So when I got access to Kiro, which was centered around this very problem, I expected it to go way beyond what claude-code's vanilla quality output was. But.. I was pretty disappointed.

It failed my expectations for a few reasons:

Rigid documentation structure (the steering docs) that requires significant context management with the dynamic path matching configuration.

The way it runs into phases based on a single "vibe" prompt without good back-and-forth feedback made me feel like it was just hallucinating a bunch of random stuff. Didn't really see how this was improving over CC.

No support for persona-based subagents that can operate in independent contexts.

Only supports Claude 3.7/4 with no support for frontier models like Opus or GPT5. I mean what even is the point if you don't have access to the latest and greatest?

Bizarre pricing with “spec” and “vibe” requests. Somehow they’re repeating all the mistakes cursor made instead of leaning into the "cool-down" pricing that anthropic has done (which I personally like).

Wondering if you guys feel the same way? What exactly is the point of Amazon trying to compete with Anthropic or OpenAI if they don't own or cannot vend the frontier models? This alone is enough of a reason not to switch to a AI DevX experience they're vending.

Comments URL: https://news.ycombinator.com/item?id=44985469

Points: 5

# Comments: 1

Another idea that I've been thinking about is context hierarchy:

Low -> High Utility

Base (AI reads tool desc/manpage,etc.) > General human advice (typically use grep this way, etc.) > Specific advice (for this project / impl this is how you use the tool).

Currently the best interface to provide our insights are via MCPs. At https://toolprint.ai/ we're building a human (or machine) driven way supplement that knowledge around tool-use to Claude/Cursor, etc.

A practical way in which we dogfood our own product is with the Linear MCP. If you connect that and ask an agent to create a new issue, it predictably fails because there's no instructions on which linear project to select or the correct way to provide a description around Linear's quirks. When we connect the linear mcp via the toolprint mcp, it gets pre-primed context around these edge cases to improve tool use.

Comments URL: https://news.ycombinator.com/item?id=18318025

Points: 5

# Comments: 0