I write to work things out, not to report conclusions I already have. Most of what I publish starts as a question I’m stuck on in something I’m building.
What I’m working on
When is AI actually the answer, and when is it lookup tables or deterministic workflows wearing an AI costume? I watched this exact question go unasked through enterprise blockchain. I hold ten patents in it, and most of what people wanted to build didn’t need it. The same pattern is running again.
What does it take to ship into a system you can’t afford to break? Building beside a legacy stack is easy and gets ignored. Building inside one is slow, political, and the only version that changes anything.
What does a regulated environment actually do to product decisions? Not the compliance checklist—the sequencing. Which thing you build first changes completely when being early and wrong costs you a licence.
What is genuinely different about building with AI agents right now? I’m building this at one7x Labs, so this is the one where I know least and am writing from the middle of it.
Why subscribe
It’s called Dipender’s Newsletter, on newsletter.dipenderb.com. Being straight with you: this is early. I’ve been building far longer than I’ve been publishing, and there isn’t a back catalogue to browse yet. What arrives is a build note or a working-out—what I decided, what it cost, what surprised me—when I have something worth writing down rather than on a schedule.
It’s for people who build products or run the systems around them, and who’d rather read someone still figuring it out than someone selling the answer. If that’s not what you want, don’t subscribe. I’d rather you read the work instead.