
TripoAI's TripoSplat Turns One Photo Into a Rotatable 3D Asset With 262K Gaussians
VAST-AI open-sources a single-image to 3D Gaussian generator with learnable density control, adjustable output size up to 262K primitives, and SIGGRAPH 2026 backing.
20 stories

VAST-AI open-sources a single-image to 3D Gaussian generator with learnable density control, adjustable output size up to 262K primitives, and SIGGRAPH 2026 backing.

Sakana Chat adds its orchestrator model Fugu, a refreshed Namazu, and in-browser Python execution — making frontier-grade AI accessible without an API

Princeton researchers unveil AQuA, a two-part LLM agent framework that iteratively discovers trading factors and evolves models, hitting a 2.50 Sharpe on US equities.

A controlled benchmark of 10 LLMs against 26 embedding models on 37 tasks finds parity in quality but a 1,431x cost gap.

Cerebras Q2 cloud revenue surges 281%, beats estimates, lands OpenAI and AMD deals — but stock drops 15% after hours

Claude in Chrome's side panel now runs a full Cowork session, syncing conversations, skills, and connectors across desktop, web, and mobile
Arena.ai analyzed tens of thousands of Claude outputs and found Opus 5 writes 3x longer, uses 2.3x more em dashes, and has grown structurally complex but lexically simpler than Opus 4.5.

Higgsfield's Cinema Studio 4 adds 30+ camera presets, dual-source lighting, 50+ color palettes, and granular emotion controls for AI-generated film production

Zed's new standalone app turns agent coding sessions into shared, reviewable, real-time multiplayer threads backed by a CRDT-based version control layer

SpaceXAI's Grok 4.6 matches Claude Fable 5 on agentic knowledge work benchmarks at a fraction of the cost, with API pricing starting at $2/M input tokens.

Upstage's Solar Pro 4 jumps 28 points on the AI Intelligence Index, with massive gains on agentic tasks, long-context work, and hallucination control

Google Research finds frontier LLMs like GPT-5 and Gemini-3 encode up to 98% of facts, but recall failures — not missing knowledge — are what drive factual errors.

Google's LiteRT runs Gemma 4 E2B on a $175 Raspberry Pi 5 with voice, vision, and robotics entirely offline at 300 words per minute

Cohere releases a free 2.4B vision-language model that reads documents at full resolution, beating larger rivals on DocVQA and grounding tasks

Google adds 14 new MCP-powered app connections to Gemini, spanning music, travel, health, and productivity, as the assistant crosses 1 billion monthly users

Grok 4.6 matches GPT-5.6 Sol on composite benchmarks at half the price, with a big focus on long-running coding agents

Google DeepMind's SL2T model brings real-time ASL-to-English dictation to Gboard and Live Transcribe on Pixel 11, trained on 100,000+ hours across 50+ sign languages.

Liquid AI's 3B vision-language model beats models up to 2.6x its size on screen understanding and grounding, running on-device at 20 tokens/s on a phone.

Sarvam's Indus platform goes live for 2,500 Maharashtra officials, marking India's most ambitious state-level sovereign AI deployment to date

Hands-On Modern RL is a free, open-source curriculum that takes you from CartPole all the way to GRPO, RLHF, and agentic systems — with runnable code at every step.