Astral Intelligence (A*I)

AI for Scientific Reasoning

Accelerating discovery by reducing human coordination bottlenecks

Founder: Peter Wang

Prior exit: Sequents

Open with the thesis: A*I is infrastructure for scientific reasoning, not a model company.

Science is not compute-limited. It is reasoning-limited.

• Data abundance, insight scarcity
• Fragmented domains, siloed expertise
• High uncertainty, irreversible decisions
• Tooling optimized for analysis, not thinking

The bottleneck is no longer information — it is alignment.

Frame the market around reasoning flow, not data volume.

What actually slows discovery

Human coordination failure

• Too many competing hypotheses
• Weak synthesis across data and models
• Slow belief updates
• Lost negative results
• Decisions justified after the fact

Reframe AI-for-science as a coordination product.

Why now

Three shifts converged

• Multimodal AI can reason over structure
• Language models can operationalize intent
• Biology crossed a combinatorial threshold

Scientific tools did not fail — human workflows did.

Highlight that A*I leverages the shifts without depending on any single model.

Our insight

AI should not replace scientists. AI should accelerate how scientists think together.

A*I focuses on

 

• Hypotheses
• Evidence
• Uncertainty
• Belief updates

Not about

 

• Black-box predictions
• Autonomous discovery claims
• Replacing scientific judgment

Position A*I as the coordination layer for scientific thinking.

What A*I is

A hypothesis-centric scientific reasoning engine

• Natural language questions
• Explicit hypothesis objects
• Evidence aggregation
• Confidence tracking
• Full reasoning trace

SQL, Python, and ML models live underneath, not on the surface.

Stress that the surface is hypothesis and evidence, not tooling.

What A*I is not

Deliberate constraints

• Not a wet-lab company
• Not promising biological outcomes
• Not replacing scientists
• Not selling black-box predictions
• Not a consumer or hardware product

Build credibility by naming what we will not do.

Mental model

From question to belief update

• Hypothesis made explicit
• Operationalization into analyses
• Evidence gathered
• Confidence updated
• Competing hypotheses compared

The unit of progress is belief convergence, not answers.

Explain that belief convergence is measurable and auditable.

Core primitives and architecture

First-class objects

 

• Hypotheses (versioned, comparable)
• Evidence (data, models, experiments)
• Belief states (confidence + uncertainty)
• Reasoning traces (auditability)

Model-agnostic reasoning substrate

 

• Hypothesis and evidence layer (A*I-owned)
• Tool execution layer (SQL, Python, bio tools)
• Model routing layer (frontier + open-weight)
• Verification and audit layer

Models are interchangeable. Reasoning is not.

Use this slide to pre-empt the “why not pretrain” question.

Why biology first and who buys

The hardest domain

 

• High uncertainty
• Long timelines
• Irreversible decisions
• Expensive mistakes

Initial use cases

 

• Aging and longevity biology
• Target validation in drug discovery
• Cross-domain synthesis
• Scientific decision reviews

Enterprise and institutional buyers

Biotech startups, biopharma R&D, translational research labs, computational biology groups

Emphasize that biology validates the platform for broader domains.

Business model and go-to-market

Enterprise-first model

 

• Paid pilots
• Institutional contracts
• Expansion inside organizations
• Long-term platform licensing

Design-partner GTM

 

• 2-4 serious design partners
• Deep integration
• Measured reasoning acceleration
• References before scale

Make clear this is not a freemium or volume game.

18-month execution plan

 

Four phases that de-risk the next

 

• Phase 0 (Months 0-2): framing and validation
• Phase 1 (Months 3-6): core reasoning engine
• Phase 2 (Months 7-12): design partner pilots
• Phase 3 (Months 13-18): institutional readiness

Each phase de-risks the next.

Mention success metrics: belief convergence, behavior change, readiness for follow-on.

Capital strategy and closing

18-month budget summary

 

Total: ~$3.3-$3.5M

 

Senior small team, cloud-first, spend where trust is built

Funding plan

 

Initial: $500k-$3M to fund Phase 0 → Phase 2

 

Follow-on: up to $6M triggered by institutional readiness

A*I is infrastructure for scientific thinking.

We sell better decisions, not breakthroughs.

If humans can think together faster, discovery follows.

Close with the conviction statement and funding path.