|
Ex-NASA dev reveals his Agentic Engineering Workflow
David Ondrej
Dexter (Dex), a former NASA developer who coined the term "context engineering" and built an AI software factory that ran for four months autonomously, argues that most developers skip a critical step he calls "program design" — defining measurable goals, architecture, and even the expected call stack *before* letting an agent write code. He warns that completely removing humans from code review leads to serious problems, as agents accumulate "slop code" that neither they nor a neglectful developer can debug when hard bugs appear. His recommended workflow progresses through four stages — product definition (with measurable success metrics), system architecture, program design, and only then agentic implementation — with the core principle that a small upfront investment in clear direction dramatically reduces costly fixes and review time later.
|
|
OpenAI Just Announced The New Plugin Standard for AI Agents
The AI Automators
OpenAI og en bred koalition af AI-virksomheder – herunder GitHub, Cursor, VS Code, ChatGPT og nu Google – har annonceret en ny åben standard kaldet "agent-plugins", der gør det muligt at genbruge færdigheder og MCP-servere på tværs af mange forskellige kodningsagenter. Standarden består af en simpel mappestruktur med en `plugin.json`-fil, en `mcp.json`-fil og en færdighedsmappe, som udgør en bærbar kerne, mens klientspecifik funktionalitet som kroge håndteres separat. En bemærkelsesværdig fraværende aktør er Anthropic, hvilket potentielt kan føre til et splittet økosystem, hvor Claude-brugere skal forholde sig til ét format og alle andre til et andet. Videoen demonstrerer praktisk, hvordan eksisterende Anthropic-plugins kan migreres til den nye standard via Codex, og understreger, at standardens succes afhænger af bred implementering.
|
|
Why Crypto Is Betting Big On AI Agents
CNBC
Major crypto companies including Coinbase, Kraken, and Circle are pivoting to target AI agents as their next wave of users, betting that autonomous software will drive demand for crypto wallets, stablecoins, and blockchain payment networks. Unlike humans, AI agents operate entirely online and require always-available, programmable payment infrastructure — making stablecoins and smart contracts a natural fit for machine-to-machine transactions that traditional banking systems weren't designed to handle. Companies like Coinbase are already allowing AI agents to hold their own crypto wallets and autonomously execute trades, while Circle is positioning its USDC stablecoin as the payment layer for this "agentic economy." The strategy represents a fundamental shift for an industry that has long struggled to find real-world utility beyond speculation, with supporters arguing that AI-powered tools could also democratize sophisticated financial capabilities for everyday retail investors.
|
|
The AI Agent Every Company is About to Build | Vercel CEO Guillermo Rauch
Riley Brown
Vercel CEO Guillermo Rauch discusses how companies are building internal AI agents to run business operations more efficiently. He describes Vercel's own internal agent called "V," which nearly 1,000 employees access via Slack to handle tasks like customer support, data analysis, content creation, and knowledge management — with sub-agents delegating to specialized tools like Codex or V0. Rauch argues that the future of competitive advantage lies in companies building, tuning, and governing their own custom agents (rather than relying on off-the-shelf tools like ChatGPT), and that a "god agent" orchestrator model — routing requests to specialized sub-agents while enforcing data access controls — is the right architecture. He also introduces Vercel's open framework "EV," designed to make it easier for any company to build and deploy their own internal agent with proper security, identity management, and infrastructure.
|
|
This FREE Open-Source Tool Gives Your LLM Internet Access
Sharbel A.
Agent Reach is a free, open-source tool with over 61,000 GitHub stars that gives AI agents access to internet sources they normally cannot reach, including X (Twitter), Reddit, GitHub, and niche communities. Unlike standard LLMs that rely on outdated training data or hallucinate search results, Agent Reach allows agents like Hermes or Claude Code to pull live, real-time information from these platforms using cookies or logged-in sessions. The video demonstrates three practical workflows: creator research (finding real video ideas from live community discussions), tool research (vetting AI tools by reading actual user complaints on GitHub and Reddit), and business research (identifying emerging demand before it appears in polished reports). The creator cautions users to start in read-only mode and use burner accounts, as aggressive automation can trigger platform bans — a lesson he learned firsthand when his own X account was suspended after he posted about the tool.
|
|
Marketing Agents Masterclass (GROW your startup)
Greg Isenberg
In this episode, Greg Isenberg and guest Cody Schneider walk through how to build two AI-powered marketing agents focused on cold outbound via email and LinkedIn. The core strategy involves using Apify to scrape LinkedIn post engagements from niche influencers, identifying "hand-raiser" signals (likes/comments) to find warm leads, then running a "waterfall enrichment" through tools like Kitleads, Apollo, and Origami to find verified email addresses and phone numbers. Cody also covers the full infrastructure stack needed—including email inbox setup with burner domains (via HyperTide or Instantly), email verification with Million Verifier, and deploying the entire system as scheduled code (cron jobs) in the cloud so it runs autonomously and manages outbound inbox responses.
|