Wrap
LangSmith LLM Gateway: runtime controls for production agentsRuntime controls on the LLM gateway for production agents — wrap/ship, not a training launch.
LangChainJul 30Editorial
Setting type
Agentic Firehose
Pulling labs, repos, and the catch-up path…
Harnesses, graphs, loops, APIs you call this week — tagged with Ng’s application skills. Not a training desk. New here is the decoder. Starting a project this week? Catch up is the four-skill map — not more news.
No splash for Production. The lanes below still scan.
Cookbooks for building AI apps: context, harness, graphs, cloud.
Wrap
LangSmith LLM Gateway: runtime controls for production agentsRuntime controls on the LLM gateway for production agents — wrap/ship, not a training launch.
LangChainJul 30Editorial
Wrap
Introducing the Agents API — managed cloud agents on the Codex harnessCallable managed Agents API for orchestration, long-running sessions, and tool use — swap DIY agent loops for a hosted Codex harness path when wrapping/shipping cloud agents.
OpenAISep 10Editorial
Wrap
Gemini 3.8 Flash on Vercel AI GatewaySame-day Gateway route for Next.js / AI SDK apps already on Vercel.
VercelSep 2Editorial
Wrap
DeepSeek V4.1 Flash now available on AI GatewayOne gateway model id for DeepSeek V4.1 Flash — A/B a cheap fast tier against Astra/Fable/Gemini on the same AI SDK router without a separate DeepSeek account.
Vercel ChangelogSep 9Editorial
Wrap
Claude for Commerce AgentsReference tool/action surface for shopping and checkout agents you'd wrap around an existing commerce API.
ClaudeSep 2Editorial
App
Next.js App Router: a streaming AI app with the AI SDKBuild the chat UI and the route handler this week. Cloud lab for shipping an application, not wiring a network.
Vercel AI SDKEditorial
Cloud
Deploy a Python ADK agent to Cloud Run this weekGCP lab: Vertex/ADK as an agent app on Cloud Run. Ship the product, not a TPU blog.
Google CloudEditorial
Harness, loops, graphs, MCP, permissions, evals as acceptance.
Harness
Managed agents: decouple the brain from the handsAnthropic’s engineering note on splitting the planner from the tools is the production version of a harness. Same idea whether you like Claude or not.
Anthropic EngineeringEditorial
browser-use’s Mac harness is the app on the laptop in front of you: Accessibility, a computer-use loop, tool permissions.
browser-use/macos-harnessEditorial
BYOH: MCP and CLI into the loop you already run. An embedded chat agent will not have your skills, memory, or local tools.
yakko.devSep 1Editorial
One command turns a stdio MCP server into a Streamable HTTP URL with OAuth 2.1. Demo the harness without deploying.
mcptunnelsEditorial
Eval you attach to the wrap: flag agent mistakes on every thread without standing up your own judge.
LangChainAug 18Editorial
Org-provisioned HTTP/SSE MCP for every user — harness permissions, not filesystem MDM.
Claude CodeEditorial
MCPAdapter + elicitation → LangGraph interrupts for app-side agent graphs.
langchain-ai/langchainEditorial
Builder glossary for harness/loop language FDEs ship with.
GitHub BlogEditorial
Redesign tool surfaces for how models call them — harness craft, not an IDE feature list.
Agent ConnectEditorial
Harness
Cursor joins the AI SDK harness layerSwap Cursor, Claude Code, Codex, and Pi behind one HarnessAgent interface — coding-agent wrap without rewriting the app.
VercelAug 27Editorial
Treat agent/env/skill/memory as IaC: describe resources in-repo, run ant apply, commit claude-lock.json so CI updates the same resources instead of recreating them.
Claude API release notesSep 3Editorial
Faster cold paths for coding-agent / cloud-agent sandboxes — ship agent loops that spawn remote sandboxes without the previous routing tax.
Vercel ChangelogSep 8Editorial
Server-side auto allow/deny/pause for agent + MCP tool calls (evaluation field on events) plus terminal attach via ant beta:sessions connect — ship Managed Agents harnesses with live approval loops.
Claude Platform release notesSep 10Editorial
How agentic coding blew up CI and how Anthropic scaled test-impact analysis
Claude BlogSep 14Editorial
Gotchas moving huge agent preprompts onto self-hosted Ollama
patrickmccanna.netSep 14Editorial
Clone-to-ship: RAG, eval harnesses, agent wrappers. Not train-a-model.
Repo
mksglu/context-modeSandbox tool output / shrink context (~98% claimed) for coding agents — drop into harnesses to cut token cost before the LLM
GitHub19m ago30k ★Live
Repo
addyosmani/agent-skillsClone a curated skill pack for coding agents — drop production engineering skills into Claude Code/Cursor/Codex-style harnesses instead of inventing prompts from scratch
GitHub4d ago108k ★Live
Watchlist posts for people building AI apps. Originals, not architecture notes.
X
Open Code Review: deterministic pipeline + agent only for reasoningCoding-agent craft: Git diffs → review changed files → codebase context search — agent where judgment helps, not everywhere.
Tech with Mak / XSep 14Live
X
8 questions before the next AI release: what eval blocks prod?Eval/prod craft: one good metric, unacceptable failures, who approves — shipping weekly without answers is release-note addiction.
Suraj Sharma / XSep 14Live
X
AI product unit economics: $ per successful task, not tokensShip craft: margin after model+infra+human review, cost of failed runs, break-even requests/day — vanity token spend is not a product.
Suraj Sharma / XSep 14Live
Harness
How to design an agent harness: work inside the system of recordAgents share data and policy with humans in one continuously updated view, with an audit trail. Harness design, not a valuation hook.
Yarchi / XAug 27Editorial
The Anthropic plugin for llm now tracks SDK 1.x. Upgrade the wrap layer this week so Claude calls keep compiling.
Simon WillisonAug 24Editorial
Drew Breunig: model cost stopped papering over a weak harness. Context and routing are the job again.
Drew BreunigAug 23Editorial
Simon Willison’s CLI upgraded to OpenAI Python 3.x and httpx2. The adapter you call through, not a cookbook and not a new model.
Simon WillisonAug 22Editorial
Course
Become an AI engineer: LLM apps, APIs, context, tools, deployWrap, RAG, tool-use, evals, then ship. A six-month app-builder path, not a lab-training syllabus.
Rahul / XAug 11Editorial
Stack
Vibe-code stack: Next.js, Vercel, Convex, Clerk, StripeSoftware fundamentals for shipping an AI app this week. The wrap layer, not an audience-growth thread.
Alex Finn / XAug 4Editorial
FDE
No-BS FDE guide: wrap and ship in the customer’s stackForward-deployed craft: you are in the account building the AI app. Not a salary-in-30-days thread.
Rahul / XAug 1Editorial