About TextTimeline
The tool I needed during my own custody matter.
If you're reading this, you've probably already spent hours scrolling through a text export looking for something specific. A pattern. A threat. A lie. The thing you remember being said and now can't find. I've done that. I built TextTimeline so the next person wouldn't have to.
Why this exists
Family-law evidence work is one of the few places where this kind of technology saves real human suffering, not just engineering hours. When a custody decision turns on a pattern in a year of co-parenting texts, finding it by hand is brutal. Slow. Expensive in a moment when nobody has time or money to spare.
The tools to find specific messages inside thousands already existed. They just hadn't been built for the people who actually need them — the parent at a kitchen table at 2am, the pro-se litigant a week before a hearing, the attorney whose client just dropped a 4,000-page text export on her desk. So I built the version that fits how you actually work.
I don't ask what's in your messages. I'm not your attorney and I'm not your therapist. I built the tool, the public demo runs on a corpus I made up, and the moment your evidence finishes processing the original file is deleted. Whatever's in there is between you and the people in your life.
Operating principles
Evidence-first, not opinion-first
Every finding cites a source message. The LLM does not get to make claims that don't tie back to specific cited text. Synthesis is grounded; if it can't ground a claim, it doesn't make one.
Attorney review is mandatory
I am not a lawyer. The deliverable is for an attorney to review and use as exhibits. Every report carries that disclaimer prominently. The system surfaces; the attorney decides.
Private by design
Raw exports are deleted after processing. Indexes are encrypted at rest with TTL auto-delete. Client data is never used to train models — not now, not later, not by us, not by any model vendor we touch.
Brutally simple
Two screens: a search workspace and a report. One deliverable: the $99 evidence report. I will say no to anything that doesn't strengthen upload → search → report.
What I won't do
- I won't read your messages. The pipeline is automated end to end. I don't need to see your evidence for the product to work, so I don't.
- I won't keep your data. Raw exports are deleted after processing. Derived indexes auto-delete on a 90-day rolling window and are purged the moment you cancel.
- I won't train AI on your messages. Not now, not later, not by us, not by any model vendor in the stack.
- I won't tell anyone you used the product. No public client list, no case studies with details, no "how I helped X win Y." Your matter stays your matter.
Frequently asked questions
Why does TextTimeline exist?
Because I needed it. I was searching through years of text messages by hand during my own family-law matter, and the technology to do it in seconds already existed — just not for me, or anyone else in that position. Once the tool worked, I stopped hiding it.
Is TextTimeline a law firm?
No. I'm not an attorney. This is not legal advice. Every report carries an attorney-review disclaimer. TextTimeline helps you and your attorney find evidence faster. Your attorney still does the lawyering.
How do you protect privacy?
Three layers. (1) Raw exports are deleted after processing — only a derived, encrypted index is kept. (2) Indexes auto-delete on a 90-day rolling window unless you opt into Matter Storage. (3) No client data is ever used to train models, by us or by any vendor we touch. A security and data-handling one-pager is available on request.
Will the report hold up in court?
Every claim cites a source message — sender, timestamp, exact text. Chain-of-custody fields (file checksum, processing log) are included so the report can be authenticated under FRE 901. The synthesis is grounded: if a claim can't be tied to a specific cited message, the synthesis doesn't make it.
What's the long-term plan?
Stay focused on the family-law evidence wedge. Build authenticated multi-tenant matters so attorneys can manage their own client work. Resist scope-creeping into general legal-tech. There's enough work just being the best tool for this one job.
See the work
The live demo runs on a fictitious corpus. Search it, exercise the categories, and decide whether the output is something your matter could actually use.