Opus 4.5 performs really well in Kiro IDE

[Ethan]

I haven’t subscribed to Kiro yet—just tried it using the free 50 credits for the first time.

I ran into a CI issue where tests were failing because of snapshot mismatches. I tried updating the snapshots locally, but nothing changed and I was kinda stuck.

I explained the issue to Kiro + Opus 4.5, and it actually figured out the root cause. Turned out I was using a cached git repo (a boilerplate related to the snapshots). That boilerplate had been updated recently, but my local cache only refreshes once per week, so everything was out of sync.

What impressed me most:

Kiro didn’t just tell me what to do — it ran the commands for me.

It cleared the cache and updated the snapshots automatically. I didn’t have to run anything myself.

After that, CI passed. Done.

First impression of Kiro: pretty solid.

The free 50 credits is also a smart move—it’s enough to actually hit real problems and see the value.

Good job, Kiro. I might subscribe if Cursor starts to suck

[Ethan]

Just checked again — turns out Kiro also gives 500 bonus credits
Didn’t expect that at all. What a nice surprise

[Ethan]

I think part of why Opus 4.5 feels so good in Kiro is because the IDE isn’t just sending the raw model request — it’s wrapping it in its own optimized prompts and context handling. A lot of these tools do a bunch of prompt engineering behind the scenes to bias the model toward certain behaviors or workflows, so the “same” model can act pretty differently depending on how the IDE structures the input. That’s why Opus can seem way better in one IDE than another — it’s not just the base model, it’s what the IDE feeds it