All Stories

Browser Use CLI 3.0 adds direct CDP control and cloud browser support

Vercel Service Bindings add private internal URLs between services

Claude Code 2.1.199 fixes background agents, partial outputs, and stacked skills

Stanford AutoMem treats agent memory management as a trainable skill

Meituan LongCat 2.0 demo turns one business question into an AI SQL report

OpenAI documents the Codex App Server protocol behind rich agent clients

AI CLI adds terminal model metadata with context, pricing, and JSON output

Stas Bekman turns his dense Python cheatsheet into a free cookbook

Dabit publishes Cloud Stacked Diffs skill for agent-delivered PR stacks

Jason Zhou shares skills for cloning sites and extracting WebGL shaders

dzhng shares a skill using Fable for planning and Codex for implementation

Matt Pocock shows claude --bg for spawning named background sessions

pvncher uses Codex side chats to generate steering prompts for main tasks

LlamaIndex maps the modern agent stack from RAG hacks to goal definitions

Base44 builds a WhatsApp and iMessage superagent skill for blind users

Kent Beck says software engineering builds trust, not just code output

BridgeMind calculates Fable 5 weekly quota can vanish in 3.3 sessions

Will Depue sketches requirements for a trusted executive super assistant

Anthropic says internal Claude Tag lands 65 percent of product PRs

Factory adds GLM 5.2 Fast to Droid with Fireworks hosting

Baseten and Cursor explore MLP compression for long-agent KV caches

Code Arena adds fullstack app generation with databases, API keys, and logins

François Chollet says ARC-AGI-3 contenders use symbolic world modeling

Ethan Mollick uses Fable, Unity, and MCP to generate a playable FPS twist

Matt Pocock renames /to-prd and /to-issues as spec and ticket skills

ARC Prize names the first 2026 ARC-AGI-3 Milestone Prize winners

Avichawla explains A/B, canary, interleaved, and shadow ML testing

EdgeBench frames agent learning as a log-sigmoid law across tasks

DeepSWE table shows Sonnet 5 costing more while trailing Fable and GPT-5.5

Scaling01 compares reward-hacking rates on DeepSWE and SWE-Marathon

Quinn Slack says banks are fast-tracking open models after frontier restrictions

BridgeMind reruns Fable 5 and finds guardrail routing crushes BridgeBench scores

Satya Nadella says enterprise AI moats come from private learning loops

UK AISI finds larger token budgets stretch cyber task horizons

OpenAI reportedly floats a 5% U.S. government stake to ease pressure

Anthropic reportedly explores custom AI chip manufacturing with Samsung

openPangu-2.0-Flash specs show a 92B MoE trained on Ascend 910B

Google and Anthropic move retrieval into standalone agent services

Anthropic previewed Claude Science as a local research desktop app

Vercel open-sources konsistent to enforce TypeScript conventions

OpenUI Cloud adds production guardrails for generative UI

Codex uses computer use to improve OpenClaw iOS from Twitter feedback

Vercel adds dry-run deploys so agents can catch issues before shipping

Gemini Omni Flash and Seedance workflows edit videos with prompt-level control

ctatedev says agentic coding and review is going chatless

fal's TRELLIS.2 turns 3D asset folders into LoRA adapters

lgrammel shows AI SDK tools running inside local Docker sandboxes on Mac

v0 adds Nano Banana 2 Lite for image generation in the prompt bar

Hermes Agent clears every P0 and P1 issue after an all-nighter

OpenWiki keeps repo docs up to date for agent consumption

SkillBench profiles agent sessions to find token-efficiency bottlenecks

QuiverAI's Arrow 1.1 generates structured SVG assets through Vercel AI Gateway

MCP Server Architecture Patterns paper maps five server shapes and four anti-patterns

