Reads the literature
She reads the papers and shows her sources. Every fact links straight back to where it came from — nothing unsourced.
Meet Lois · your AI lab partner
Meet Lois.
Meet Lois.
Your lab partner.
Lois works alongside you like a careful postdoc who never sleeps — and never guesses. She reads the papers, pulls the data, keeps track of your ideas, and does the legwork. And the part that matters most: when she isn't sure, she tells you. Nothing she hands you is made up.
A research prototype from Perslis Research. She runs on your hardware, and you decide what she can touch.
Ask her something.
Watch her ground it — or admit she can't.
This is Lois's research runtime, live. Ask her anything: a gene, a drug, a disease, how something works. She looks it up in real sources while you watch (UniProt, PDB, AlphaFold, ClinVar, Ensembl, PubChem, ChEMBL, Reactome, KEGG, InterPro, Europe PMC, OpenAlex, arXiv, Wikipedia), then answers only from what she fetched. Every cited sentence links to its receipt; anything the evidence doesn't back is marked unverified instead of dressed up.
Conversation
Hi — I'm Lois. Ask me anything. I'll look it up live in the real databases and the research literature, show you each source as I read it, and put a receipt on every fact. If the evidence isn't there, I'll say so.
A hosted language model plans her lookups and phrases the answer, but it is never the source of a fact: each sentence is checked against the record it cites, and a sentence with no receipt is labelled unverified. This public console sends your question to that hosted model, so don't paste anything confidential. A lab install runs on local models instead.
Everything Lois can do,
in one lab partner.
Not a chatbot bolted onto your workflow — a partner who does the parts of research that eat your day, and does them with her receipts showing.
Pulls the data
Structures, sequences, variants, chemistry, pathways — gathered from the world's databases into one place, so you stop tab-hopping.
Never makes things up
When the evidence isn't there, she says "I don't know" instead of a confident guess. That honesty is the whole point.
Tracks your hypotheses
She writes down what she expects before she looks — so she can't quietly move the goalposts, and neither can you.
Connects your lab
Your instruments, LIMS, notebook, and datasets — brought into one console she can actually reason over.
Does the legwork
Organizes files, prepares data, runs analyses, and kicks off long jobs — then checks back with you when they're done.
Works while you sleep
Hand her a task and come back to finished results. She works on her own and shows you exactly what she did.
Remembers
She keeps the thread across days and sessions, building on what you did before instead of starting cold every time.
Stays yours
Runs on your own hardware. Your data doesn't leave the building unless you decide to send it.
Keeps you in control
Scoped access you grant and can pull back at any time — and a quiet review before she ever touches your systems.
Speaks your language
Works in English, 中文, and 日本語 — while the facts themselves (accessions, IDs, sequences) stay universal.
Shows her work
Every answer comes with the trail behind it — the sources, the path, the checks — so you can follow it, not just trust it.
She'd rather say "I don't know"
than impress you with a guess.
Lois doesn't just sound confident — she shows her work. Every answer arrives with its evidence attached, and where the record is silent she hands you an honest gap instead of a plausible-sounding invention. For research, an honest "unknown" is often the most useful answer you can get: it tells you where the next experiment should go.
Built for people who
can't afford a wrong answer.
You see everything she does. She earns access in stages, on your terms, and gives it back when the work is done. She's grounded in a knowledge floor that only lets through what can be traced to a source — so the thing you get to trust is the evidence, not her tone of voice.
What that feels like day to day
You ask a real question. She comes back with an answer, its sources, and an honest note about what's still uncertain. You check the trail, push back, and she updates. It feels less like prompting a tool and more like working with a colleague who does the tedious parts and keeps you both honest.
How she answers without weights.
Ask her about a protein she holds and the facts come back from her floor, each one pinned to its source. Ask for something she has no evidence for and she says so.
Answering from the record
Symbolic recall on the public demo — no model in the loop at all.
Continue exploringHow she reasons