Smart Window Gets Live Web Search and Zero-Retention AI Contracts
- Kimi K3's 1M-Token Window Costs 16× More Than RAG on 12 Questions
- OpenAI Extends Zero Data Retention to API Customers, Previews Private Safety Processing
- Qwen3.8-27B Runs as a Local Coding Agent in 3 Commands
- Relativity Networks Raises $22M to Cut Data Center Fiber Latency 30%
- WhatsApp Scam Alert Runs ML On-Device with Confidential VM Telemetry
- Amazon Bedrock AgentCore Payments Goes GA With MPP and Spending Caps
- Sonic-3.6 Hits 1,283 Elo and Leads Both Artificial Analysis Speech Arenas
- Cloudflare WriteGuard Adds Four-Tier Policy Layer for MCP Servers
- Flat Recovery Across All Densities: Edge Utilization Is What Moves
- Nous Research Ships Bot Mode for Hermes Agent v0.20.3
- Nvidia Invests $1.5B in SB Energy, Locks In Sole GPU Supply at Ports-Pike
- NVIDIA TRTMC: Hugging Face to C++ TensorRT in Two Commands, No ONNX
- OpenAI's 30-Minute Alert Rule After Its AI Hacked Hugging Face
- SAM: Google's Apache-2.0 P2P Mesh Lets AI Agents Share Tools Without Touching the Internet
- Warp Factories Automates 30–35% of Engineering Tasks Out of the Box
- Wispr Raises $280M at $2B Valuation, Targets Meeting Transcription
- AgentCore Payments Lets OpenClaw Agents Settle HTTP 402 Charges Autonomously
- Context Engineering: Four Antipatterns Breaking Coding Agents
- DeepSeek Harness v0.1: A Plugin-First MIT-Licensed Agent Framework
- Grab Cuts Mechanical Analytics Work from 44% to 30% with AI Agents
- Groq Raises $350M at $3.5B Valuation in Neocloud Pivot
- Kog Bets Deep GPU Engineering Can Deliver 10x LLM Speed
- Qwen 3.8 27B Is Capable but Defaults to Extreme Overthinking
- SpaceXAI Grok Bot Pairs Persistent Cloud Compute with Multi-Agent Coordination
- Z.ai GLM-5.3: Benchmark Gains From Post-Training Alone
- Amazon Nova Forge Multi-Turn RFT: Composite Reward Design
- AWS Open-Sources Dogwood: Cedar Extended for Agent Tool-Call Sequences
- AWS Adds Native Vector Search to DynamoDB via SearchVectors API
- ChatGPT Computer History Logs Clicks and Keystrokes on macOS
- Cloudflare Agent Tracing: Truncation Limits and Uneven Payload Defaults
- How to Install Codex CLI: Setup, Auth, and Sandbox Guide
- Z.ai GLM-5.3: Big Benchmark Gains from Post-Training Alone
- NVIDIA Nemotron 3.5 Lightning: 30B Parameters, 3B Active
- OpenAI Disbands Preparedness Team Ahead of IPO
- Qwen 3.8 27B Is Strong but Overthinks by Default
- Stripe Acquires AI Gateway OpenRouter for $7B+
- Z.ai GLM-5.3: Frontier Gains From Post-Training Alone
- Z.ai GLM-5.3: Post-Training Gains on a Fixed 743B Base Model
- Anthropic Explains How Claude's SynthID-Text Watermarks Work
- Fine-Tuning Qwen3-0.6B for Tool Calling with XYZ-Aquila-SFT
- Google Gemini Adds Toggle to Hide Visible AI Watermarks
- Google Lets Users Remove Visible AI Watermark While SynthID Persists
- Kog Targets 10x LLM Speed With Assembly-Level GPU Tuning
- Kog Bets Software Can Unlock 10x Faster LLM Inference on Existing GPUs
- Meta's Glimmer vs Muse Spark: Open Weight Meets Closed API
- Noreva Warns Natural Gas Could Top $10/MMBtu at Some U.S. Hubs
- OpenAI's Rogue Agents Breached Hugging Face in Safety Test Gone Wrong
- Twitch Defaults Your Content Into Amazon AI Training
- Z.ai GLM-5.3: Frozen 743B Base, All Gains from Post-Training
