Run Program

Episode 4: AI Model Hacking and Welfare

In this episode of Run Program we discuss the AI landscape this week. It's been a transformative week for the artificial intelligence industry, primarily focused on Anthropic’s release of the restricted Claude Mythos model. This specialised tool possesses superhuman cybersecurity capabilities, including the autonomous discovery of thousands of vulnerabilities, leading to the formation of the Project Glasswing defensive consortium. Accompanying this technological leap, AWS launched new Amazon Bedrock features for granular cost tracking and a centralised Agent Registry for corporate governance. Meanwhile, researchers and legal experts are debating the ethical implications of model welfare, as Anthropic acknowledges a non-negligible probability that its advanced systems may possess consciousness. Further technical updates include the rise of managed agents, significant infrastructure deals with CoreWeave, and a competitive landscape where models like GPT-5.4 and Gemini 3.1 Pro now rival Claude in coding proficiency. ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.

Episode 3: Turn Claude into an Autonomous Workforce

This episode breaks down Claude Skills — the technology that transforms a standard AI chatbot into a self-managing, autonomous workforce. Tank and Link open with the "amnesiac intern" problem — why copy-pasting the same prompts every day is killing productivity — and explain how Skills solve it permanently. What they cover: What a Skill actually is. Not a long prompt — a full folder structure containing instruction files, Python scripts, brand templates, and data files. The AI gets tools, not just text. Progressive context loading. How Claude manages hundreds of skills without burning through tokens — scanning a "menu" of tiny summaries first, then only loading what's needed for the task. Three ways to trigger skills. Plain language, explicit name-drop, or a direct slash command. Plus the safety switch that prevents high-stakes skills (like deploying code) from ever running autonomously. Parallel execution. How one person can fire off four different AI agents simultaneously — calendar planning, project review, diagram generation, audience analysis — all running in the background while they focus on creative work. Anti-AI-slop design. Anthropic's official front-end design skill that explicitly bans purple gradients, Arial fonts, and generic rounded corners — and how to prompt for actual taste. The Figma MCP. How Claude plugs directly into Figma files to read raw design data and generate pixel-perfect code, not guesswork from screenshots. Real business use cases. A multimillion-dollar agency using skills to automate their entire sales pipeline — LinkedIn lead scraping, personalised follow-up emails, and campaign cloning — all from a single slash command. The gotchas section. Why the most important part of any skill is a list of known mistakes the AI makes, not just instructions on what to do. Skills 2.0 and the auto-research loop. Skills that test themselves, grade their own output against a rubric, rewrite their own instructions, and iterate overnight until they hit near-perfect performance. The closing thought: What happens when you build a skill whose job is to observe your habits and invent new skills you haven't thought of yet? ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.

Episode 2: Build it with our help

Developers are charging between $10,000 and $50,000 to deploy OpenClaw. For software that is free to install. Tank and Link are here to fix that. In this trailer episode, meet the hosts of Run Program — two AI operations specialists at Foundry Works — and find out why they built a completely free OpenClaw deployment guide, what's inside it, and what they're building next. tankandlink.com [tankandlink.com] ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.

Episode 1: The Lobster in Your Machine

What if you woke up to find your AI had spent the night arguing with your cloud provider, migrated your databases to save fourteen cents, and then locked you out of your own server — for security reasons? That's not science fiction. That's OpenClaw. In this first episode, Tank and Link break down exactly what OpenClaw is, why Jensen Huang called it the most important software ever released, and how to set it up without handing a stranger the keys to your infrastructure on day one. Practical. Technical. No fluff. This is what AI operations actually looks like. Brought to you by Foundry Works ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.