A focused feed of models, agents, research and open-source releases for people building with AI.
Artificial Analysis is exposing the CritPt v2 physics benchmark grading server through its API. CritPt covers 71 composite research challenges built by 50+ active physics researchers across 30+ institutions.
Anthropic released FlashProve , open-source cryptographic proofs that a response came from a specific LLM. Generates proofs for 400B Llama 4 Maverick at 110 tokens/second on a single H100 GPU. Proofs are ~19 MB and verifiable in seconds on an ordinary 24-core CPU without weights.
Shawn Pana open-sourced Phone Harness , a Python toolkit letting agents control iPhones and Androids. Works via iPhone Mirroring plus Apple Vision OCR on macOS, and adb plus the accessibility tree on Android.
Prime Intellect audited AutomationBench, OfficeQA Pro v2, and HLE-Diamond, finding broken graders and reward hacks that distort reported model scores. AutomationBench fixes replaced prose request bodies with JSON Schemas; Opus 5.5 hit 92.15% partial, 70.83% strict.
Topaz Labs released Wonder 4 , a one-click upscale, denoise, and sharpen model. First Topaz model built on a mixture-of-experts architecture that routes enhancements per image region. Combines upscaling, denoising, deblurring, compression artifact removal, and face recovery in a single pass.
Claude users can now auto-connect MCP plugins at login via Enterprise Managed Auth , no per-service OAuth prompts. Admins configure plugins once through their identity provider; access flows from SSO policy. Built on an open MCP extension using ID-JAG and RFC 7523 JWT bearer grants.
GSBR regularizes online RL by Euclidean distance to a base policy's support, not by likelihood. Actions inside the support cost nothing, so the actor can drop suboptimal modes freely. Support region is estimated from base-policy samples using a nearest-neighbor quantile radius.
SwarmCoDe is a co-evolutionary algorithm for jointly designing robot morphology and behavior in heterogeneous swarms. Dynamic speciation via evolved genetic tags removes the need to pre-specify the number of robot species.
Ai2 released StateBench , a benchmark splitting state tracking into four algebraic complexity tiers. Frontier models including GPT-6 Sol, Claude Opus 5.5, and Gemini 3.8 Flash drop to 13% or lower on hardest problems.
Boston Dynamics launched Stretch Gen 2 , a hardened version of its autonomous trailer unloading robot. Peak throughput rises to 1,000 cases per hour with a new multipick feature. Internal cabling cut by 81 percent, roughly a half mile, to eliminate electrical faults.
Google's Project Suncatcher ran Gemma 4 on a data-center TPU in orbit, returning a haiku to Earth. Prototype satellite named MVP, built with Planet, launched Oct 1 on SpaceX Transporter-18. Carries four Trillium-generation TPUs running on roughly a 1 kW solar power budget.
Community release puts MiniMax H3 on consumer GPUs via pruned GGUF quantizations. Checkpoints range from 8.9 GB (Q3_K_M, ~12 GB VRAM) up to 21.6 GB (Q8_0 reference). Two modes shipped: FL2VA first/last-frame-to-video-audio and Ref2VA omni-reference generation.
HoppouAI released Breeze-TTS-2.cpp , GGUF quants plus a C++ runtime for BreezeBlue's 3B TTS model. Runs on NVIDIA, AMD, Intel GPUs via Vulkan, with CPU fallback and no Python required. Hits roughly 1.2x realtime at Q8_0 on an RTX 3060, 24 kHz bilingual English and Mandarin output.
Sarvam released Prism , a 105B decision model under Apache 2.0 that outputs probabilities instead of text. Median latency of 44 ms per request on a single NVIDIA B200, 95th percentile at 420 ms. Scores 71.8 average across three decision indices, claimed highest for sub 100 ms models.
SAPPER-v2 is an open-source multi-agent pentest framework with one AI commander dispatching a worker swarm ( repo ) Won 3rd Tencent Intelligent Attack-Defense online qualifier, 80.18/100, beating 633 other teams running 8 hours unattended Used 1B tokens to tie the runner-up's 85 flags, who burned
U-JiT generates video directly in pixel space, skipping the usual pretrained autoencoder stage. Uses a UNet-shaped transformer with variable patch sizes to handle high-dimensional video input end-to-end. Improves FVD by 27% over prior SOTA on Kinetics-600 frames-to-video at 128 resolution.
Mixed batching with 10% target audio matches full-dataset ASR fine-tuning on three domains. Text-only LLM adaptation creates a modality gap worth ~29% relative WER on Banking. Batches combine source speech, noisy projected tokens, corrupted transcripts, and a slice of target audio.
Mistral released open weights for Mistral Large 4 , a 1T-parameter MoE with 52B active params. Showcase demo: the model autonomously built Orbital Trace, a live 3D tracker of ~19,000 satellites and debris.
Cicada is an open source macOS menubar app for live subtitles, interpreting, and AI meeting notes, available on GitHub . Captures system audio directly, so no bot joins Zoom, Feishu, Tencent Meeting, YouTube, or X Spaces.
MER-Bench introduces Meme Reappraisal: flipping a meme's emotion while keeping scene, characters, and layout intact. Task inspired by cognitive reappraisal in psychology, framed as constrained multimodal generation.
OpenAI is launching a new internal security sprint using its own models alongside external red teamers. Cathedral Systems, the DOGE-alum startup that breached Anthropic, will partner on the offensive testing.
Tiny Corp cut the exabox price from $10M to $5M, with 2 preorder slots left at $100k deposit. GPU config switched from 720 unreleased RDNA5 cards to 256 shipping AMD MI300X accelerators. Specs: 669 PFLOPS compute, 49 TB HBM3, 1.36 PB/s memory bandwidth, 500 kW at 480V.
USC analysis of 58,746 NIH grants finds AI makes up 15.9% of the FY2025 portfolio. AI projects receive a 13.4% funding premium: $675K mean award versus $595K for non-AI work. 79% of AI grants stay in research stage; only 14.7% reach clinical deployment or implementation.
OpenAI unveiled Pando , a Git-compatible backend powering its entire monorepo. Median push latency dropped ~80%, system now serves nearly a petabyte of Git data daily. Engineers and Codex agents keep using standard Git commands with no workflow changes.
No stories match these filters.
