Aster is an autonomous research lab built around thousands of agents working in parallel. It fits founders and research teams testing whether open-ended investigation can be decomposed into machine-run experiments.
What it does
Aster’s YC profile describes a system that orchestrates thousands of research agents toward a single goal. The company’s current public research includes work on ProteinGym, NanoChat, NanoGPT and logic-gate language models.
The important distinction is between benchmarked research and open-ended research. Aster’s YC launch says its system set a ProteinGym result in 30 minutes, 57 times faster than single-agent systems and at one-third the cost for the same result. Those are company-reported claims tied to a named experiment, not a general claim that it has automated science.
The current homepage says Aster is a Public Benefit Corporation focused on autonomous research. It publishes research writeups and presents the system as a way to compress long research cycles. The company’s public pages do not establish that every research problem can be expressed as a benchmark or that the results translate directly to commercial R&D.
Pricing and fit
Aster’s inference pricing lists Free, Pro at $20/month with $30 in tokens, Max at $100/month with $150 in tokens and Enterprise custom. The pricing page is for inference access, not a complete price for an autonomous research program.
That makes Aster interesting for two different buyers. A technical researcher may want to explore the inference system. A company may want a research partner for a problem where parallel search and experimentation are valuable. Those are not the same purchase.
The founder profile describes Emmett Bicker as a former Magic researcher who moved into autonomous research. The company’s public material is ambitious and unusually specific about its experiments; it still needs to earn trust one research task at a time.
What could make it the wrong choice
Open-ended research has a nasty failure mode: an agent can generate more hypotheses than a human team can validate. A benchmark result can show that the loop works on one defined task without showing that the system can choose a commercially important question, design a valid experiment and recognize a dead end.
My editorial take
I would shortlist Aster for a research team with a clear experimental surface and enough technical expertise to inspect the process. I would not treat the public speed or cost claims as a substitute for evaluating the actual research question. The product’s edge is parallel exploration; the buyer still owns scientific judgment.
Quick facts
| Field | Sourced detail |
|---|---|
| Product | Autonomous research system and inference platform |
| Buyer | Technical research teams and builders of AI systems |
| Public pricing | Free; Pro $20/month; Max $100/month; Enterprise custom |
| Public claims | ProteinGym result and speed/cost comparisons, company-reported |
| Main gate | A research problem with inspectable evaluation and human review |
Sources checked
| Source | Checked |
|---|---|
| YC profile | 2026-09-19 |
| Aster homepage | 2026-09-19 |
| Aster pricing | 2026-09-19 |
| YC launch | 2026-09-19 |
Cohort context
Aster is listed in Spring 2026. In our 2026-09-18 directory snapshot, 112 of 193 listed companies in that cohort have YC’s primary industry label B2B (58.0%). This is a current-directory comparison, not an original intake count or a performance ranking. Nine-cohort dataset.
Public website snapshot
Observed 2026-09-19 16:16 UTC in raw homepage HTML. This records visible metadata and advertised links, not agent execution or product quality.
| Signal | Homepage observation |
|---|---|
| Product description metadata | Observed |
| Canonical link | Not observed in this response |
| H1 or H2 heading | Observed |
| Typed structured data | Not observed in this response |
| Docs/developer link | Not observed in this response |
| Pricing link | Observed |
| llms.txt link | Not observed in this response |
| Markdown alternate | Not observed in this response |
Public observations · Collection method. Missing links here do not establish that a capability or file is absent elsewhere.
