Last updated Nov 29, 2025
aiscience
In 2024, at least one notable new drug, material, or production method discovered using AI‑driven predictive models (rather than traditional brute‑force lab discovery) will reach a milestone significant enough to be featured as a "science corner" topic on the show.
I'm excited about seeing what comes to market this year. I'm sure we're going to have a science corner at some point this year that says, look at this amazing new thing that was discovered in software, and it works, and it's going to be really cool.View on YouTube
Explanation

Friedberg’s prediction was that in 2024 there would be “a science corner … that says, look at this amazing new thing that was discovered in software, and it works,” i.e., a notable new drug/material/production‑method–type technology discovered using AI models rather than traditional brute‑force lab work.

In 2024, the All‑In podcast did exactly this:

  • On May 3, 2024, they released the episode “In conversation with Sheryl Sandberg, plus open-source AI gene editing explained,” whose Science Corner segment is explicitly titled “Science Corner: Open-source AI gene editing with OpenCRISPR-1.” (zeno.fm)
  • That segment links to Profluent’s OpenCRISPR-1 GitHub and work. (zeno.fm) OpenCRISPR‑1 is described by Profluent as an AI-created gene editor—a Cas9‑like protein and guide RNA fully developed using large language models trained on massive CRISPR sequence datasets, rather than through traditional discovery or directed evolution. (profluent.bio)
  • Profluent’s 2024 announcements and technical write‑ups report that OpenCRISPR‑1 successfully performs precision editing of the human genome in cells, with activity comparable to or better than SpCas9 and improved specificity, and that it has been released as an open-source, licenseable tool for research and commercial use. (profluent.bio) Independent coverage (e.g., Chemical & Engineering News) characterizes this as a generative‑AI‑designed, new‑to‑nature gene editor created by protein language models. (cen.acs.org)

This clearly satisfies the normalized prediction:

  • AI-driven predictive models, not brute-force discovery: OpenCRISPR‑1 was generated by LLM-based protein design models trained on vast CRISPR sequence data, then experimentally validated—exactly the “discovered in software, then shown to work” pattern Friedberg described. (profluent.bio)
  • New therapeutic/biotech modality: While it’s a gene-editing tool rather than a finished pill, it is a novel, AI-designed molecular system intended for therapeutic and biomanufacturing applications, analogous in importance to a new drug platform. (profluent.bio)
  • Featured as a Science Corner in 2024: The All‑In episode explicitly labels the segment as a Science Corner about “open-source AI gene editing with OpenCRISPR‑1,” within the 2024 calendar year. (zeno.fm)

Because at least one such AI-discovered biological technology reached a major experimental and commercialization milestone in 2024 and was indeed featured as a Science Corner topic on the show, Friedberg’s prediction is right.