01 / watch
Learn how it works
Weekly explainers on what's under the hood: tokens and caching, agents, LiDAR, Gaussian splatting, digital twins.
> kernel&chai
Deep tech.
Strong chai.
Zero hype.
How modern tech actually works — AI agents, developer tools, 3D and spatial computing — explained by an engineer who builds it. And production audits for teams shipping AI apps.
$ context 50,000 tokens
$ turns 20
$ sent 1,257,000 input tokens
without caching $2.67
with caching $0.47
# same work. 83% less.Three ways in
02 / calculate
See what your agent really costs
Plug in your context size, turns and prices. See the bill with and without prompt caching, turn by turn.
Open the calculator →03 / fix
Get your AI app production-ready
A fixed-price review of cost, security and reliability. A written report ranked by impact, in 5 business days.
See the audit →who's brewing
An engineer who ships, not a commentator
I'm Shubham, a principal engineer. I've architected and built a production digital-twin platform end to end: reality capture from LiDAR, photogrammetry and 360° walkthroughs, 3D on the web, cloud infrastructure, and the AI on top.
Kernel & Chai is where I explain what I've learned in plain words, and where teams can hire me to look at their AI apps with the same eye I use on my own systems.
One email a week. One idea, explained properly.
The week's video in writing, the releases that matter to builders, and nothing else. Unsubscribe in one click.