Opinion
You don't need an AI second brain.
You need a capture habit and one file you own. Second-brain systems, AI-powered or not, fail at the same step every time: not retrieval, not organization, capture. And capture fails because it happens somewhere other than where your thinking happens, which in 2026 is inside AI conversations.
I've watched the second-brain pitch cycle through three generations of tools in fifteen years of marketing and building: elaborate notebook methods, then PKM apps with graph views, now "AI second brains" that promise to remember everything for you. I've also watched what happens to almost everyone who builds one. The system gets built. The system gets abandoned. The notes from month one are the only notes in it.
Why second brains actually fail
Not at retrieval. Retrieval is a solved, even over-solved, problem; every one of these tools has spectacular search. They fail at capture, because capture is a habit, not a feature, and the habit breaks the moment it costs a context switch. When keeping a thought means leaving the place where you're thinking, opening another app, and deciding where the thought goes, you do it for two weeks and then you stop. The collector's fallacy does the rest: saving starts to feel like knowing, the vault fills with unread clippings, and the genuinely important stuff, the decisions you made, still isn't in there.
The ground shifted: your thinking has a location now
Here's what the second-brain conversation hasn't caught up with. A huge share of knowledge work now happens inside AI conversations. The decision gets made mid-chat. The number gets settled mid-chat. The follow-up gets promised mid-chat. Which means capture finally has an obvious home: the conversation itself, at the moment the keepable thing appears, handled by the model that just watched it happen.
What to build instead
Three parts, deliberately boring. A command you can type without leaving the thread. A single plain-text file as the destination, grouped by project, dated, newest first. And ownership as a hard requirement: Markdown on your machine, committable, portable, hostage to no platform. That's the entire system. No graph view, no tags taxonomy, no weekly review ritual you'll skip.
This is exactly what JotThread is, and equally, what it refuses to be. It's not a knowledge base and it will never "surface insights." It's a capture habit with a file: /jot mid-conversation, one line back, everything in jotthread.md, a board when you want the overview. The free version of the habit, using nothing but a prompt, is in the capture guide; start there and see if the habit sticks before spending a dollar.
Keep what matters. Skip the brain.
Common questions
Isn't Claude's memory already a second brain?
It's half of one: context that makes Claude smarter about you. It isn't an organized record you can open, search in an editor, or take with you. Context and record are different jobs; see the memory guide.
Do PKM apps like Notion or Obsidian have a place?
Sure, as libraries for finished thinking: published notes, references, documentation. The failure mode is using a library as a capture tool. Capture needs to cost two seconds and zero context switches.
What should a capture system cost?
Almost nothing, in both money and attention. Anything with an onboarding course is already too big. The system described here is one command and one file.
Not a second brain. A capture habit.
One command, one Markdown file, no platform. JotThread is deliberately the smallest tool that solves the actual problem. $29 once.