← all use cases
for research & writing ·one memory layer

Notes that build, not reset.

Long projects lose their thread between sessions. 3ngram keeps your sources, your evolving argument, and the reasons behind each turn.

start free ↵how it works →
how it shows up ·memories your assistant keeps
  • note

    Section 3 leans on the 2024 replication study; verify the sample size.

    claude · −2d · draft/ch3
  • decision

    Reframed the thesis around retrieval, not storage.

    chatgpt · −8d · thesis
  • fact

    The primary source contradicts the widely cited figure; footnote it.

    claude · −5d · sources
what changes ·with the loop closed
  • Resume a draft with its reasoning intact
  • Sources and caveats stay attached to claims
  • The argument compounds across sessions
why it matters ·the shape of the problem

You reframed the thesis last week, flagged a shaky source on Tuesday, and cut a section for a reason you have already half forgotten. 3ngram keeps those moves on record, so the next session picks up the argument instead of rediscovering it.

Long-form work is a running argument with yourself. The sources pile up, the thesis shifts, sections get cut and resurrected, and the only record of why is scattered across sessions that have already ended.

3ngram keeps the connective tissue. As you work, your assistant captures the notes, caveats, and decisions that shape the draft; when you come back, tomorrow or next month, it recalls them by relevance. The tenth session knows what the first nine argued, and the writing compounds instead of circling.

friction points ·what breaks, and what 3ngram does instead
01

The argument resets every session

You spent last week reframing the thesis around retrieval instead of storage. A fresh session opens, and the assistant helpfully argues the old frame back at you.

3ngram: The reframing is a decision on record. The next session starts from the current thesis, and the assistant builds on the turn instead of undoing it.

02

Caveats detach from claims

Section 3 leans on a study whose sample size you meant to verify. That caveat lives in a note from two drafts ago, and the claim now reads more confident than it should.

3ngram: The caveat is captured with the claim it belongs to. When the section comes back up, the verify-this note comes back with it.

03

Cut material loses its why

You cut the counterexample for a good reason. Three weeks later a reviewer asks about it, and the reason is gone.

3ngram: 3ngram keeps the reasoning behind the cut. Corrections and reversals create new versions instead of overwriting, so you can trace how the argument got to where it is.

faq ·the questions we hear most
Does 3ngram store my documents or drafts?
No. It stores the memories your assistant captures as you work: notes, decisions, facts, and the reasoning behind them. Your draft stays wherever you write it; source sync for document systems is on the roadmap.
What happens when the argument changes?
Corrections create a new version and retire the old one instead of overwriting it. The superseded note keeps a link forward to what replaced it, so you can always see how a conclusion evolved.
Which tools does this work with?
Any MCP-compatible client: Claude Desktop, Claude Code, ChatGPT, Cursor, Codex, and more. The same memory follows the project across all of them.

Keep the Reasoning Attached to the Draft

3ngram remembers sources, caveats, and the argument's turns across Claude, ChatGPT, and any MCP client, so drafts compound instead of resetting.

get started free ↵