Why now
AI is moving faster than research can follow.
The length of task an AI system can finish on its own has doubled about every seven months since 2019.
- 2019 · secondsLanguage models finish tasks that take an expert a few seconds.
- 2023 · minutesGPT-4 handles tasks of several minutes.
- 2025 · hoursThe newest models work through tasks that take an expert hours.
- If the trend holdsAI finishes a working day’s task on its own around 2027.
The gap
Good research can be reproduced.
Someone else, with the same data and code, should get the same result. With today’s AI apps, a lab easily loses that control.
What a lab loses when nobody checks
- Understanding
- Results arrive faster than anyone can explain them. Papers cite numbers no one in the lab can reproduce.
- Adaptability
- Skills we stop practising fade. The next unfamiliar problem meets a lab that can only prompt.
- Independence
- Every step runs through one vendor’s model, pricing and terms of use.
- Data
- Unpublished data, samples and ideas leave the institute to be processed elsewhere.
The rules
Today’s AI tools don’t meet the rules science already has.
An analysis done in a chat assistant usually can’t be run again. The code, the data and the decisions stay in a chat log, and nobody records who accepted the result.
A person is accountable; AI cannot be an author.
COPE position statement on authorship and AI tools, 2023 · ICMJE Recommendations
How it should workEvery value, figure and paragraph names the person who accepted it.
Use of AI is disclosed and documented.
Nature Portfolio editorial policy · DFG statement on generative AI, 2023
How it should workEvery AI suggestion is labelled with its model and kept with what was accepted or discarded.
Every result can be traced back to its data.
How it should workEvery number links to its run, code and data, and can be run again.
AI text published to inform the public needs a label, unless a person reviewed it and answers for it.
EU AI Act, Article 50, applies from 2 August 2026
How it should workAI text stays marked until a person reviews it.
So we build products that bring AI into research and keep the work reproducible and verifiable.
- Runs againEvery number links to its data and code.
- Has a name on itA person accepts each result, and AI text stays marked until they do.
- Stays with youOn your institute’s servers, with open models.
Products
OS, the shared workspace for research groups.
A web app for the whole group, in the browser: chat, tasks, data, analysis and writing on one record.
- OSWeb app for the whole group
DrawingsFigure editor with a drawing library
CompanionMac app that turns mail into tasks
SeqDeskFor sequencing facilities
OS in use: eight short demos of a fictional lab at work.
The demo starts when it is in view.
OS runs in the browser. Nothing to install.
Chat, tasks, data, analysis and writing on one record, with every result tied to its run.
OS is short for Ongoing Science.
More about OS →
Working togetherChat, people, calendar, tasks and projects.
Analysis and writingData, analysis, figures and the writer, from the data to the paper.
Your fieldAdd-ons, built with founding labs.
- 02 · On its own, in your browserDrawings
A figure editor with 3,054 scientific drawings in one style, for papers, posters and teaching. Institutes license it for everyone, who sign in through their own institution. Inside OS it is an area of the workspace; on its own it runs in the browser, and your figures stay there.
About Drawings →
- 03 · Part of OS, on your MacCompanion
Part of OS, as a Mac app for your mail. It sorts incoming mail and turns what needs doing into tasks in OS. A language model service of your choice helps it sort and extract, with guardrails that keep your mail safe.
About Companion →
- 04 · For sequencing facilitiesSeqDesk
Orders, samples, runs and pipelines, from the request to the delivered data. It runs on its own, or connects to OS: the reads arrive in the group’s project, ready to start pipelines and analyses.
About SeqDesk →The lab
ResearcherOrders a run and sends the samples.order, samplesSeqDesk
Sequencing facilityChecks samples and metadata, plans the run.run planRun 2417
SequencerReads for every sample, filed with the order.readsYour compute
PipelinesQC, counts and results for each sample.data and results
For institutions
Your institute’s AI, on your infrastructure.
- Self-hostedOn your own servers or with a European provider you already trust. Research data does not have to leave.
- Open modelsWhat your lab writes does not train someone else’s model.
- Good research practiceNamed checks, traceable numbers and labelled AI output, designed around the DFG code, GDPR and the EU AI Act.
- Three partsThe app in the browser, a collaboration server for the shared record, and a compute server where your data is.
Try it · Manuscript writing
Your words.
A second pass.
Our AI model is tuned for manuscript-style writing. Paste a paragraph from your draft and try it here. Review its edits on the page, then choose which changes to keep.
Experimental Open full-page editor ↗ Compare edits blindly ↗ Compare Methods drafts ↗
Select a passage to edit just that part, or work on the whole fragment. Nothing is sent until you ask for a suggestion.
We do not log or save your manuscript text or the suggestions we generate. Your text is sent through our website to our GPU server only when you request an edit. Drafts stay in this tab and disappear on reload. We keep short-lived usage counters to prevent abuse; standard hosting metadata is logged separately. How the demo handles data
Coming soon
Preprints you can verify.
A new kind of preprint server. It holds the manuscript together with its data and code, and shows which results in the paper can be verified end to end.
Example · preprint v1 · fictional lab
Colonisation reshapes the colonic epithelium within six hours
Weber, Becker, Rossi, Morgan
- ✓ Re-run
Six hours of colonisation changes the expression of 1,146 genes.
Run #14 · data and code attached · result matches - ✓ Re-run
41 of 200 hallmark hypoxia genes respond.
Run #14, step 5 · result matches - ● Checked
Vegfa rises in qPCR at 6 h.
Data attached, no code · checked by the authors - ○ Text only
This matches earlier reports in conventional mice.
Cites the literature · nothing to re-run
Paper, data and code together
Post the manuscript with the data and code behind each figure and number. Readers can open any value and see where it comes from.
See what is verifiable
Every result says whether it was re-run from data and code, checked by the authors, or is text only. No more guessing.
An API for journals
Journals can ask for a reproducibility report automatically when a manuscript is submitted.
GET /v1/preprints/{id}/reproducibility { "results": 4, "rerun": 2, "checked": 1, "text_only": 1 }
Science is never finished. With OS, the next person can pick up where you left off.
The mark is a half-open interval: closed where the work began, open where it continues.
Be one of the first labs to run on OS.
We are looking for research groups and institutes to pilot OS with us, and for founding labs to shape the add-on for their field.
Or write to support@ongoingscience.com