All Stories

Pauliusztin's software production line splits agent work into plan, implement, and review

Nous Research adds cron-backed dashboard scripts with sandboxed paths

The writing-agent repo lays out a Python agent stack from configs to observability

UC Irvine study says ChatGPT made students faster on math but less likely to retain it

OpenTrons OT-2 and Claude Code power a garage Alzheimer's screening workflow

Corey Ganim says AI should filter email and draft replies, not send them

Techhalla shares a Seedance 2.0 prompt recipe for stable 15-second shots

Techhalla's Blender MCP workflow turns viewport scenes into stylized video

Cursor helps build an Electron app that fixes Samsung webcam flicker

willmcgugan's writing app targets distraction-free drafting with typography controls

OpenKnowledgeAI turns papers into visual explainers and markdown notes

NUS survey maps World Action Models from prediction to action

Opus 4.8 forgets names in long Claude Code sessions, hinting at KV limits

Bank of Korea finds AI cuts task time 3.8 percent but not output

Codex helps reproduce a touch-sensing robot paper in a few hours

hf-claude lets Claude Code use over 100 open models

Sebastian Raschka shows how to run local coding agents with open-weight models

Vercel skills CLI auto-matches a goal to the right Claude Code skill

Josef Chen shows routing and voting hit a co-failure ceiling across 67 models

reach_vb says your top five Codex prompts shape how you work

cwolferesearch shows how realistic cyber evals are built

Hamel Husain says binary LLM judges beat Likert scales for actionable evals

Micron says the global RAM crunch should ease by 2028, slowly

NVIDIA's SOLAR automates speed-of-light bounds from PyTorch and JAX code

swyx argues evals should report dollar budgets, not just token counts

TriAttention compacts vLLM KV cache and frees memory from paged blocks

OpenAI plans GPT-5.6 Sol on Cerebras at up to 750 tokens per second

PaddleFormers 1.2 adds DeepSeek-V4 training support and 128K long contexts

Dwarkesh explains why computer use needs grindable simulators

OpenAI wires Codex into Excel and PowerPoint with add-ons

OpenRouter's MCP demo runs a model council for design comparisons

Microsoft ships MAI-Code-1-Flash into GitHub Copilot Business and Enterprise

Google AI Studio says builders shipped 200,000 apps in a month

Conductor 0.70.0 adds multiple run scripts per project

TanStack Model Sets save and reapply model selections across charts

TanStack NPM Stats adds playback, scrubbing, and export modes

A volunteer-built Venezuela needs map turns social posts into response routes

OpenAI previews GPT-5.6 Sol, Terra, and Luna through Codex and API

Claude Code 2.1.195 writes files explicitly and fixes background jobs

TanStack AI MCP Apps turn chat outputs into interactive widgets

thdxr keeps system-prompt facts cacheable with a separate state layer

Magnitude claims 75.5% on Terminal-Bench 2.1 with a grammar-based harness

ComfyUI's Seedance workflow turns Blender motion into stylized video

Manus says KV-cache hit rate is the key production metric for agents

Ornith 1.0 posts 82.4 on SWE-bench Verified, but looks benchmaxed

Kimi K2.7 edges GLM-5.2 on a rotating kebab HTML texture test

GLM-5.2 NVFP4 serves on vLLM and SGLang with lower memory use

Nick Frosst points to a local coding model that fits in 20 GB of RAM

Meta reverses some layoffs and data-labeling reassignments in AI orgs

Agent Arena shows token efficiency diverging sharply across models

OpenAI's internal CTF benchmarks show Sol ahead on efficiency and score

Claude Tag adds proactive memory and identity on top of Claude Code

JV Roig's 172B-token study finds document Q&A still fabricates answers

OpenAI makes Codex Remote generally available with QR pairing

Codex can spin up a VM and hand work off to it with one prompt

Microsoft adds Copilot Skills to Excel for finance workflows

Omarsar maps dynamic workflows as a test-time compute pattern

