Kairosby Sightegy

Kairos shows its work.

Kairos is Sightegy's decision-intelligence platform for biotech. It reads every kind of evidence, from genetics and omics to trials, filings, deals and markets, lays out the paths and shows what each one rests on, what is contested, and what Kairos refused to guess. You choose.

Join the beta Inside the engine

The platform

One platform. Three studios.

Founder Studio, Strategy Studio and Venture Cohort run on the same source-traced record, so every reading, brief and base rate shows what it rests on.

01Founder Studio

Read your venture from the other side of the table.

Four chairs read your venture: Founder, Investor, Partnering and Operating. Every finding is marked Evidenced, Modelled, Contested or Not answered, so you see what a reader will push on before they do.

1Tuned to the fund you're pitching
2At most three blockers, in the order a reader notices them
3Briefing prompt: tell your own AI which claims it may not state

Founder Studio/Example Bio/Venture reading

AST-201 · oral allosteric TYK2 inhibitor

Entering Phase 1 · psoriasis, psoriatic arthritis

Illustrative Tuned toSeries A · life-science fund

{{ v.e }} Evidenced{{ v.m }} Modelled{{ v.c }} Contested{{ v.n }} Not answered

{{ s.name }}

Before you send thisin the order a reader notices them

{{ it.n }}{{ it.t }}{{ it.d }}

Export

Dossier PDF

Briefing promptThe prompt to paste into whatever model builds your slides: it carries the claims that model may NOT state, and why.

02Strategy Studio

Decision briefs where every citation opens its source.

Ask the question your board is asking. Science specialists read genetics, expression, pathways and pharmacology next to clinical, regulatory and market evidence, and every agent cites the tools it ran. A brief typically takes 20–40 minutes.

1Synthesize writes the executive summary
2Every citation opens the source's own words, on its page
3Competitive, Forecast, Catalysts, Evidence, and Compare

Strategy Studio/AST-201/Decision brief

Which indication should AST-201 lead with?

Psoriasis or psoriatic arthritis · 4 sources · 3 specialist agents

Illustrative SynthesizeExport PDFCopilot

CompetitiveForecastCatalystsEvidenceCompare

Executive summary

The evidence favours leading with psoriatic arthritis and proving the mechanism in psoriasis first. Human genetics links TYK2 to both indications, but two oral rivals read out before AST-201. Phase 1 finished with no dose-limiting toxicity and supports once-daily dosing: the edge that matters most in psoriatic arthritis, where joint efficacy is still open.

EvidencedHuman genetics supports the target
EvidencedOnce-daily dosing supported
ContestedOnce-daily wins in PsA clinical and commercial agents disagree
Not answeredJoint efficacy in PsA no source states it yet

Copilot

What would change this answer?

A joint-efficacy readout in psoriatic arthritis. Until then, the lead rests on genetics, safety and dosing.

Sources · each opens its passage

03Venture Cohort

Base rates that count the failures.

1,128 Swiss & Nordic ventures, including the 97 that dissolved and the 80 that never incorporated. Survival curves, deal flow and a playbook built from what actually happened, not only from the winners.

1Failures counted, not filtered out
2Failures included survival: every tick a dated failure
3Year replay: move through the cohort's history

Venture Cohort/Swiss & Nordic ecosystem

Every venture, failures included

Illustrative curve

Ventures1,128

Failures counted177

Dissolved97

Never incorporated80

Survival since foundingFailures included

Years since foundingTicks mark dated failures. Ventures still operating are censored, not dropped.

Year replay Year 10 · 77% still operating

Track Record Forecasts scored as they resolve, in the open.

Discovery Studio The fourth studio, on the same record.

How Kairos thinks

Language models read. Models compute. A record keeps score.

Every kind of evidence, from genome to term sheet, goes through the same three steps, and every path Kairos lays out shows which threads it rests on.

Solid: running todayDashed: building next

01Language models

Read

Language models extract dated facts, and each keeps a fingerprint of its source file.
02Models

Reason

Models compute base rates that count the failures, and lay out the paths.

NextA probability that moves only on dated, cited evidence, step by step.
03Record

Record

Every forecast is logged before the outcome, with a due date.

Scored when it falls due

Every kind of evidenceseven families, biology first

Biologygenetics and omics
Clinicaltrials and readouts
RegulatoryFDA and EMA
Corporate & financialfilings and reports
Deals & IPlicensing, M&A, patents
Marketepidemiology, rivals, sales
Venturesrounds and outcomes

Technology

Inside the engine

Five layers, one scroll: how Kairos reads, connects and reasons, at the level we can share.

The reasoning engine

Every probability, explained.

Next · 2027Illustrative

This is the engine we're building next: a base rate that counts the failures, then one visible step per piece of dated evidence. No step without a source. No source after the date you ask about.

AST-201 · probability of Phase 2 success

Start from the base rate. Switch a signal off, or drag the date back, and watch the answer move.

Posterior, as of {{ wfDateL }} {{ wfPostT }}

{{ c.lab }}

As ofonly what was knowable then

Dated signalseach moves the odds by its likelihood ratio

logit(p) = logit(29%) + Σ ln LR_i · only signals dated on or before the as-of date

logit(p) ={{ e.t }}={{ wfEqR }}→p = {{ wfPostT }}

Built to be checked, not taken on trust.

We don't race anyone's accuracy claims on past data. Kairos puts its forecasts on the record before the outcome, counts the failures, and shows the source behind every line.

Pre-registered forecasts

On the record

Every forecast is logged before the outcome, with a due date, and scored when it falls due. Never back-filled.

Base rates that count the failures

1,128ventures177failures counted

97 dissolved80 never incorporated

No invented answers

When no source states it, Kairos answers “Not answered” rather than invent.

A time-split benchmark, published with its method.

Kairos sees only what was knowable on a date, and is scored on what happened after.

What Kairos has read

about 700K

facts extracted from company filings and reports, each traced to its source file.

288,625drug–target–indication records, each with mechanism and phase (Sep 2026)
43,555clinical trials
27,807R&D programmes, indication, stage and status each cited
1,128Swiss & Nordic ventures, failures included
15+public sources
150+tools for your own AI agents

ClinicalTrials.govU.S. FDAOrange BookEMASEC EDGAR filingsCompany reportsOpen TargetsPatentsLicensing and M&A dealsVenture financing histories

Your data stays yours. Your agents are welcome.

Kairos is built for the decisions you cannot afford to leak: workspaces only you can open, and evidence your own AI tools can call.

Owner-only workspaces

Your programmes live in a workspace that only you can open.

Never trains on your data

What you bring stays yours. It is never used to train models.

Connect your own agents

150+ tools. Point your own AI agents at Kairos and they work from the same cited record.

The Briefing prompt

Paste it into whatever model builds your slides. It carries the claims that model may not state, and why.

Next · planned

Hosted beta in Zurich

A hosted beta for design partners, hosted in Switzerland.

Evidence API

Kairos's source-traced evidence, callable by other science agents.

The beta programme

Bring your hardest decision.

Kairos is in beta with design partners. Bring one programme and one real decision. We run Kairos on it with you, and what you need shapes what we build next.

01
Early access
Use Kairos ahead of release, and each studio as it ships.
02
Briefs on your own programmes
Readings and decision briefs on the questions you actually face, in your own workspace.
03
A direct line to the founder
Work with Bogdan Rivoal, who designed and built Kairos end to end.
04
Shape the roadmap
Your decisions set what Kairos learns to answer next.

Join the beta Partner with us Meet the founder