The AI tool landscape has split into parallel universes.
In one, developers quietly adopt open-source command line tools like OpenCode that connect to local models, document code, and fix actual problems. Podcasting 2.0's Adam Curry found it transformative - helping him understand diffs, avoid cloud lock-in, and maintain control. He'd pay $100 monthly for such transparency.
In the other universe, financial media peddles fantasy. On CNBC, an analyst declared a basic project the "most successful open source project in history" before suggesting AI agents would soon perform open-heart surgery - then awkwardly retreated to designing kitchens.
Meanwhile, real technical progress happens quietly. Agent 'skills' solve a fundamental bottleneck: ballooning system prompts that leave no room for actual work. Instead of endless context, skills let agents load expertise just-in-time.
Adam Curry, Podcasting 2.0:
- This thing has changed my life.
- I would pay these guys a hundred dollars a month. I'd cancel everything.
According to Anthropic's Tariq, the real innovation isn't markdown files but folders containing scripts, credentials, and assets that agents can explore. This turns static instructions into executable knowledge.
The most practical applications? Verification and code review - tasks with clear inputs and outputs. Tools like Skill Creator now help non-engineers test, benchmark, and refine these skills systematically.
While one path builds modular tools solving real developer pain points, the other promises planetary disruption with little substance. The gap between carpenter and architect fantasies has become a chasm.
Tariq, Anthropic:
- A common misconception we hear about skills is that they are just markdown files.
- The most interesting part of skills is that they're not just text files. They're folders that can include scripts, assets, datas, etc.