Midjourney preview adds style-space search and faster batch drafting

Julius turns spreadsheets into video generation workflows

Next.js 16.3 preview adds Instant Navigations and route shells

François Chollet says agentic coding needs explicit API contracts

Google Gemini adds study notebooks with diagnostic quizzes

Vercel uses Linear Agent to guide large PR reviews in code order

BabelTele compresses language to 28% of its length without losing meaning

CRED says AI now writes 90% of its code and is changing team speed

PaddlePaddle PP-OCRv6 lands on Hugging Face with better accuracy

i10X launches Superagent as an AI chief of staff

Higgsfield Supercomputer 2.0 automates campaign creation and monitoring

Manus connects Semrush SEO data directly into promptable workflows

Google's AI Talk Radio agent generates a full talk show in one API call

Dingyi shows an agent recreating a live website from a screenshot

BridgeMind's vibe-coding workflow builds BridgeBench V3

OpenAI makes Codex available in the ChatGPT mobile app with device pairing

Google Research turns memory consolidation into a sleep paradigm

UBS says 60% of firms are curbing AI spend and routing cheaper models

Microsoft MAI-Image-2.5 lands near the quality-price frontier

ElevenLabs shows how to make a Hermes agent talk back

Databricks claims 392 tok/s for GLM-5.2 on Artificial Analysis

Exponential View's State of the AI Economy pegs real revenue at $110B

IBM debuts a sub-1nm nanostack chip with vertical transistors

Tom Greenwald says GLM-5.2 provider speeds peak on launch day then fade

DeepSeek's recruitment graphic shows a broad scale-up across the stack

Nous Research adds a creative-ideation skill that routes prompts through 22 methods

Google AI Studio says users built 1M Android apps

Exa Connect adds agent access to ZoomInfo, Crunchbase, and Similarweb

Eric Xing's Critique of the Agent Model defines agency across five dimensions

Perplexity adds Computer for Counsel for legal research workflows

Agent-native memory paper breaks memory into storage, retrieval, and maintenance

Sakana AI brings Fugu-Ultra to OpenRouter as a multi-model system

Hermes Agent's /learn skill turns docs into a learning plan

Cursor in Notion lets specs turn directly into tasks and PRs

HoloDesktop CLI lets agents operate a local desktop

cwolferesearch defines the agent loop around model, tools, context, and environment

Tetsuwan turns wet-lab work into cloud lab workflows

Manus adds autoscale and reserved hosting modes for published apps

Google Flow Agent grounds image and video generation in Maps Street View

DynamicWebPaige's Chrome extension records browser actions into skills

Google Research finds reasoning unlocks parametric knowledge in LLMs

Portless doctor checks proxy, DNS, and CA trust in one command

OpenAI and Anthropic reportedly store agent traces in ClickHouse

Factory lets teams pick models in Slack and run Droid in different computers

University of Chicago paper says AI use surveys undercount adoption

gpt-realtime-2 demo packs wake words and games into one voice loop

ByteDance's Seedance 2.5 adds a licensed-IP pass for secondary creation

Figma Weave puts 20-plus workflows directly in the design canvas

Huawei's Atlas 950 SuperPOD signals a bigger domestic compute push

Kog open-sources a 2B model that runs at 3,000-plus tokens per second

Arthur Conmy joins Anthropic as Google DeepMind talent keeps leaving

Hermes Agent adds animated pets that mirror task state in the UI

Hermes Agent can run cron scripts without paying for an LLM call

Zed adds an outline panel for jumping through search results

Base MCP ships a Venice plugin for private model inference

Self-Harness lets agents mine failures and revise their own harnesses

Dreamina workflow guide shows how to turn one AI video hit into a repeatable pipeline

Firecrawl becomes an official plugin in Grok Build

ChatGPT for iOS adds host personalities, goal editing, and faster thread recovery

