The promise
Three months.
Every year.
Returned.
Palimpsa is the academic operating system. Every claim traceable. Every paper monitored. Every idea dated. Every analysis reproducible. The system that gives you a quarter of your year back — provably, measurably, on a maxed-out account.
How we get to three months
Motion's headline is the 13th month — one extra month per year for knowledge workers. Researchers lose a far higher fraction of their year to non-research work: re-reading papers they've forgotten, re-formatting citations between journals, fact-checking their own draft, rebuilding analyses whose code went missing, addressing desk-rejects on issues that were always catchable. We can credibly promise three months a year because the surface area of waste is that much bigger.
The contract: each row below is a feature, not a vibe. If the feature ships and you use it, the hours come back.
| Source of returned time | Tier | Active user (h/yr) | Heavy user (h/yr) |
|---|---|---|---|
| Re-reading papers you've forgotten — answered by semantic memory | 3 | 60 | 120 |
| Citation format / existence / freshness — automated by heuristics | 2 | 30 | 60 |
| "What did I read about X" — answered without re-reading | 3 | 40 | 80 |
| Daily arXiv triage — pre-scored relevance digest | 5 | 60 | 100 |
| Retraction-flagging — eliminates panic-edit cycle | 5 | 5 | 20 |
| "Did I already answer this?" — answered without re-reading the draft | 4 | 30 | 60 |
| Lit-review completeness — caught before submission, not after R2 | 4 | 20 | 40 |
| Pre-submission audits eliminate ~1 desk-reject cycle/year | 9 | 80 | 160 |
| Reproducibility receipts eliminate rebuild-the-analysis episodes | 10 | 60 | 120 |
| Defensibility ledger eliminates "I had it first" disputes | 11 | 5 | 30 |
| Total | ~390 h | ~790 h |
A research-active academic year is ~1,800 productive hours. 390 hours = 2.4 months back for an active user; 790 hours = 4.9 months for a heavy user. The promise rounds the honest middle.
Why this compounds
ChatGPT and Claude.ai are conversations with a brilliant amnesiac stranger. Every prompt starts cold. Palimpsa is different in kind: every feature is grounded in your reading list, your drafts, your supervisor's pinned comments, your experiments. Year 2 saves more than year 1, because the memory is bigger. Year 5 saves more than year 2.
The user who adopts Palimpsa builds a moat against the user who doesn't.
What we will not promise
- Faster writing — we don't write. AI is a critic, never a writer.
- Better grades or more citations — that's your work.
- AI does it for you — you will always do the writing.
- A specific paper count — confounded by field, supervisor, luck.
The promise is time returned, because that's what the system actually delivers and what the user actually loses.