A second brain is just a place outside your head where useful thoughts can be captured in seconds, filed with almost no effort, and found again months later by searching the words you actually remember. On iPhone, you do not need an elaborate method to have one — you need four working parts: fast capture, a few honest projects, trustworthy search, and a tiny review habit. This guide builds that in Thoughts, explains where AI genuinely helps (and where it is decoration), and keeps the whole thing maintainable by a busy person.
What most “second brain” systems get wrong
The popular methods were designed by people who enjoy knowledge management, for libraries of other people's ideas: highlights, literature notes, evergreen principles with linking ceremonies. They assume desk time, a keyboard, and enthusiasm for maintenance. Applied to a phone-first life — ideas in queues, decisions in corridors, recipes at midnight — the maintenance loses to the moment, capture stops, and the system becomes a beautifully labelled empty building.
The useful core underneath the ceremony is small: get it out of your head, put it somewhere findable, revisit it when relevant. Everything else must earn its place.
Part 1: Capture faster than forgetting
Your second brain's value is capped by capture friction. Voice-first capture on iPhone wins because the phone is present at the moment of thought and speaking needs no desk. In Thoughts, you say the thing; on-device intelligence tidies the transcript, suggests a title and a project; you move on. One thought per capture, context first — the technique from brain dump on iPhone doubles as your capture standard for everything.
Text capture matters too, for quiet rooms and pasted snippets. What matters is that both land in the same library with the same filing, not in separate apps for separate moods.
Part 2: Projects named after your actual life
Replace abstract categories (“Knowledge”, “Resources”, “Areas”) with projects you would say out loud: Kitchen renovation, Launch, Recipes, Money, Dad's house. A thought's home is where you will look when it becomes useful, and that is a project, not a topic in a taxonomy. Start with a handful. Split only when browsing gets slow or sharing boundaries differ. Everything undecided goes to Unfiled, your inbox — the full rationale is in organising notes by project.
Let the filing suggestion do the first pass and correct it when wrong. Corrections are cheap; hesitation at capture time is expensive.
Part 3: Search you trust (and titles that feed it)
Retrieval in Thoughts is literal: titles, text and tags, matched on the words present. That predictability is a feature — no wondering what the algorithm felt like today — but it asks one thing of you: distinctive words in titles. “Boiler service — quotes, Nov” will be found by boiler, quotes or Nov. “House stuff” will be found by luck. Spend your three seconds of effort there, not on folder choreography.
Test retrieval monthly with a real question from your life. If search fails you, the fix is almost always a better title on the thought you eventually found by scrolling. Improve it on the spot; your future searches inherit the repair.
Part 4: A review measured in minutes
Second brains die of write-only memory. The antidote is embarrassingly small: five minutes, weekly, attached to a moment you already have. Empty Unfiled into projects, sharpen two or three titles, move commitments to reminders, delete courageously. Then open one active project and reread its latest thoughts before the week needs them. The fuller ritual — including using an AI assistant to summarise what changed — is in a weekly review your notes and AI can do together.
Where AI helps, honestly
Three places. On-device: tidying speech into readable notes and suggesting titles and projects — private, offline, unglamorous, and the reason capture stays fast. At save time: turning a ChatGPT or Claude conversation into a distilled thought filed in the right project, so the library absorbs what your chats produce — the notepad for ChatGPT pattern. At retrieval time: asking an assistant to search, read fully and summarise across projects, per asking AI about your notes.
Notice what AI does not do in this design: own the library, decide what matters, or edit your record. Assistants are visitors with library cards — the same card in ChatGPT and Claude, as one second brain for both describes. The brain is yours, on your phone, useful with every assistant switched off. Keep the sensitive projects local and they are not even visitable; the split is explained in what stays on your device.
Your first week
Day one: create three to five projects named after live concerns; leave everything else to Unfiled. Days two to six: capture by voice whenever a thought repeats itself in your head — repetition is your brain begging for external storage. Day seven: five-minute review, then ask your library one real question by search. If it answers, you have a second brain. Grow it by use, not by design documents.
Frequently asked questions
What do I actually need for a second brain on iPhone?
Four parts: capture faster than forgetting (voice-first), a few projects named after real areas of life, search fed by distinctive titles, and a five-minute weekly review. In Thoughts, on-device tidying and filing suggestions keep the effort near zero.
Do I need tags, backlinks or a formal method?
No. Those suit desk-based research libraries. For phone-first capture, projects plus literal search over titles, text and tags retrieves reliably, provided titles contain the distinctive words you would search for.
What role does AI play in a second brain?
On-device AI tidies, titles and files captures privately. Connected assistants like ChatGPT or Claude can later search and summarise your synced library and save distilled answers back into it, while you keep ownership and editing control.
How do I stop my second brain becoming a junk drawer?
Review little and often: five minutes weekly to empty Unfiled, sharpen titles, move commitments to reminders and delete freely. Capture without review is write-only memory, however good the app.