Standing entries
Things I've thought through, written briefly so the thinking has to be honest.
Shorter than essays, longer than tweets. Each one is something I've returned to enough times that writing it down was cheaper than re-thinking it.
How I learn anything new in 2026.
Build first, books second, podcasts third. Action beats analysis of other people. But books have their place, and so do long-form podcasts — I just can't sit through them, so I read summaries or pull the crux and focus on where I can actually use it.
The primary surface is an LLM, used as a tutor. The protocol I've refined over the last year:
- Create the curriculum. State the destination.
- Integrate every good source — papers, talks, threads — into one notebook.
- Dedupe to bullet points. Build a concept map: most fundamental → most complex, like a dependency tree.
- Take the points in batches of five. Repeat each one back to the LLM in my own words. Get clarification where I'm fuzzy.
- Consolidate every twenty points. Compress into chunks I can hold.
- Spaced repetition on the load-bearing ones.
- Use cases, real-world links, metaphors, links to things I already know.
- If five points won't stick, iterate on those five until they do, even if it breaks them into more sub-points.
Courses are sometimes useful, but nothing beats this loop. Mastery, not exposure.
An alter ego, not an assistant.
Eight months ago I had multiple active projects, a calendar I didn't trust, and work eating into my evenings. mineOS is what I built to fix that — a personal operating system that runs as my alter ego. A markdown vault, a set of agents, a few rituals around the edges. The hardest part wasn't building it. It was accepting the truth it showed me — the things I was avoiding and the patterns that repeat.
The thesis underneath it: AI works best as the invisible layer that does what I can't, compounds toward goals I can't hold across context-switches, and turns ADHD into shape rather than friction. Not a co-pilot. A second self. Despite giving up some control over the calendar and some admin aspects of life, I feel more in control now.
The moat is not the code.
The tech is rarely the hard part of a community-and-AI play. The ops, the network effects, and the flywheel are the hard part. So I don't open-source the core — open-sourcing it serves neither us nor anyone else. If you have a counter-view, I want to hear it; I'm not closed on this.
Pricing shape is freemium with tiers, plus a few other monetisation options. A free floor, paid capability, and a couple of revenue streams that aren't pure tier-charge. Industries we won't serve as paying customers: tobacco, gambling, drugs. Individual liberty is a separate question; this is about what danadone amplifies.
AI is the asteroid for hierarchical organizations.
As the number of levels in a hierarchy goes up, the effectiveness of the organization goes down. Each person at each level has their own read of the situation, their own interests. They slightly modify the order from above to suit themselves. Classic principal-agent problem from economics, applied to organizations.
AI-enabled organizations break this. The orgs being built today don't need seven levels. Most don't even need three. The lower-level employees, in touch with reality because they're using the new tools, are now often more strategically valuable than the leadership reading reports about them. The script has flipped. It will keep flipping.
Not against in general. Against for danadone.
I'm not against VC universally. But Indian VC is largely blind right now — a mix of status-seekers who don't know what works, and tech people biased toward IIT and IIM credentials in a year when AI has flipped the asymmetry. Founders who live the hard grind in 2026 are more in touch with the landscape than VCs who were good tech founders five years ago.
danadone doesn't need VC money, and we don't need the pressure that comes with it. That's the calibration today. It might change. The principle won't.
What stays off the brand.
Quantitative-private by default — income, net worth, equity, valuations. Conditions can be public; specifics stay private. ADHD is on-brand; the pharmacology is not. Religion is private; the cultural identity is not. Family details that aren't mine to tell — Prathyusha's career, my parents' health, photos of Agasthya beyond a few — stay off. Failures land here only after they're resolved, not while they're fresh.
The general rule: anything that would be a problem in five years if it were public today, stays private. When in doubt, ask.