Techhalla's Seedance 2.0 prompt guide keeps style stable across a 15-second shot

Vercel says tsgo makes its frontend monorepo type checks 7x faster

OpenCode lands in Conductor with multi-provider model switching

GL.iNet Comet Pro turns a spare machine into a remote KVM for agent PCs

Codex users can refresh global agents.md from past PRs and sessions

Codex workflow tip stores feedback in a tmp file while complex skills run

gstack packages Garry Tan's exact Claude Code setup into a 114k-star repo

DeepSeek publishes Engram for conditional LLM memory

Vercel redesigns Workflows trace viewer around spans and metadata search

DeepSeek-V4-Pro scores 80.9% on DimAgent and stays close on Terminal-Bench

OpenRouter lists providers for GLM 5.2 and one endpoint hits 125 TPS

BabyVision benchmark says multimodal models still miss tasks kids solve

Seedance 2.0 native 4K benchmark raises the bar for AI filmmaking

React Doctor for Performance turns slow React code into a CLI fix

Gemini Spark beta turns Google AI Ultra into a personal agent

Claude mobile adds Cowork for background tasks and scheduled runs

Box adds HTML preview, editing, versions, and secure sharing

Dexhorthy says agent planning docs should stay outside version control

guinnesschen says skills should be ranked by downstream invocations

guinnesschen says a fully specified prompt saves more tokens than a small one

Google Labs Jules reframes coding-agent evals around useful behavior

OpenRouter detects 1Password and offers to save API keys

NotebookLM makes flashcards editable instead of forcing regeneration

Manus AI turns SEO research into an executable workflow

OpenRouter adds stablecoin payments with Coinbase

supermemory turns long-term recall into an inspectable memory graph

Binary quantization shrinks embeddings and cuts CPU to GPU transfer costs

GLM-5.2 tops DeepSWE v1.1 with a 44% pass@1 score

Qwen distillation turns DeepSeek thinking style into a local orchestrator

pi-ai 2 splits SDK imports into selective modules and a compat path

Notion frames itself as an agent-first workspace in Akshay Kothari interview

Simon Willison ports Moebius image pinpointing to the browser

Apodex 1.0 turns open-ended research questions into evidence-backed analysis

Vercel decouples deploy from release with a feature flag switch

ComfyUI and Beeble AI Canvas blend generated objects into live video

ComfyUI workflow uses annotations to steer motion across a scene

Techhalla builds an action scene for less than a Starbucks coffee

ElevenLabs launches Ads Engine to localize Google, Meta, and LinkedIn ads

AiBattle_ chart shows GPT-5.5-Cyber topping CyberGym and GLM-5.1 leading open-source

Codex loops keep a JetBlue refund chase running until the case closes

rauchg's team tunes a page by optimizing layout, WebGPU shaders, and scripts

Aravind Srinivas says the model is no longer the product, the harness is

Avichawla's BM25 explainer shows why lexical search still matters in RAG

Yacine MTB says reverse engineering is the most verifiable task for models

Techhalla shares a Grok Imagine prompt recipe for consistent music-video cuts

Cursor says Composer got 10 to 20x more compute to train a GPT-sized model

Meta AI adds an Artifacts tab for chats, docs, web pages, and slides

Samsung says ChatGPT and Codex are rolling out to employees at scale

LiteParse parses a SpaceX PDF faster than a screen-record zoom

Cursor's /automate skill turns one task into a trigger-based workflow

Hugging Face gets LocalLaws, a 2.2M-law U.S. legal corpus

ClickUp Brain can now propose dedicated agents for recurring bugs

NickAdobos says let models ask a user question without blocking the loop

Recovery Lab simulates disconnects, evictions, and mid-tool-call failures

Anthropic partner slug suggests Claude Sonnet 5 is close

Dax Raad says OpenCode won by optimizing terminal feel, not avoiding framework work

PAI turns personal health data and notes into a Claude Code life dashboard

OpenAI appears to be preparing ChatGPT Library for Codex

dolphin-summarize starts extracting GLM-5.2 architecture from GGUF too

Pew says ChatGPT use hit 44% as chatbot adoption goes mainstream

jakubkrehel's make-interfaces-feel-better skill passes 30k installs

Emollick says coding agents are too software-brained for knowledge work

LangChain shows how to build a Claude Code-style agent with Deep Agents

Aaron Levie says headless agent traffic needs guardrails, logs, and source truth

BrennerBot turns scientific method into a reusable agent workflow

pratyushmaini says a 30B MoE can train on 10T tokens for about $500K

Together AI and 5C deploy GB300 NVL72 racks for next-gen inference

Epoch's chart says open-weight models trail the frontier by about four months

GLM-5.2 climbs the DeepSwe chart while undercutting Gemini 3.5 Flash on cost