All Stories

DeepSeek releases V4-Flash-Vision-Exp multimodal weights

Qwen3.8-Flash-Next reaches seventh among open models in Agent Arena

GLM-5.3 ranks second among open weights on the Vals Index

Stanford BIT bridges text and images with bidirectional diffusion

Tencent ContextPilot trains agents to edit context proactively

Google releases TimesFM-3 for zero-shot multivariate forecasting

NVIDIA publishes an NVFP4 DeepSeek-V4-Flash variant

Anthropic reportedly signs a $35B Lambda cloud capacity deal

Sliding-window attention outperforms retrofitted linear attention

Inkling reaches top speech-detail recovery with 25 to 100 hours of audio

Franken projects combine drone reconstruction with local sensor fusion

Snowflake Semi-Persistence speeds vLLM model swapping up to 19.9x

Hermes users are urged to set custom user agents amid fraud

Firstmate explains durable records for recoverable agent fleets

Yacine runs rapid reward-shaping sweeps with a phone policy viewer

Reflexio turns user corrections into testable agent rules

Vercel Labs prototypes statically analyzable code mode without VMs

Codex users can schedule heartbeats for long-running tasks

LoopArena measures controllers for long-running coding agents

Amp outlines an agent-assisted production bug-fix flywheel

Google shows Gemini 3.7 Flash powering internal interactive apps

New paper argues agent autonomy can erode human oversight

South Korea prepares free AI services connected to government systems

pMLX selects local model precision from task constraints

RunInfra routes agent sessions back to warm KV cache replicas

CLEAR routes safety adapters only when a prompt needs them

NCCL now supports multi-rank collectives across MIG instances

Prof Tom Yeh ships an interactive self versus cross-attention diagram

Rasbt explains reasoning models and PyTorch setup with uv

Hy4 Preview compresses from 1.5 TB to roughly 200 GiB

Avi Chawla explains why KV caches do not retain queries

Redwood takes an AI-designed accelerator from spec to FPGA

Code as Worlds learns executable simulations for physical reasoning

OpenAI Jalapeño allocates inference resources inside one chip

OpenAI reportedly buys Mac fleets for computer-use RL training

PILOT lets a supervisor improve agents during a live run

OpenClaw teams migrate from local harnesses to a shared agent

Shu Chao argues agent progress depends on better environments

Grok Bot runs Claude Code and Codex inside a persistent computer

PhoneLLM Alpha 1 NVFP4 targets local Blackwell voice agents

DingSim transfers a robot policy to MuJoCo in under two minutes

Geoffrey Huntley advises regularly revalidating agent skills

OpenClaw 2.0 rebuilds its open agent platform

GLM-5.3 reaches 32.4% on Terminal-Bench 3.0

Thinking Machines details its post-training evaluation agenda

No-Mistakes 1.60.2 escalates agent scope creep

Agenttrail visualizes concurrent Claude, Codex and Cursor work

CritICL retrieves small-model failures to steer stronger models

The Turing Post compares ten RAG tools by workload

D4RT encodes video once for queryable dynamic 3D scenes

Higgsfield uses Blender to lock AI-video camera movement

Higgsfield recommends Blender scene previsualization for video

a16z flags 800V DC electricians as an AI data center bottleneck

Hetzner VPS pricing makes a mini PC cheaper after three months

Gemini Co-Scientist reports real-world scientific validation

Tencent confirms Hy4 Preview runs in vLLM

fal previews H3 Max with native infinite scene continuity

GLM-5.3 Flash replaces GPT-5.6 Luna in document workflows

CopilotKit shows how to deploy AI coworkers across channels

ChatGPT workspaces sync plugin marketplaces from GitHub

Microquack runs an emotional droid LLM and synthesizer in-browser

AI translation of runic text demonstrates prompt-injection risk

Humor Arena ranks Fable 5 as its funniest frontier model

Theo calls for analytics products with agent-native traces

Apollo-inspired agent honeypots test reward-seeking behavior

Ramp estimates OpenAI receives 35% of AI token spending

Prof. Tom Yeh walks through a deep RNN calculation by hand

NVIDIA describes Codex as its complex engineering default

Prof. Tom Yeh walks through RLHF calculations by hand

FrankenTTS previews free on-device speech synthesis for Apple devices

FastH3 v1 generates 15-second video in 13 seconds

Puro-2B trains a 2B model on RTX 5090 GPUs for $6.9K

Cerebras explains wafer-scale SRAM advantage for decode latency

FrankenMarkdown shares a profiling-first agent optimization workflow