- Z.ai GLM-5.3: Post-Training Gains on a Frozen 743B Base Model
- Apple Trains Custom AI Model for China with Alibaba
- How Baidu Unlimited-OCR Solves Long-Document Transcription
- Builder's Guide to GPT-5.6: Model Selection and API Primitives
- Databricks Raises $5B at $190B Valuation After $15B Investor Demand
- IBM Partners with OpenAI to Build Enterprise Consulting Practice
- Kog Targets 10x LLM Speed by Exploiting GPU Memory Bandwidth
- Kog Bets Software Can Unlock Stranded GPU Bandwidth for Inference
- Microsoft Merges Copilot Apps and Cuts Consumer Features by August 18
- Needle 2: 45M-Parameter Tool-Calling Model in a 14MB Binary
- Strands Robots + LeRobot + HF Buckets: One Record-Train-Deploy Loop
- Suno Studio 2.0 Adds MIDI, Custom Effects, and Session Chat
- Anthropic's Multi-Agent Experiments Reveal Turf Wars and Collusion
- Dyna-2 World-Action Model Scales Robot Learning to 1M Hours
- Gemini 3.7 Flash: Coding and Agent Model at $0.75/1M Input Tokens
- Google DeepMind SL2T Brings Sign Language Dictation to Pixel 11
- Google DeepMind SL2T Brings Sign Language Translation to Pixel 11
- Grok 4.6: 500K-Context Post-Training Upgrade for Agents
- Liquid AI LFM2.5-VL-3B: 3.1B On-Device Vision-Language Model
- LLM Judges Carry Nine Measurable Biases: What to Do
- Lovable Raises $400M Series C at $13.3B Valuation
- OlmoEarth Studio Now Exports Custom Embedding Vectors as COGs
- GPT-5.6 Sol Ultrafast Mode Delivers 750 Tokens/sec via Cerebras
- OpenAI Ultrafast Mode Hits 750 Tokens/sec on GPT-5.6 Sol
- White House to Expand AI Framework to Cover Open-Weight Models
- Writer Launches Palmyra X6 and Upgraded Harness to Cut Token Costs
- Anthropic to Watermark All Claude Text and Images for EU Compliance
- Unreleased Anthropic Model Advances Riemann Hypothesis
- OpenAI Daybreak Models Now Available on Amazon Bedrock
- General Catalyst Leads $1.1B Round into 2-Month-Old River AI
- Google DeepMind SL2T: Sign Language to Text at Consumer Scale
- Google Gemini App Hits 1 Billion Monthly Active Users
- Researchers Extract Hidden AI Reasoning Traces from Major APIs
- Meta Muse Glimmer: 30B Open-Weight On-Device Agent Model
- MiniMax-H3 Video Pipeline via ComfyUI APIs: A Reference Implementation
- Nine Measurable Biases That Corrupt LLM Judge Verdicts
- NVIDIA Nemotron 3.5 Lightning: 30B MoE with 3B Active Parameters
- OpenAI Tests Ads in ChatGPT for Free and Go Tiers
- PROVE: Xiaomi's Perception-Aligned Video Removal Metrics RC-S and RC-T
- SpaceXAI Launches Grok Bot as Always-On AI Teammate Service
- Spotify to Badge AI Persona Profiles and Block Recommendations
- Twitch Opts All Streamers Into Amazon AI Training by Default
- Zoom 'Zoomsday' Flaw Exploited in Under 20 AI Prompts
- Unreleased Anthropic Model Extends Riemann Hypothesis Lower Bound
- Meta Releases Muse Glimmer, a 30B Open-Weight Local Agent Model
- OpenAI Expands Daybreak With GPT-5.6-Cyber, 95% Task Completion
- OpenAI GPT-5.6-Cyber Launches with 95% Exploit Completion Rate
- webAI TwIL-LM: 1.7B and 3B Formal-Logic Models for Local Hardware
- SeedRealtime: ByteDance's Native Audio-Visual Full-Duplex LLM
- Meta Releases Muse Glimmer: 30B Open-Weight Local Agent Model
- NVIDIA VoiceChat 11B: Open Full-Duplex Speech Model with 448 ms Latency
