The AI research copilot Kairos is now a working tool you can talk to. Kairos Research Copilot takes a natural-language research question — “How can I maximize secreted expression of this protein in Pichia?” — and runs the full inspectable loop against it: retrieval, reasoning, design, and verification, with every step streamed to your screen as it happens.
It is invite-only, and every output is built to be challenged rather than trusted. This is what “provenance, not persuasion” looks like when you can watch it think.
What it does
You provide two things: a research question (required) and an optional target protein sequence. From there, the copilot runs four stages in sequence, each visible in a live timeline:
- Retrieving evidence. Kairos searches across its corpus of full-text papers for cited, relevant evidence. No citations are fabricated — if the retrieval service is unreachable, the tool says so plainly rather than inventing a source.
- Generating candidate designs. If you provided a protein, Kairos produces verifiable coding-sequence candidates on the same four-axis scorecard as every Kairos design run — manufacturability, host-likeness, 5′ initiation, and protein developability.
- Reasoning to strategies. The copilot connects the evidence and the designs into candidate strategies for your question, streaming its reasoning token by token so you can follow the chain of thought in real time.
- Delivering a structured dossier. The final result is a structured research report: evidence-grounded hypotheses, candidate designs with their scorecards, and recommended next steps — each carrying its provenance.
Streaming, not a black box
Most AI tools return a finished answer and ask you to trust the process. The copilot does the opposite. As it works, a live timeline shows each stage lighting up — Retrieving, Designing, Reasoning — with the elapsed time ticking beside it. The reasoning step streams its output token by token, so you watch the argument form rather than appearing fully written.
If something fails — retrieval is down, the API is unreachable, a step errors — the timeline shows the failure and stops. No silent fallback to a plausible-sounding guess.
The loop, made interactive
The copilot is the interactive embodiment of the Evolrix AI research loop — read, reason, design, verify — built so that a researcher can point it at a real question and watch the whole loop run against it in minutes. The evidence is cited. The designs are verifiable. The reasoning is inspectable. The disclaimer is visible.
Honest scope
The copilot produces evidence-grounded hypotheses and verifiable dry-lab design artifacts. It does not predict expression or titer. Every strategy it proposes is a hypothesis to test at the bench — and the dossier says so. The value is in the cited evidence and the verifiable designs, not in a promise the tool cannot honestly keep.
References
- Ahmad M, et al. Efficient Expression of Lactone Hydrolase Cr2zen for Scalable Zearalenone Degradation in Pichia pastoris. Toxins (Basel), 2025. PMID:41591157
- Zha J, et al. Advances in Metabolic Engineering of Pichia pastoris Strains as Powerful Cell Factories. J Fungi (Basel), 2023. PMID:37888283
- Miao L, et al. Metabolic engineering of methylotrophic Pichia pastoris for the production of β-alanine. Bioresour Bioprocess, 2021. PMID:38650288
- Maity N, et al. Statistically Designed Medium Reveals Interactions between Metabolism and Genetic Information Processing for Production of Stable Human Serum Albumin in Pichia pastoris. Biomolecules, 2019. PMID:31590267
- Sharp PM, Li WH. The codon Adaptation Index. Nucleic Acids Res, 1987. PMID:3547335