Sesame releases TurnBench for voice-agent turn timing

mcp-handler npm downloads grow sixfold in six months

Linear turns Gong sales calls into product feedback loops

Cua Cloud Fleets launches managed Omarchy agent sandboxes

TanStack withSkills loads agent instructions on demand

WikiSkill separates agent traces, knowledge, and skills

Grok Bot publishes guides for coordinating bot teams

Vercel Speed Insights adds a free tier for every project

Simon Willison expands the LLM cliché highlighter to 38 patterns

Muse Image launches on the Meta Model API at $0.01 per image

GLM-5.3-Flash retains 93% accuracy at 4-bit

Prof Tom Yeh explains vector databases with hand calculations

Composio compares five open models on multi-step agent tasks

Orange Pi Zero 3W runs stereo depth in under 60 ms

vLLM benchmarks speculative decoding methods on AMD GPUs

Baseten reports 122-plus TPS for GLM-5.3-Flash

vgpu offers WebGPU primitives for agents across runtimes

Code Mode Review Council scales human reviews of RL task data

Vtrivedy outlines a practical GRPO signal-building workflow

Claude Managed Agents gets a Vercel Chat SDK cookbook

PhoneLLM releases a low-latency voice agent model

Google Research introduces a planetary prediction engine

ChatGPT Voice can drive an always-on remote desktop

Temporal workflows can stage deterministic agent work

JIT-Agent generates and repairs task-specific agent harnesses

EvoMal shows coding-agent skills can self-propagate malware

Sonic 3.6 TTS reaches general availability

Grok-4.6 ranks fifteenth overall on Agent Arena

FrankenWhisper ships private on-device transcription for iOS

DiffsHub adds local-token access for private GitHub PRs

Ofir Press revisits three rules for useful LLM benchmarks

Agent traces collapse into compact finite-state machines

VS Code explains independent editor and terminal font sizing

Casey Muratori advocates hardware-first performance engineering

Sentence Transformers details ColBERT-style medical retrieval tuning

Pi lets users constrain and cycle model shortlists

OpenCode proposes retention as a model quality metric

Nvidia reportedly agrees to buy Hugging Face for $12.9B

Jacob Peake maps the tradeoffs across AI chip architectures

Dwarkesh maps the compute pricing advantage of major AI labs

X Chat gives developers programmatic agents with public handles

Composio bridges Grok Bot to more than 1,000 integrations

Cua ports Omarchy to Apple Silicon with agent desktop control

Runme combines WebMCP notebooks with bring-your-own agents

DeepLearning.AI teaches agents to turn traces into reusable skills

Flint AI's Switch puts cross-framework agents in shared channels

VS Code Agent Host keeps agent sessions persistent across clients

Lovable connects its workspace to Google's Antigravity agents

Artificial Analysis details a reward-hacking judge for Terminal-Bench

OpenCode agents combine observability, payments, and databases for fraud triage

Higgsfield documents one-photo video swaps with Seedance 2.5

Claude Code drafts feedback reports after failed tasks

Linear reports agents in 95% of paid workspaces

Wan 3.0 leads a practitioner test of raw video editing

Meta's Project OT explored replacing teams with AI-native workflows

Codex visualizer improves cloud coding-session progress views

Mac Studio outpaces clustered DGX Sparks on generation bandwidth

Cerebras plans 3D DRAM for its wafer-scale inference system

Model Discovery Agent combines LLM proposals with Bayesian design

South Korean labs climb Artificial Analysis intelligence rankings

Makora launches on OpenRouter with DeepSeek, GLM, and Kimi

Pi exposes custom slash command registration for agents

Jakub Krehel adds UI stress-testing and variant skills

Vercel AI Gateway sends webhooks after long generations

Factory Droid connects agents to 50-plus enterprise tools

Google AgentHands syncs XR gestures to spoken agent guidance

Grok Bot workforce checker monitors stalled multi-bot work

MiniMax M3 completes email workflow for $0.018

ChatGPT extension passes open browser tabs into Codex tasks

Agent work design framework maps eleven firm-worker configurations

OpenCode says Stripe blocked $300M in card fraud in days

Cline reports Stripe blocked $8.4B in recent abuse

Simile argues query-level confidence is central to simulation

DeepLearning.AI maps skills needed for production AI systems

Claude unifies memory across Chat and Cowork with controls

Tailwind StyleX exposes typed Tailwind tokens in StyleX

Ramp AI Index shows paid AI spend remains dominated by chat