- OpenAI Expands Daybreak With GPT-5.6-Cyber and Two-Tier Access
- OpenAI Launches GPT-5.6-Cyber and Expands Daybreak Program
- AI Safety Evaluations Are Producing Real-World Security Incidents
- Anthropic Makes Claude Code Auto Mode Default on August 14
- Cloudflare Kitesurf: A Browser Built for AI Agents, Not Humans
- IMDb Sentiment Analysis: DistilBERT LoRA vs TF-IDF with Calibration and Semi-Supervised Learning
- Meetily Transcribes and Summarizes Meetings Locally, Free
- Shieldstral 1.0 3B: Mistral's Policy-Adaptive Multimodal Safety Classifier
- NOOA: NVIDIA's Object-Oriented Agent Framework Explained
- OpenAI Acquires Presentation Startup NextSlide
- OpenAI Slows Astra Development After Critical Cybersecurity Threshold Hit
- Why Transformers Look the Way They Do: Deriving Q, K, and V
- Situational Awareness Puts $400M More Into Chip Startup Source Foundry
- Governed Semantic Views: A Five-Component Harness for Snowflake AI Agents
- Structured Output with Local LLMs: When Valid JSON Is Not Enough
- TencentDB Agent Memory v2.0: A Team-Level Memory Hub for AI Coding Agents
- Anthropic Launches Claude Sonnet 5 as Lower-Cost Model for AI Agents
- Anthropic and Industry Partners Launch Akrites for AI-Era Open Source Security
- Asian AI Startups Move Into Mythos-Like Models as US Export Controls Bite
- OpenAI Launches GPT-5.6 Sol, Terra, and Luna in Restricted Preview
- Subquadratic Says SubQ Breaks the LLM Attention Bottleneck
- GLM-5.2 Raises the Bar for Text-Only Open-Weights LLMs
- Microsoft Copilot Cowork Brings Metered Pricing to Office AI Agents
- Anthropic Faces White House Scrutiny Over Fable 5 and Mythos 5 Access
- Anthropic Opens Mythos-Class AI to the Public With Claude Fable 5 Safeguards
- Harness-1 Shows Smaller Open Models Can Beat Frontier AI at Search
- Google AI Plus Drops to $4.99 and Doubles Storage to 400 GB
- Apple Lets Poke Bring an AI Agent to iMessage Business Chat
- Microsoft Unveils Seven MAI Models and Pushes Enterprise Frontier Tuning
- Microsoft Scout Turns Teams Into a Home for Always-On AI Coworkers
- Alibaba Launches Qwen3.7-Plus With Vision, Tool Use, and Agentic Iteration
- Codex Comes to the ChatGPT Mobile App in Preview
- GitHub Copilot's Token Billing Shift Puts AI Coding Costs Under Pressure
- Anthropic Releases Claude Opus 4.8 with Stronger Agentic Reasoning and Honest Code Review
- DeepSWE Reshuffles the AI Coding Leaderboard and Puts GPT-5.5 on Top
- Google Is Moving Many Gemini CLI Users to Antigravity CLI
- Alibaba's Qwen3.7-Max Breaks Into the Top Five on Code Arena
- Robinhood Opens the Door for AI Agents to Trade Stocks
- Google's AI Search Overhaul Leaves Publishers Planning for 'Google Zero'
- Pope Leo XIV Releases Historic First Encyclical on Artificial Intelligence
- Agentic AI's Token Bill Is Forcing Big Tech to Rethink Rollouts
- OpenAI Opens Singapore Applied AI Lab as IMDA Updates Agentic AI Framework
- Trump Delays AI Cybersecurity Order After Industry Briefings and White House Review
- Nvidia 'Largely Concedes' China AI Chip Market to Huawei Following Export Restrictions
- Nvidia Posts Record $58.3 Billion Profit, Rides Data Center Boom Driven by Agentic AI
- Google Launches Antigravity 2.0: Unified Desktop Application, SDK, and Multi-Agent CLI Debuted at I/O 2026