Is There a Compute Surplus On the Horizon?
Benny & The Squirrel: Ep. 19 with Sachin Dev Duggal, Founder of SekondBrain.AI
This week we have a real treat for you all as our guest was my longtime friend Sachin Dev Duggal, who has a Substack you should all follow (@Sachmans). For those of who don’t know Sachin, he is an award-winning entrepreneur and technologist who’s spent the last decade building, scaling, and stress-testing SaaS and software platforms globally.
Sachin’s perspective comes from being deep in the work: across product, AI, and platform strategy — seeing both the upside of hyper-growth and the hard lessons that only show up when things don’t go to plan.
In this conversation, we talk candidly about where SaaS is heading, what AI genuinely changes — and what it doesn’t — and why the future of software may depend less on tools — and more on how we design systems people can actually trust.
ENJOY!
Instead of giving you a bland summary of the episoe here I am just going to repost Sachin’s linkedin summary as I thought it was brilliant. (Sachin LinkedIn Profile)
Sachin’s Summary
A few days ago I jumped on Benny & the Squirrel as a guest. 🎙️
No script. No deck (well Benjamin always has slides because he’s one of those walking talking encyclopedias). Just a proper, long, messy conversation with two Brit’s and one American (a new ratio!!)
We hit hard a phrase that keeps floating around the markets like it’s a fact:
“SaaS is dead.” 💭- about 2 years ago I wrote a similar LinkedIn with certain caveats.
And here’s the honest / quantum (Schroeder’s) version (no pun intended):
It’s true.
And it’s false.
It’s true because the original SaaS promise quietly snapped. ⚡
We told ourselves:
Software is expensive.
Software is hard.
Everyone needs the same thing.
All three are collapsing in real time.
- The craft barrier has fallen.
- Feature-building is cheap(er).
- And “out-of-the-box” became a polite myth… the moment customers started paying 7–9x the licence cost just to make the thing usable.
But it’s also false because… not all SaaS is the same. 🧠
What’s ending isn’t “software as a service.” What’s ending is software as a wrapper.
A wrapper on:
a workflow.
a database.
a UI that pretends configuration is value.
What survives are the systems that own something non-replicable:
Risk.
Regulation.
Distribution.
Data gravity.
That’s why this conversation quickly became an AI conversation too. 🔥 Because most AI right now is scaling output faster than we’re scaling truth.
Fluency isn’t reliability.
It’s just a very convincing accent.
And the deeper point (the one I keep coming back to) is this:
Language is how we transmit meaning.
It’s not where meaning lives.
So if you build AI on files + prompts… you’re building on the wrong atomic unit. You’re asking a system to “understand” a container.
Whereas the real substrate is concepts. And the relationships between concepts. That’s fidelity because context isn’t a blob. It’s a neighbourhood. 🗺️
That’s what we explored on the episode:
- why SaaS compresses into features, not suites.
- why AI needs reconstruction, not endless regeneration.
- why “trust” is the bottleneck, not compute.
- why there is a serious risk of over investment in gpu etc in the mid term