feed
last scrape: 3h ago · 961 in archive · 9 today
this week
Breakdown of Codex agent scaffolding: prompts for one-offs, skills as markdown "house styles" for reuse, plugins packaging full workflows with integrations, plus MCPs/hooks/scripts—decision tree in the article.
AI Agents Need Scaffolding: Prompts to Plugins
Walkthrough of installing vidIQ MCP connector in Claude Code to audit channel metrics (VPH trends, outliers), compare to competitors like Nate O'Brien, spot title/thumbnail gaps, and prompt a live analytics dashboard.
vidIQ MCP enables Claude to audit YouTube channels
Demo of hooking up vidIQ MCP to Claude Code's desktop app for a self-audit of the creator's channel: views-per-hour trends, competitor benchmarks, title/thumbnail gaps, and prompting a simple dashboard to track metrics like subs and watch time.
Claude + vidIQ MCP Audits YouTube for Growth Leaks
YC Founder Firesides chat with Trigger.dev co-founders on three product pivots: v1 "Zapier for devs," v2 embedded async tasks, v3 hosted SDK execution—hitting PMF as AI agents drove 90% usage.
Trigger.dev Pivots to AI Agents, Hits PMF with 90% Usage
YC Founder Firesides interview with Trigger.dev co-founders, who recap pivoting from a "Zapier for devs" background job SDK (v1-2, self-hosted) to a hosted platform (v3) for reliable async execution—now mostly powering customer AI agents after YC and a Series A. They cover PMF lessons, use cases, open-source strategy, and bets on checkpoint/restore compute.
Trigger.dev: Async Infra Powers 90% AI Agents
Walkthrough of Cursor's /goal command best practices: enable via /features enable goal, prompt with explicit verifiable "done" states (e.g., Playwright visual checks, quantified targets like "20 issues"), initial alignment chats, and scaffold via npx goalbody for goal.md + state.yaml files.
Codex /goal tips: Define verifiable 'done' states
Walkthrough of Codex's /goal slash command for long-running agent tasks (like code migrations), with setup steps, prompting templates emphasizing verifiable "done" states and pre-task alignment chats, plus intro to the open-source Goal-buddy prompt builder and /mission alternative. Links workshop for deeper best practices: aibuilderclub.com.
Master Cursor /goal: Fix Premature Stops on Complex Tasks
Walkthrough and demo of Symphony, OpenAI's Elixir tool for spinning up Codex agents on Linear issues — highlights prompting an agent with a 2k-line spec to build a custom version (e.g., Python), plus hooks for repo cloning/PRs. Compares to MultiOn/Conductor but skips real-project readiness details.
OpenAI's Symphony: Autonomous Codex Agents for Linear Issues
Walkthrough of Symphony (announcement), OpenAI's open-source tool for letting coding agents autonomously claim and complete Linear issues in isolated workspaces. Demos basic setup via repo clone or agent-built from a 2k-line spec, plus hooks for PRs, and calls it barebones vs. MultiOn or Conductor—not production-ready yet.
Symphony: Agents Autonomously Claim and Complete Tasks
Verdent Manager overview: coordinates AI agents by decomposing ideas into parallel tasks with long-term memory of your stack/preferences, specialist skills for testing/deployment, and Slack/Telegram messaging — invite-only, with Eco Mode/BYOK for costs.
Verdent Manager Coordinates Idea-to-Deployed App Builds
Overview of Verdent Manager, a coordinator layer for Verdent's AI coding agents that decomposes ideas into parallel tasks, applies long-term memory and specialist skills (testing, deployment), and supports Slack/Telegram messaging plus Eco Mode/BYOK for costs. Speaker explains features without demoing app builds.
Verdent Manager: AI CTO Builds Apps from One Prompt
Casual "Let It Cook" live stream where TypeScript PM Daniel Rosenwasser and James Montemagno demo VS Code insiders, LSP tweaks, agents, and early TypeScript 7 previews while building web stuff—not 7 complete sites.
TS7 Native Go Port Delivers 10x Faster Type Checking
Casual live stream on the VS Code channel where TypeScript PM Daniel Rosenwasser joins host James Montemagno to chat updates, demo insiders features like browser attachment and open telemetry, then build websites using early TypeScript 7 and LSP in VS Code.
