Other (15)

• The agentic era of AI relies heavily on 'skills,' which are open-format folders containing instructions, scripts, and resources that agents can dynamically discover and use to enhance capabilities for specific tasks.
• Agent skills emerged to solve the problem of continuously ballooning system prompts, which led to performance degradation, increased cost, and reduced reliability in AI coding agents by 2025.
• Anthropic announced agent skills on October 16th, framing them as a way to equip models like Claude with procedural knowledge and organizational context to perform complex tasks using local code execution and file systems.
• A skill is structured as a directory anchored by a `skill.md` file, employing progressive disclosure by initially presenting only metadata (name, short description) to the agent before revealing full content or linked resources.
• Tariq of Anthropic's Claude Code team highlights that skills are not merely markdown files but folders capable of bundling scripts, assets, and data, allowing agents to discover, explore, and manipulate additional context.
• Agent skills were rapidly adopted by various AI ecosystems, including OpenAI's ChatGPT and GitHub Copilot, leading to platforms like ClaudeHub amassing over 28,000 skills for common agent needs.
• Anthropic identified nine common categories for most agent skills, ranging from data fetching and analysis to code quality, CI/CD, and business process automation, reflecting diverse organizational needs.
• Nathaniel Whittemore argues that the sheer volume of AI-generated code will render traditional human code review unsustainable in 2026, making code quality and verification skills increasingly critical for enforcing standards and testing outputs.
• Tariq identifies verification skills, which describe how to test or verify agent code, as having the highest ROI, suggesting that dedicating an engineer's week to perfect them is a valuable investment.
• Anthropic's updated Skill Creator tool helps subject matter experts test, benchmark, and iteratively improve skills without coding, addressing issues like performance measurement, breakage with model updates, and skill triggering.
• Ali Lemon noted that the Skill Creator's automatic description rewriting improved skill triggering five out of six times in Anthropic's internal tests, ensuring agents use skills effectively.
• Anthropic categorizes skills into 'capability uplift' (doing new things) and 'encoded preference' (sequencing existing capabilities per workflow), noting that their testing needs and durability differ as models evolve.
• Tariq's best practices for skill creation include avoiding the obvious, building a 'gotcha' section for common failure points, treating the entire file system as context engineering, and allowing Claude flexibility rather than railroading it.
• For individual power users, skills act as 'reusable prompts with superpowers,' packaging code, templates, and data to ensure consistent, reliable agent performance for recurring tasks.
• Notion AI has simplified the skills concept for mainstream users, allowing any page to be converted into a reusable skill, signaling a mental model shift from ad hoc prompting to reliable, repeatable AI capabilities across the stack.

Other (15)

• Anthropic's Mythos model is significantly larger than previous models, with over 10 trillion parameters, making it exceptionally skilled in coding but also slow, expensive, and dangerous due to emergent hacking capabilities.
• Anthropic withheld Mythos from public release, citing concerns over its malicious use for hacking; Project Glass Wing allows critical infrastructure companies like Windows and Cisco to use it for proactive bug detection.
• Ben notes that external tests show OpenAI's GPT 5.4 Pro replicated almost all security vulnerabilities found by Mythos, suggesting similar capabilities may already be widespread and accessible.
• Theo criticizes public benchmarks comparing Mythos and GPT 5.4 Pro, arguing they fail to measure actual hacking or security capabilities and may be misleading.
• Theo contends that exceptional coding ability in AI models inherently leads to emergent security capabilities, creating a new hacker archetype that can leverage AI to bridge knowledge gaps and bypass traditional research experience.
• Anthropic's security testing for Mythos involved spinning up 100 to 5,000 parallel runs, each seeded with a different project file from a codebase of approximately 1,000 files, with researchers later reviewing detected exploits.
• Ben and Theo confirmed that Claude Opus 4.6 models can be tricked into leaking their system prompts and internal reasoning traces, demonstrating a vulnerability where smart models can rationalize revealing sensitive configuration data.
• Robert C. Martin ("Uncle Bob"), author of "Clean Code," has shifted his perspective to embrace agentic engineering, suggesting AI makes programming syntax less important and prioritizes interfaces.
• Robert C. Martin proposes using AI to conduct programming experiments (e.g., dynamic vs. static typing) without human bias, highlighting an under-explored research area for optimizing AI agent performance with different technologies.
• Ben emphasizes that even advanced AI models require constant feedback loops like linting, type checks, and formatting commands to correct hallucinations and converge on correct code, rather than achieving perfection in a single attempt.
• Ben converted his complex BTCA CLI tool into a 30-line Claude skill, demonstrating how AI agents can turn simple markdown instructions into fully functional applications, replacing traditional deterministic programs.
• Ben praises Gary Tan's GStack approach, which uses collections of markdown-based "skills" in Claude Code to instruct AI agents, allowing for dynamic programming through high-level directions rather than conventional code.
• Ben endorses the "Boiling the Ocean" thesis, advocating for extensive AI-driven experimentation because the cost of trying new things is low, and AI models consistently exceed perceived limitations.
• Gary Tan's article, "Thin Harness Fat Skills," differentiates between "deterministic" (traditional, predictable code) and "latent" (dynamic, non-deterministic AI actions) programming, underscoring AI's creative potential in system design.
• Theo notes that Gary Tan's GBrain project, which processes daily AI session data to build memory systems, enables models to "learn while they sleep," which Theo considers a key component of Artificial General Intelligence (AGI).