Kiro explained in 5 minutes

Kira is a fork of VS Code just like cursor and windsurf and VS code is notorious for adding engineering overhead just to keep the fork up to date with Microsoft releases. Other AI

code editors like Klein and Rue are VS Code extensions by nature and we also have tools like ADER, Cloud Code and Codeex CLI that lives directly in your terminal and some are web- based code

editors like Vzero, Lovable and Bolt as well as cloud-based engineering tools like Codeex. And given this landscape of

like Codeex. And given this landscape of AI code editors, Kira fits in the category of native app and Amazon is well equipped to handle the engineering overhead that comes with forking VS

Code. So my first question after trying

Code. So my first question after trying out Kira was why why did Amazon release Kuro even though it only has basic features and especially when they're somewhat late to the party? The biggest

difference I see is a feature that Kira calls spec. Spec is Kuro's opinion on

calls spec. Spec is Kuro's opinion on how your prompt should be interpreted into structured components. For example

when you send your prompt to Kira, it will split them into three major markdown files: requirements, design and tasks. And Kira will work with you

and tasks. And Kira will work with you to build this out before acting on them.

And if you're used to other tools like Klein and Clot Code, coding agents typically take the driver seat on the coding part. But with spec, Kirro is a

coding part. But with spec, Kirro is a bit more greedy in how it wants to translate your business requirements into a formalized document that it knows how to work on. And this design choice made by Curo is somewhat opinionated

compared to other tools, which is not necessarily a bad thing, but anytime you make an opinionated stance on how things should be done, it adds a bit of a learning curve and a point of contention for users. Another feature worth

for users. Another feature worth mentioning is steering. Steering is

essentially a repository of documents that describe your project. For example

you can describe what your project is about, the text stack you used, and how the codebase is structured in the steering document. And while this

steering document. And while this feature in itself is not groundbreaking at all, it's still interesting how Kira is calling this feature steering when other AI code editors typically call them rules. Another feature worth

them rules. Another feature worth mentioning is what Kira calls hooks.

Hooks are akin to web hooks in API where an event within your IDE will trigger a certain workflow. For example, when you

certain workflow. For example, when you create a new file in Kira, Kira can trigger a predefined workflow such as updating a readme file automatically by using hooks. And based on these major

using hooks. And based on these major features of Kira, we can infer how Kira wants to establish itself in the industry. While some other tools like

industry. While some other tools like Ader and Klein can be treated more transactionally where you view them as a task completion tool, Kira wants to have a bigger scope in the software development life cycle and be more

vertically integrated. And the next

vertically integrated. And the next question is this. Beyond the peripheral features like spec, steering, and hooks how good is Kira as a coding agent?

Because while these features make the coding experience more enjoyable, what truly matters is the underlying agentic mechanism of the AI code editors. For

example, I explained in my other video how client recursively make tool calls to load relevant files into the context so that it can generate a good response that client is delivering. And because

client is open source, we can see exactly how the coding agent works behind the scenes. And for Kira, because Kirro is proprietary software, we have minimal visibility in how Kirro handles

token usage, context management, and product latency. So, anytime Kira is

product latency. So, anytime Kira is running slow, it's really hard to tell whether the context is bloated or maybe the system prompt is not engineered well, or maybe it's just the API connection to LLM providers is slow. And

I'm certainly not advocating that all AI code editors need to be open- source like ADER, Klein, and Codeex. Because

honestly, if the coding agent works seamlessly, I'm much less inclined to know how the sausage is being made. So

until Kuro can establish that trust by writing some solid code is going to face some friction in product adoption especially after pricing will be announced by Amazon for Kuro soon. So

what does the release of Ko tell us about the industry? Although Kuro's

release is seemingly late, every new release made from major companies like Amazon is an indicator that AI code editor industry is still lucrative and competitive. Typically, we see primary

competitive. Typically, we see primary market in the industry get the most amount of funding so that these companies can innovate the state-of-the-art models that we use today. But now, we're seeing the

today. But now, we're seeing the secondary market gaining more capital as the industry widens its focus on encapsulating the state-of-the-art model into an agent to solve domain specific

problems. So, the release of KO signals that we still have a lot of runway in the industry and there certainly is enough pie to go around given the size of the market and all the different mediums that the coding agents can be

delivered into. the terminals, cloud

delivered into. the terminals, cloud on-prem and extensions. And Kira is yet to prove itself in how it stands in this competitive space. And we'll see based

competitive space. And we'll see based on their opinion and approach and how AI should be integrated to coding will actually end up establishing them in the industry.