Thank you for writing this. I actually disagree with a lot of the "PRDs are dead" commentary we're seeing online; the thinking is precisely along the lines of your post (just ship it, don't think about it!).
Love this breakdown, Leah - especially the reminder that velocity without validation is just noise. AI might speed up the 'how,' but it still can’t replace the 'why' behind great product decisions.
It feels like I’m being a party pooper when I point this out to folks that are genuinely excited about how accessible it’s become for folks to make things.
Shipping things fast comes from having an actual business goal that you’re pushing towards. Bringing this language to the vibe coding crew creates false advertising
"Yes, vibe coding tools like Lovable et al. help you to ship things faster, but only as long as these ideas struggle with the “development” part and don’t need Ideation and Validation."
Absolutely!
What's more, one of the common misconceptions is that development is a common bottleneck. It might be perceived so since, for years, it was the expensive part. However, it was rarely so. The price tag for development made us more reluctant to bet on everything. Now that we got the AI discount, suddenly people are like, "build, build, build!"
Yet they miss the part where they answer the question "why?"
Even if building something does not change the business's bottom line, it is still costly. It adds to the complexity of the code base and makes future changes more error-prone and more expensive. And thus, future AI code generation will be less impactful.
To add insult to injury, we feed ourselves with success stories that are edge cases rather than the norm. At the same time, what really provides leverage is (early and frequent) validation. Except that part can't be outsourced to AI.
I understand the article and agree with it. And it can be summed up when you say "Building a tech and software product that is commercially viable will always be hard".
but I have two questions:
1 - If we use AI along the process of development (at least MVP development) wouldn't we be gaining time in the process for validation with the users and iterate it along?
2 - What do you mean by complex systems? An Ecommerce can be one of those examples?
A complex system integration could be in mind something that requires "offline" knowledge or data that is not easily accessible by any AI agent or the data is for whatever reason not processable by it.
Security measures in big enterprises are also oftentimes going completely against what AI needs (complete access), there's a million things that an AI can't solve regardless of it's capabilities simply by not having access (or lacking offline context).
We absolute will gain time and you *should* use AI, but it doesn't absolve you from mastering the other disciplines, the reality is that your bottleneck will always be somewhere.
Absolutely agree, Leah. I was part of a startup that launched a product prematurely. We thought we could make improvements on the fly but instead, customer trust took a hit. It's critical to ensure readiness before shipping.
Thank you for writing this. I actually disagree with a lot of the "PRDs are dead" commentary we're seeing online; the thinking is precisely along the lines of your post (just ship it, don't think about it!).
I wrote this post with Ravi, which looks at things in the other extreme: seeing specs as the source code: https://blog.ravi-mehta.com/p/specs-are-the-new-source-code
Would love your thoughts :)
I've said it today somewhere on a different LinkedIn Post.
Prototyping is not the same as building a solution. Prototyping got sped up a LOT, building received a little boost as well.
Whether you draw a prototype, write it out or vibe code it, is all the same in the end, it's to convey meaning about what you envision.
Love this breakdown, Leah - especially the reminder that velocity without validation is just noise. AI might speed up the 'how,' but it still can’t replace the 'why' behind great product decisions.
It feels like I’m being a party pooper when I point this out to folks that are genuinely excited about how accessible it’s become for folks to make things.
Shipping things fast comes from having an actual business goal that you’re pushing towards. Bringing this language to the vibe coding crew creates false advertising
"Yes, vibe coding tools like Lovable et al. help you to ship things faster, but only as long as these ideas struggle with the “development” part and don’t need Ideation and Validation."
Absolutely!
What's more, one of the common misconceptions is that development is a common bottleneck. It might be perceived so since, for years, it was the expensive part. However, it was rarely so. The price tag for development made us more reluctant to bet on everything. Now that we got the AI discount, suddenly people are like, "build, build, build!"
Yet they miss the part where they answer the question "why?"
Even if building something does not change the business's bottom line, it is still costly. It adds to the complexity of the code base and makes future changes more error-prone and more expensive. And thus, future AI code generation will be less impactful.
I see that as a caricature in early-stage product development, where it is even more pronounced than in ongoing product development efforts. There, the invalidation rate is as high as 90% (https://pawelbrodzinski.substack.com/p/90-of-times-validation-means-invalidation).
To add insult to injury, we feed ourselves with success stories that are edge cases rather than the norm. At the same time, what really provides leverage is (early and frequent) validation. Except that part can't be outsourced to AI.
I understand the article and agree with it. And it can be summed up when you say "Building a tech and software product that is commercially viable will always be hard".
but I have two questions:
1 - If we use AI along the process of development (at least MVP development) wouldn't we be gaining time in the process for validation with the users and iterate it along?
2 - What do you mean by complex systems? An Ecommerce can be one of those examples?
A complex system integration could be in mind something that requires "offline" knowledge or data that is not easily accessible by any AI agent or the data is for whatever reason not processable by it.
Security measures in big enterprises are also oftentimes going completely against what AI needs (complete access), there's a million things that an AI can't solve regardless of it's capabilities simply by not having access (or lacking offline context).
We absolute will gain time and you *should* use AI, but it doesn't absolve you from mastering the other disciplines, the reality is that your bottleneck will always be somewhere.
agree!
Whoa, proprietary data's just a myth now? That's wild!
Absolutely agree, Leah. I was part of a startup that launched a product prematurely. We thought we could make improvements on the fly but instead, customer trust took a hit. It's critical to ensure readiness before shipping.
Totally get you, Jiri! It's all about shifting folks' habits, right?