>It’s actually worse than that. A centralised SaaS product has to architect for diversity - every customer’s feature permutations, every edge case, every conflicting workflow, all coexisting in a single multi-tenant system. That architectural complexity is enormous. A custom build for a single customer doesn’t carry any of it. One set of features, one workflow, one tenant. Orders of magnitude less complex to build, and orders of magnitude less complex to maintain and run in production.
SaaS also covers all kind of legal requirements (accessibility, auditing, security, payment handling, and so on), and has someone to support, blame, and come and fix it when things go sour. Plus the architecture to cover scaling needs.
With some thing Claude churned, you're on your own.
The software engineers of today are what craftsmen and artisans were at the beginning of the Industrial Revolution. It will be interesting to see how similarly this revolution will play out.
In a major (pre-AI-IPO) hype bubble, we can't be sure of any "deeper thing" causing software valuations to fail.
Especially when the economy goes shock after shock unrelated to AI too (tarrifs, a bad economy, unpredictable White House tennants, fears of war).
Valuations are fickle and can reverse course just as well as continue it.
The "deeper thing" is mostly taking what the article says for granted, when it hasn't been happening in any real scale. No real movement of companies getting rid of their Salesforce or Microsoft suite or such just yet - a handful of cherry-picked examples that might have replaced some smaller SaaS with a custom thing with different degrees of success.
>It’s actually worse than that. A centralised SaaS product has to architect for diversity - every customer’s feature permutations, every edge case, every conflicting workflow, all coexisting in a single multi-tenant system. That architectural complexity is enormous. A custom build for a single customer doesn’t carry any of it. One set of features, one workflow, one tenant. Orders of magnitude less complex to build, and orders of magnitude less complex to maintain and run in production.
SaaS also covers all kind of legal requirements (accessibility, auditing, security, payment handling, and so on), and has someone to support, blame, and come and fix it when things go sour. Plus the architecture to cover scaling needs.
With some thing Claude churned, you're on your own.
Yes, and that's a service. There is no product.
We’re all focused on the symptom. Software valuations are falling, and the discourse is about why software valuations are falling.
The deeper thing is this: the traditional economics of software - the idea that building software creates an asset - is breaking.
The software engineers of today are what craftsmen and artisans were at the beginning of the Industrial Revolution. It will be interesting to see how similarly this revolution will play out.
In a major (pre-AI-IPO) hype bubble, we can't be sure of any "deeper thing" causing software valuations to fail.
Especially when the economy goes shock after shock unrelated to AI too (tarrifs, a bad economy, unpredictable White House tennants, fears of war).
Valuations are fickle and can reverse course just as well as continue it.
The "deeper thing" is mostly taking what the article says for granted, when it hasn't been happening in any real scale. No real movement of companies getting rid of their Salesforce or Microsoft suite or such just yet - a handful of cherry-picked examples that might have replaced some smaller SaaS with a custom thing with different degrees of success.