Instead it: • inspected my repo • checked the provider architecture • researched the APIs • found and reopened an old Together .ai issue • created Baseten and IREN issues • wrote the implementation…

Something I spent real, deliberate effort designing. A repo I've been building for months. Got extended by AI inferring the architecture on its own. Not autocomplete. Not boilerplate. Actual…

Something I spent real, deliberate effort designing. A repo I've been building for months. Got extended by AI inferring the architecture on its own. Not autocomplete. Not boilerplate. Actual…

I wouldn't build any product on a single dependency. AI native development is no different. If that's a controversial take, question the architecture, not the technology.

The strongest case for AI: Many gains are invisible. Docs. Tests. Boilerplate. Summaries. Exploration. Not glamorous. But those tasks quietly drain engineering capacity. That’s where teams should…

What’s interesting isn’t Fibonacci. It’s that modern models now infer the missing senior-engineer context automatically: - iterative over recursive - edge cases - performance assumptions -…

What’s interesting isn’t Fibonacci. It’s that modern models now infer the missing senior-engineer context automatically: - iterative over recursive - edge cases - performance assumptions -…

What’s interesting isn’t Fibonacci. It’s that modern models now infer the missing senior-engineer context automatically: - iterative over recursive - edge cases - performance assumptions -…