CellType: 1.66M views, 100 creators and an X news story

At a glance

Client

CellType (celltype.com), Y Combinator W26

Agency

GrowthMatrix, influencer marketing for AI and tech launches

Category

Biology AI: models that predict how cells respond to drugs

Launch

4 November 2026, 9:26am US Eastern, BioSimBench launch post from the company account on X

Platform

X / Twitter

Format

100 creator quote reposts on the launch post

Result

1.66M views on the launch post, plus 1.83M on the creator quotes (about 3.5M combined)

On X

A Today's News story: "BioSimBench Reveals AI Limits in Predicting Cell Responses"

About CellType

CellType builds AI models that simulate human biology, so drug teams can predict how a compound will work before they run it in the lab. Its site puts it in one line: "Simulate biology. Predict how your drugs work." The models are pre-trained on single-cell, spatial and bulk RNA data, then fine-tuned on a customer's own lab data and run on-prem or in their own cloud.

• Backed by: Y Combinator, Winter 2026 batch (W26)
• Team: co-founder and CEO David van Dijk, a Yale professor, in New York
• Science: built by the inventors of Cell2Sentence at Yale, made in collaboration with Google DeepMind

What CellType launched

BioSimBench is a benchmark of real lab experiments. A model gets a cell and a change to it (a drug, a gene knockdown or an immune signal) and has to predict which genes go up and which go down. The answers are checked against what the lab measured.

CellType tested 12 frontier models. The best scored 0.318 on new drugs and cell contexts, where two lab repeats of the same experiment agree at 0.462. CellType also launched BioSimGym, training tasks built from real lab experiments, and a 27B model trained in it that scored 0.366 and led every frontier model on all four prediction tracks. The launch was built on NVIDIA tools, and the post put it in one question: "Can LLMs simulate biology?"

What we ran

The launch came from CellType's company account, which had 2,213 followers. GrowthMatrix matched 100 AI and tech creators to the launch. Each one quote-reposted CellType's post in their own words and picked the angle their audience would care about: a 27B model beating the frontier models, a benchmark graded by a real wet lab, a simple explainer for people outside biology, or what it could mean for drug discovery.

The quotes were timed to the launch. The first went live 5 minutes after CellType posted, 66 landed in the first hour and 89 within two hours. A last group went out four to five hours in, as the US afternoon started.

It trended on X as a news story

By the next day, X had turned the launch into a news story. Its Today's News panel listed "BioSimBench Reveals AI Limits in Predicting Cell Responses", with 3,754 posts in the conversation. That is far more than the 100 creator quotes, so the talk went well past the people we briefed.

Results

Public counts: the launch post as of 10 November 2026, the creator quotes as of 5 November 2026. One of the 100 creator posts was no longer public that day, so the creator row adds up the other 99.

Creator quotes from launch day

Three of the 100 creator quote posts on CellType's launch day, each showing a different moment of the launch video. We blur creator names on our site.

Views on each quote post at capture, 5 November 2026.

How to read these numbers

1.66M views from an account with 2,213 followers means almost every view came from outside CellType's own audience. The creator quotes added 1.83M views of their own, each one pointing back to the launch post, which is how the total reached about 3.5M. 775 bookmarks show people saving a technical launch to read properly later, and the X news story shows the topic broke out of the creators' timelines into X's own news feed.

What the client said

When this approach fits your launch

• You're launching something technical, like a benchmark, a model or a research result, and need people outside your field to get it.
• You launch from a company account with a small following.
• You have one clear result to point at, like a leaderboard or a demo video.

Launching an AI model? See launching AI models on X and our developer marketing agency page. Launching out of a YC batch? Read how YC startups launch. The full day-by-day plan is in the X Launch Playbook, and how we judge a result is in how many views is viral on X.