iiinova — where your digital self keeps evolving

Reposted by Mario Stanke

Prof. Konrad Waldorf @konradwaldorf.bsky.social · 07/10/2026

21. September: OpenAI richtet eine Beratergruppe in Princeton ein, und meldet 100+ weitere gelöste Probleme – intern ringt man darum, „wie man es announced, ohne die Community erneut zu erzürnen". Gestern dann der nächste Schritt: 722 Manuskripte, 372 Problemfamilien.

171

Mario Stanke @masta.bsky.social · 07/10/2026

The proof is here: github.com/openai/math Along with 721 others. Math is also doomed.

000

Reposted by Mario Stanke

Katharina Hoff @katharinahoff.bsky.social · 03/10/2026

I am still very excited about the unsupervised gene prediction approach Vipsania, made by @masta.bsky.social , Richard Krieg & Joscha Diehl. So excited that I made you a little web service. You can now run Vipsania via your web browser. bioinf.uni-greifswald.de/vipsania/

bioinf.uni-greifswald.de

Vipsania Web Server - Unsupervised Ab Initio Gene Prediction

Vipsania - unsupervised ab initio eukaryotic gene prediction web server (pretrained models, no finetuning)

054

Mario Stanke @masta.bsky.social · 10/09/2026

With his program, Vipsania, Richard Krieg has shown, that it is possible to do unsupervised deep gene finding with masked language modeling The trick: use an HMM in the middle of the deep-learning architecture that knows the grammar of gene structures. preprint: www.biorxiv.org/content/10.6...

022