Shh, he’s napping…

Meet Lilo.

A small scientist who lives on your desktop.
Always on your side.

Watch a day with Lilo

01

Small.

He lives right on your Dock: out of the way, never in it. Look away and he reads, dozes on his books, or tidies his antenna. Say Hi Lilo and he’s already listening.

02

Curious.

Ask him something hard. He searches the literature, reads the papers, runs the numbers, then answers with quotes he has checked are really there.

  • Literature
  • Statistics
  • Figures
  • Single-cell
  • Protocols

03

Every day.

Each morning, the new papers that matter to you, already read and ranked. Today’s reminders, last night’s timers, your follow-ups. One short brief, waiting when you sit down.

04

Always on
your side.

The workspace

Six rooms.
One Lilo.

When a question needs more than a speech bubble, open his workspace. Everything he knows about a project goes with you from room to room.

Chat

Bring him
your data.

Drop in a PDF, a CSV or an image. He reads the pages he cites, and when a question needs code he writes the Python and shows it to you. You press Run; the plot lands right in the chat.

  • PDFs, tables and images in one conversation
  • Code runs when you press Run, never on its own
  • Rename, regenerate, export as Markdown or print
Chat with LiloMistral
counts_per_cell.csv

How many cells look empty? Plot the UMI counts.

import pandas as pd
df = pd.read_csv("counts_per_cell.csv")
df.umi.plot.hist(bins=40, logx=True)
Run this code?Not nowRun

412 of 5,120 barcodes sit below 500 UMIs. That first bump is probably empty droplets. Want me to filter them and redo the QC plot?

Papers

He reads
so you can think.

Tell him what you care about once. He searches PubMed, arXiv, bioRxiv and medRxiv on your schedule and ranks what’s new against your interests. For bigger questions, a rapid review: search, screen, extract, report.

  • Every statement quotes its paper, and he checks the quote is really there
  • A full search record for your methods section
  • Your Zotero library, BibTeX, RIS and PDFs
Paper news · Tumour stromaWeekly
  1. 9.1Fibroblast states map to distinct niches in pancreatic tumoursbioRxiv · 2 days ago · Core to your question
  2. 8.4A spatial atlas of stromal remodelling after chemotherapyPubMed · 4 days ago · Same model system
  3. 7.6Benchmarking cell-type deconvolution for spatial dataarXiv · 6 days ago · Methods you use

Myofibroblast-like cells sat closest to tumour glands[3]

“…were enriched within 50 µm of malignant ducts…”

Quote found in paper 3

Lab

A second pair
of hands.

Give him a protocol PDF, or just paste the text, and he turns it into steps and timers you check once. At the bench he reads each step aloud, keeps the clock, and writes down what you tell him, with the step and the time.

  • Reusable protocols and a notebook for every run
  • Lilo, note: used 2 µl instead of 1.
  • Calculators for dilutions, solutions, master mix and cell seeding
  • Send a protocol to your iPhone with one QR code
RNA extractionStep 4 of 9

Now

Centrifuge at 12,000 × g for 15 min at 4 °C.

BackDone

“Used 2 µl instead of 1.”Step 4 · 14:32

Dilution · C₁V₁ = C₂V₂

25 µL stock + 975 µL buffer

10 mM × V₁ = 250 µM × 1 mL

Analysis

Your data,
in a notebook.

Real Python notebooks, right next to your datasets. Start from a ready-made notebook for tables, flow, single-cell, bulk RNA or images. When a cell fails or begs a next step, his suggestion sits right under it. Nothing runs until you press Run.

  • CSV, FCS and FlowJo, h5ad and images, up to 10 GB each
  • Figures as PNG, editable SVG or PDF
  • Extra packages, or point it at your own Python
PBMC quality controlRun all

pbmc_10k.h5ad10,241 cells · 1.8 GB

[ ] adata = lilo.load("pbmc")
sc.pp.calculate_qc_metrics(adata, qc_vars=["mt"])
sc.pl.scatter(adata, "total_counts",
              "pct_counts_mt")
0:04 · 10,241 cells

Lilo suggests

8 % of cells sit above 20 % mitochondrial reads. Filter them before clustering?

adata = adata[adata.obs.pct_counts_mt < 20]
Adds a cell below. You run it.DismissAdd cell

Save figure · qc_scatter

PNGSVGPDF

Text stays editable in Illustrator.

Plan

Your week,
worked out.

Pick a protocol and a day 0, and he lays out every step: day 3 at 14:00 or 72 h after step 2, even across a clock change. Weekends, holidays and clashes with your calendar are flagged, never moved behind your back.

