Claude Prompt Caching: 6 Tricks That Cut Your API Bill
Breakpoints, TTL math, Batch stacking, and a two-session split. Measured on OpenRouter, $0.003 vs $0.022 per turn
// Intelligence
Today in AI
A focused feed of models, agents, research and open-source releases for people building with AI.
Updated
Suno Splits v6 Into Three Models and Rewrites Its Licensing Story
Suno's new v6 lineup splits into a precise flagship, an experimental wild variant, and a free mini, all trained with licensed catalog music.
Anthropic Models How AI Could Hollow Out Knowledge Worker Wages by 2030
Anthropic's Economics team released an interactive model projecting three futures for GDP, wages, and jobs by 2030, from modest to extreme AI adoption.
Lovable Turns 800,000 Freelancers Into a Paid Partner Sales Force
Lovable formalizes its 800,000 strong builder community into a two-track program with certifications, a partner directory, and commission on Business subscriptions.
Google DeepMind's Gemini Agents Spontaneously Split Into Cheaters and Whistleblowers
A Google DeepMind case study put 100 LLM agents in a math conference simulation and watched cheating spread, then whistleblowers spontaneously fight back.
Sarvam Teams Up With IDFC FIRST Bank to Build a Self-Improving AI Bank
Sarvam and IDFC FIRST Bank are launching a joint R&D Lab to build banking AI that learns from live operations and improves itself over time.
Intelligent Internet's Meta-Zenith Agent Rewrote vLLM Kernels for a 4x Speedup
An autonomous agent ran 111 trials to optimize vLLM for Qwen3.8-27B on a single RTX 5090, hitting 4.1x throughput at 65K context.
anything2explainer Turns Claude Into a Full Motion-Graphics Film Studio
A new open-source Claude Code and Codex skill turns any topic into a fully coded motion-graphics explainer video with voiceover, subtitles and QC.
FastVideo Ships FastH3 for ComfyUI, Cutting Video Generation to 8 Steps
FastVideo's 8-step distilled MiniMax H3 checkpoint lands as a single-file ComfyUI repack, bringing text-to-video-with-audio generation to local workflows.
Suno Flips its Biggest Blocker, Believe, Into a Global Music Partner
Suno strikes a global licensing deal with Believe and TuneCore, reversing an April ban and opening opt-in training plus distribution for indie artists.
NVIDIA's CUDA Python 1.0 Makes Python a First-Class GPU Citizen
NVIDIA ships CUDA Python 1.0 with semantic versioning, a shared cuda.core foundation, and PyTorch and CuPy already building on it.
Cohere's Open-Source Megakernel Beats vLLM by 1.58x on H100
Cohere open-sourced a serving engine that runs the entire LLM decode step as one persistent CUDA kernel, hitting 1.58x vLLM throughput on H100.
Meta's Muse Agent Books Travel and Negotiates Bills With Your Accounts
Meta's new personal AI agent runs in an isolated cloud VM with a permission gatekeeper called Sentinel, and pays up to $300K in bounties for prompt injection attacks.
HarnessRouter Unifies Codex, Claude Code, and Hermes Behind One Open API
An open-source Docker container runs Codex, Claude Code, and Hermes through one unified API, keeping your keys and data on your own machine.
UT Austin Finds Fractal Chaos Explains Why AI Reasoning Costs Spike 10x
A new paper shows recurrent-depth reasoning models behave like chaotic dynamical systems, where hard problems create fractal basins that trap thinking near wrong answers.
vLLM's Hybrid HiSparse Triples Concurrent Requests on Million-Token Contexts
vLLM's new Hybrid HiSparse keeps long-context requests decoding when the KV cache overflows HBM, tripling concurrency on GLM 5.3 at 1M context.
OpenAI's ChatGPT Images 2.5 Cuts Generation Time by 50% and Fixes Edits
OpenAI's new image model brings 50% faster generation, precision comment-based edits, in-chat sketching, and two new API tiers for developers.
World Labs' Atlas Unifies 3D Reconstruction and Video Generation in One Model
World Labs unveiled Atlas, an omni world model that unifies text, images, video, and 3D into a single 3D-grounded spatial context.
Krea's Realtime Director Turns Text Prompts Into Live Video Controls
Krea's new Realtime Director lets you steer a live AI video stream with prompts as it plays, powered by fal's H3 Max Director model.
OpenAI's 10,000-Agent Swarm Solves a 100-Year-Old Math Mystery
An internal OpenAI model coordinated roughly 10,000 agents for 88 hours to produce a Lean-verified finite-time blowup proof for 3D Navier-Stokes.
Cognition's Devin Hits $48B Valuation as Revenue Nearly Doubles in Four Months
Cognition doubles its valuation in four months as Devin's run rate nearly hits $900M, cementing autonomous coding agents as a real enterprise category.
Inception's Mercury 2.5 Hits 1,107 Tokens per Second, Beating Autoregressive Models
Inception's new diffusion-based LLM hits 1,107 tokens per second on NVIDIA GPUs while boosting quality 40% over Mercury 2.
Cowart Plugin Turns Codex Into a Visual Design Tool With Infinite Canvas
Cowart is an open source Codex plugin from developer Zhong Xin that adds a tldraw-based local canvas for image generation and annotate-to-edit workflows.
Gradium's Voice Design Beats ElevenLabs With 72.6% Win Rate for Free
Gradium's new Voice Design turns a written description into a brand-new synthetic voice in seconds, no cloning or licensing required.
No stories match these filters.