AUO sits next to your team every day, turning what your workflow produces into where every hypothesis stands.
Open it in the morning to see what moved overnight. Pull on a thread when a readout lands. Make the call by end of day with the evidence in front of you.
It runs on what you already have.
Upload your presentations, program docs, lab-notebook exports, and decision memos. AUO reconstructs the hypotheses your team is tracking at every level, from experiment to molecule to target to indication to program, with their pillars and load-bearing claims. You commit the exit criteria: the thresholds that would change your mind and the conditions that would kill each bet.
From then on, AUO watches a local folder and the sources you approve: decks, notebooks, computational results, wet-lab readouts, meeting notes, literature. Everything routes to the hypothesis and the criterion it bears on.
AUO produces a recommendation: Stay, Pivot, Confirm, or Abandon.
It connects the dots across everything it has read and your team's most current context, then calls each hypothesis against the criteria you pre-committed. Scientific tools return verified data from authoritative sources, not generated guesses. Every verdict carries its cited evidence, its confidence, and the conditions that would change the call.
Open any recommendation to its evidence chain. Click any line through to its source, ask why AUO weighted it the way it did, override the verdict and leave your reasoning. Your reasoning becomes part of the record. Ask about your program in plain language; the answer comes from the recorded evidence and verdict trail.
From raw evidence to final call, and the share-out builds itself.
Evidence → claim → verdict → recommendation → your call. The chain is visible end-to-end; intervene at any layer and the record holds. Float a speculative bet, say an adjacent indication or a target you're curious about, and AUO researches it in the background, pulling your team in only when there's enough signal to take it seriously.
When it's time to share, the document is generated from live state: cited, current, on your brand. Ready for the CSO, the board, a partner, or an advisor. Not reassembled from slides.
Verified data. Reproducible analyses. An audit trail behind every claim.
75 of AUO's 80+ scientific skills are wired to live API integrations: Open Targets, ChEMBL, AlphaFold, PubMed, ClinicalTrials.gov, UniProt, BindingDB. Five hand-built specialist flows compose them into deterministic analyses: same inputs, same output, no black-box reasoning.
Every recommendation, override, and comment is written to an append-only verdict log with cited evidence and prior state. A language model does the interpretive reasoning at bounded, recorded points, never the final arithmetic on an explicit-number criterion.
Your workspace content is never reused across customers and never enters any model's training data. Full archive export anytime. No lock-in.
Your judgment operates on a clear picture, not a scattered one.
Better hypotheses reach the clinic. Better drugs reach patients. That is the mission.