TypeScript 7 Native Preview: 10x Faster Web Builds
Walkthrough of Printing Press, a Go-based CLI factory/library with 50+ pre-builts (ESPN, Craigslist, etc.) that agents like Claude Code invoke token-efficiently vs MCPs (35x fewer tokens, 100% vs 72% reliability) or raw APIs. Demos setup and custom CLI for Skool.
Printing Press: CLI Factory for AI Agents
Demo of Printing Press, a Go-based CLI factory (factory repo) and library (library repo) for AI agents in Claude Code. Explains CLI benefits over APIs/MCPs (token efficiency, no rate limits), shows starter pack install/setup, and demos building/sharing custom CLIs for sites like Skool.
Printing Press: CLI Factory for Token-Efficient Agents
Walks through pasting three specific prompts into Claude Code after adding the Higgsfield MCP connector: generating product images/videos from a Shopify URL, counter-objection UGC ads from scraped Amazon 1-star reviews (via Claude Chrome extension), and consistent AI influencer shots in various locations. Prompts/markdown files are free in the description.
Claude + Higgsfield MCP Builds 3 Agency Ad Tools in One Session
Weekly AI podcast recaps essays from Ezra Klein/Alex Ess/A16Z on relational job growth over apocalypse fears, plus Anthropic/OpenAI enterprise JVs and Wall Street data center bets as maturation signs.
AI Shifts: No Job Doom, Infra Boom Ahead
Narrated walkthrough of OpenAI's announcements on real-time voice models (GPT-Realtime-2/Translate/Whisper for conversations, translation, transcription), MRC supercomputer networking, and Sam Altman's AI layoff comments, with a CodeRabbit sponsor segment.
OpenAI's Real-Time Voice AI Powers Agents, Backed by MRC Networking
Demo and benchmark of oMLX on M2 MacBook Pro: setup, dashboard tour, Qwen 3.6 coding task via Codex CLI (with SSD-cached context recovery), vs. LM Studio—faster tokens/sec (47 vs. 16), better multitasking, but occasional context limit errors.
oMLX: 3x Faster Local LLMs on Apple Silicon via SSD KV Cache
Demo of oMLX using MLX's Two-Tier KV cache to page inactive context to SSD on M2 MacBook Pro: 47 t/s and 89% cache efficiency with Qwen 3.6 (vs LM Studio's 16 t/s), enables multitasking, but hits occasional 400 context errors.
oMLX SSD KV Cache Enables 3x Faster LLMs on M2 Macs
Interview with Anthropic's Claude platform leads Angela Jiang and Katelyn Lesse on the new Managed Agents: how it evolved from basic APIs to cloud-wrapped, scalable agents with file systems, skills, and tools like code execution; production hurdles; internal uses; and visions for self-writing harnesses. Follow-up to their Code with Claude event announcement.
Claude Managed Agents: Scalable Path to Production AI Agents
Podcast interview with Anthropic's head of Claude platform product (Angela Jiang) and engineering (Katelyn Lesse), unpacking Managed Agents' primitives like messages API/tools/file systems, production infrastructure hurdles, internal use cases, and a future where Claude self-configures its harness.
Claude Managed Agents: Scalable Path to Production AI Agents
Interview with Claude platform leads Angela Jiang and Katelyn Lesse on Managed Agents primitives (messages API, code execution sandbox, file systems, skills), production scaling hurdles, and a future of outcome+budget goals with self-writing harnesses.
Claude Managed Agents: Production-Ready AI Infra from Anthropic
YC Founder Firesides chat with Conductor co-founders on trying 12 ideas before building their Mac app for parallel coding agents, 10x growth, $22M raise, and Cloud launch.
Conductor Founders on Pioneering Multi-Agent Coding
YC Founder Firesides interview with Conductor co-founders Charlie and Jackson: they recap pivoting through 12 ideas in YC, building their Mac app for parallel AI coding agents (Claude/Codex on codebase clones), 10x growth, $22M raise, and today's Conductor Cloud launch for persistent runs.
Conductor: Multi-Agent Coding Tool Founders Reveal YC Pivots