  • From a Lab protocol to dated steps, without guessing waits
  • Templates and repeating chores, like weekly media changes
  • Becomes Reminders after one look, and moves with the plan
Organoid differentiationDay 0 · Thu 5 Nov
  1. Day 0Thu 09:00Seed organoids in Matrigel
  2. Day 1Fri 10:00Switch to differentiation medium
  3. Day 3Sun 14:00Medium changeWeekend
  4. +72 hWed 14:00Add BMP4Lab meeting 14:00
Nothing moves on its own.Add 4 reminders

Notes

He remembers
your projects.

Say Remember that… and it stays: per project, across restarts and model changes. Runs, papers and notes come together when you need them: ask for this week and he prepares your lab meeting from what really happened.

  • Review, edit, pin or forget anything he knows
  • Weekly updates for your lab meeting or your PI
  • Built only from your records, with sources

Pinned · Stroma project

Our organoids are passaged every 7 days; use passage 4 to 12 only.

Lab meeting · last 7 days
3runs2deviations5papers saved
  • Mon RNA extraction · step 4 used 2 µl instead of 1
  • Wed Rapid review finished · 18 studies included
  • Fri Reminder · order collagenase

The research brain

The model is the engine.
Lilo is the scientist.

Under the helmet there is more than a chatbot. The AI model you pick does the talking. Lilo brings the method: what to look up, which tool to reach for, how to check the result.

  1. He decides what to do. Ask Is this effect real? and he doesn’t guess. He searches the literature, opens your file, runs the statistics, checks every quote. Each is a tool he picks up and puts down, the way you would.
  2. He reads the manual first. 38 reviewed science skills: the written-down method for critical appraisal, experimental design, statistics, single-cell analysis, primer design and more. He opens the right one before he starts.
  3. He shows his work. Every search, every file, every line of code is a step you can see, so you always know where an answer came from.
You Is the dose effect in plate_3.csv real, or just two outliers?
  1. Opened plate_3.csv · 96 wells, 4 doses
  2. Read the skill Statistical analysis · which test fits
  3. Ran a robust fit, with and without the two wells
  4. Checked 3 papers · 3 of 3 quotes found

Real. The trend holds without the two wells (p = 0.004, same direction). The outliers are both from column 12, so look at the edge first.

  • Literature search
  • Your files & PDFs
  • Python, when you press Run
  • Public biology databases
  • Zotero
  • Reminders & calendar

Put to the test.

Same model, all 199 LitQA2 questions about real papers.

Precision: how often the answer was right, when one was given.

Lilo for iPhone

At the bench,
in your pocket.

A small, free companion for when your hands are busy. The maths you’d otherwise do on a glove, your protocols step by step, and timers you can see from across the lab.

  • Calculate. Dilutions, serial dilutions, master mix, solutions, cell counts and seeding. The same engine as on the Mac, checked against 555 reference cases.
  • Protocols. Show a QR code on your Mac, scan it, and the protocol is on your phone. He reads each step aloud, or ask Siri for the next one.
  • Timers. Several at once, on the lock screen and in the Dynamic Island. Next step works without unlocking.
  • Back to the Mac. Your run, notes and times included, lands in the Lab notebook with one AirDrop.

Coming soon to the App StoreFree · No account needed

9:41

Dilution

MolarMassFold

Stock
10 mM
Target
250 µM
Final volume
1 mL

Add 25 µL stock

to 975 µL buffer

Spin 04:32

CalculateProtocols

RNA extraction Step 4 of 9

Centrifuge 12,000 × g, 4 °C

14:12BackNext step

Scan from your Mac

And the little things

Made for
a scientist’s day.

How he works

Reading is free.
Acting asks.

He reads what you give him. Before he saves a file, changes your reminders or acts for you, he shows you exactly what will happen and waits. No answer means no.

  • Only your folders. He sees the folders you add, nothing else.
  • Copies, not edits. report.docx stays; he writes report_Lilo.docx.
  • You press Run. His code waits in the notebook until you run it, in a sandbox that can’t reach the rest of your Mac.
  • Your choice of brain. Lilo Cloud, the AI provider you already trust, or keep it all on your Mac with a local model.

Lilo Cloud

You handle the lab.
We handle the computing.

No API keys of your own? Can’t, or don’t want to, host a model yourself? No problem. Meet Lilo Cloud: no keys to paste, no extra account with an AI company. He brings his own computing, free for as long as our credits last.

Free while our credits last

The film

A day with Lilo.

Ninety seconds, one day: papers before breakfast, notes at the bench, an experiment that bubbles over, a figure, a literature review and a report by night.

Early access

Early access

Request early access.

Two quick questions, then Lilo will be in touch.