Anthropic says Claude now writes more than 80% of merged code

Pauliusztin's context layer diagram separates memory from harnesses

LangChain shows recursive language model workflows inside Deep Agents

OpenCode 2.0 hotreloads skills without busting cache

Hermes Agent v0.18.0 ships a Judgement Release and new updater

Natalia Zarina says Codex built her a custom CRM overnight

Bloome runs Claude Code, Codex, and DeepSeek in one group chat

LiteResearcher pitches scalable agentic RL for deep-research tasks

Perplexity CEO says power users can burn $10K a month in agent loops

Atomic.chat physics contest shows Fable 5 leads, but GLM 5.2 is far cheaper

Databricks tops NVIDIA's SOL-ExecBench using agentic kernel-design loops

cwolferesearch highlights practical KV cache compaction papers

Meta shows quantized reasoning models overthink correct answers

Eve uses OpenUI to turn chat output into editable dashboard UI

Neural Procedural Memory stores agent know-how as activation steering vectors

DeltaDB records agent edits as ordered deltas instead of commit snapshots

OSWorld2 benchmarks computer-use agents on long-horizon desktop tasks

Spellbook launches autonomous contract management across the full deal lifecycle

Portless lets one local proxy serve multiple TLDs for dev testing

OpenClaw says 5% of production traffic runs on Pi and 5% on OpenCode

Matt Pocock packages Martin Fowler code smells into a /review skill

Jason Zhou gives a coding agent a repo map and richer grep hooks

Anthropic frames managed agents as a stable layer over changing harnesses

Acti turns the keyboard into an action layer for agent workflows

Matt Pocock's /wizard skill turns third-party setup into an interactive CLI

MCP Server Architecture Patterns paper maps five recurring server shapes

Paper adds tokens, pen tool, cursor chat, folders, and Figma paste

fofrAI publishes an Omni API skill for video editing workflows

Linear adds triage intelligence that labels bugs and opens investigation PRs

Higgsfield launches Seed Audio 1.0 for voice changes and dubbing

Sebastian Raschka publishes a 440-page book on reasoning models

Ramp and Revelio find heavy AI adopters grow headcount 10% in two years

Etched comes out of stealth with $800M and inference-first chip claims

Re-llax proposes a fully non-blocking sync algorithm for recurrent attention training

Spotify says 73% of PRs are AI-assisted in a 20M-line monorepo

OpenCode v2 adds a shell manager for backgrounded subagents

ComfyUI's Comfy MCP exposes saved workflows to agents

VS Code and TypeScript team up to ship TypeScript 7

Base44 launches Base 1 as its first web-app generation model

Google AI Studio adds full chat history when sharing apps

Cursor launches an iOS app for cloud agents and remote control

ClinePass bundles open-weight coding models behind one monthly plan

Hermes Agent moves session DB writes off the event loop

IRIS packages Gemini Live plus Hermes Agent into a public repo

Zenith harness pushes open models above frontier baselines on hard tasks

Hamel Husain says hard-to-eval AI products are usually a design smell

NVIDIA's HORIZON turns hardware design into repository-level agent work

Warp adds custom routers for task-based model selection

Pauliusztin uses an LLM wiki to design a coding harness

Dexhorthy says to layer small agent loops onto existing systems

LangChain shows how to turn a LangGraph agent into a voice agent

LlamaParse's Retrieval Harness gives agents hybrid search and file tools

JetBrains Air brings Codex, Claude, Gemini, and Junie to Windows

Candidly reads support chats live and steers resolution in real time

Gemini Enterprise adds an Inbox for review, in progress, and done work

Cursor internal screenshot shows GPT-5.6 Sol preview with 1M context

OpenAI Codex v0.142.4 adds GPT-5.6 Sol, Terra, and Luna to model picker

H company opens its computer-use agent API in beta

T3 Code mobile lets one app steer work across multiple computers

Google's Paper Assistant Tool automates scientific review

OpenHands shows a sidekick pattern that cuts LLM spend

Pauliusztin's raw, wiki, outputs folders bootstrap a second brain

Qwen maps the verification horizon for coding-agent rewards

LangChain starts rolling out Trace Judge for cheaper agent trace checks

thdxr slices bare metal servers into per-engineer VMs for agent work

vLLM wires in DSpark speculative decoding and pushes throughput higher

DAIR AI's agent reading list packs 11 core papers into one bookmark

Codex desktop shows all remote sessions on mobile and other devices

PlanBench-XL tests agents across 1,665 tools and hidden dependencies

BrowserBC turns browser sessions into reusable skills and lifts web scores

Accenture's Memex RL stores long-horizon agent work in indexed memory

Towards an Agent-First Web adds identity, purpose, and agents.txt to requests

Paper adds selected text layers as prompt presets for image generation

Instructor 1.15.4 ships a maintainer sweep of parsing and doc fixes

Mark Chen says OpenAI models are nearing self-sustained research

OpenAI adds GPT-5.6 to Codex and Amazon Bedrock

OpenAI tests a Codex effort slider and reworks real-time voice support

OpenCode signs up for Telnyx by following agent-specific onboarding steps

ChatGPT's plugin menu exposes Agent mode, Netlify, Supabase, and Slack

Think orchestrates Claude Code in Cloudflare sandboxes with zero API keys

LangChain ties DeepAgents, sandboxes, evals, and model integration together

Gergely Orosz says hosted open models vary by provider and quantization

BIS warns AI capex and debt could become the next credit shock

Anthropic says the real breakthrough is cloud AI wired into internal systems

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