ash1794
In 2018, in a handwritten journal I’ve kept on and off since then, I wrote a rule for myself:
“Every unique mark / thought / idea needs to be jotted down.”
I didn’t have a name for what I was building. A few years later the internet gave me one. But by 2018 the habit was already four years old.
Keep, since 2014
My oldest Google Keep note is dated 13 January 2014. I was in my second year at Amrita. When I exported everything recently, it came to more than nine hundred notes, running from that January to this month.
Most of it is unglamorous: shopping lists, packing lists, to-dos. But the unglamorous parts are where the record is honest. From September to November 2014 there is a complete build log for a contest platform we were making at college, one task at a time: a contest model, a password strictness enforcer, email confirmation, a captcha, a leaderboard, ICPC team registration. There is a revision list from November 2014 that runs from Huffman coding to strongly connected components. There are poems from 2016 through 2019.
Nobody told me to keep any of it.
Map everything
Then the notes started turning into systems.
In September 2019 the journal has a plan: “Every Sunday, collect metrics about myself.” The list underneath includes happy thoughts, epiphanies, productive work hours, unique people met, and time spent studying.
A Keep note from January 2020 goes further:
“Start sorting out your life. Map everything you ever wanna do. Organise it. Priortise it. Everything. Everyone… Build brand Ashwin.”
The plan had a calendar, a Trello board, and sections for commentary, writing, political views and reading. Six years later I am building this site, which is more or less what that note asked for. I spelled “prioritise” wrong in it, and I’m leaving it that way.
In January 2021 I sketched something bigger in Keep: a personal wiki, a habit tracker with codenames, a happiness and gratitude log. Around the same time I set myself a smaller, more practical rule: “spend half an hour writing a new entry in my wiki on some article that I wanna remind myself or write about.”
The name arrives
On 31 August 2021 I started a Notion database called LinkDrop. The first rows are two essays on “Building a Second Brain.” The name had finally caught up with the habit.
LinkDrop became my everything drawer: articles, books, videos, podcasts, one film, one TV series, one comic. It had 142 rows by the end, most of them added in two bursts. The first was August to November 2021. The second was September to November 2024, when I was reading about agents: OpenInterpreter, OpenHands, Devika, MetaGPT, AlphaCodium, local models on Ollama, logged right next to lists of ICPC coaches.
The last row is dated 26 November 2024. After that, nothing.
In a letter I wrote to myself in September 2022, one line reads: “Your second brain is taking shape.”
Where it went
LinkDrop stopped in November 2024. What replaced it took another year to arrive: I’ve been agentic since December 2025. Today the second brain doesn’t live in a notes app. It lives in systems that read, remember and act. There’s NoSaintsville, a multi-agent assistant I wrote in Go, with specialised agents, soul files and a learning journal it writes back to. There’s Ash OS, the operating layer for my academic work, with its canonical state kept in a radar that gets pulled at the start of every session and updated at the end. And since this month there’s Ash’s Desk, a single surface meant to hold my working day in under an hour.
It’s the same instinct as the 2018 rule, with better tools.
Capture is the easy half
Twelve years of notes are proof that I capture. They are not proof that I finish. Of the books I wishlisted in LinkDrop, I marked exactly one as finished: Dune. When I ran a 57-question self-interview on myself, finishing and shipping came out as my lowest-rated skill. By my own count, things I mark “Later” get done about 30% of the time. The Software Engineering 101 course I opened in Notion in 2023 still has no lessons.
In December 2014, running a winter coding camp for fifty students, I wrote a line that has since become the thesis of this site: “It’s always easy once you know it, but getting there and proving it is the harder part.”
Capture is getting there. The notes, the wiki, LinkDrop, the metrics on Sundays: they all get you to the point where you know what you think. Proving it is the other half, and a notes app has never been good at it.
The agent systems are where I’ve started pushing on the second half. The rules I run now are about closure more than capture. No session closes without a named artifact. Plan, work, review, compound, in that order, every time. When I audited one of my own agent systems this year, its dashboard showed everything green; of 99 tests, 4 had passed, 93 had failed and 2 were skipped. That audit was the second brain doing the part my old ones never did: checking whether what I believed was true.