iiinova — where your digital self keeps evolving

But can you animate the drawing part? 😛 added a "draw" button.

1122

Removing inner outline for loops. Beginning to look like watercolor? 😎 (thanks to @martinpacesa.bsky.social for complaining about the loops).

260

Adding support for cyclic L/D peptide mixes.

191

Try for your protein at py2dmol.solab.org (Style → Cartoon)

35214

Sergey Ovchinnikov @sokrypton.org · 16/07/2026

getting close to recreating Jane Richardson's style 😎

412315

A few updates: - multiplayer support expanded to 4 players - marketplace (buy/sell/trade) - outline to show units going behind buildings - mobile friendly (classic mode available via desktop) - battering rams can hold units - various optimizations and speedups

050

Sergey Ovchinnikov @sokrypton.org · 09/07/2026

One thing that helped with was implementing a simulated gameplay. Have Claude run 100s of games, get feedback and update ai.js. It is still doing some silly things... but getting better. Perhaps a future direction would be to let multiple LLM compete with eachother by building better ai.js?🤠 (3/4)

251

Sergey Ovchinnikov @sokrypton.org · 09/07/2026

🍹Long weekend Project: Since Claude Fable is banned for Science, I thought it might be fun to see if it can be used for something less scientific. 😎 Introducing Age of Epochs! ⚔️ An attempted reproduction of Age of Empires II in Javascript. ageofepochs.com (1/4)

2212

Sergey Ovchinnikov @sokrypton.org · 15/06/2026

To get OpenFold3 (pytorch weights) working in AlphaFold3 (jax code base) was not super easy. Required an entire weekend and $40 of Claude Code credit. See breakdown of what had to change in AF3 code base to get OF3 weights to work: github.com/sokrypton/al... Commit: github.com/sokrypton/al... (2/3)

140

Sergey Ovchinnikov @sokrypton.org · 15/06/2026

New Experimental Google Colab Notebook now integrates AlphaFold3 with OpenFold3 and py2Dmol: colab.research.google.com/github/sokry... (1/3)

15515

Adding support for interactive MSA viewing, including coloring by entropy! Auto download from AFDB (for both uniprot and pdb entries). (4/4)

0101

Save vectorized SVG for infinitely ♾ zoomable figures. (including support for contacts). See example: biorxiv.org/content/10.1... Zhidian Zhang @yoakiyama.bsky.social @yehlincho.bsky.social @jajoosam.bsky.social (3/4)

161

Adding record button to save animation. (2/4)

120

Sergey Ovchinnikov @sokrypton.org · 29/10/2025

A few py2Dmol updates 🧬 py2dmol.solab.org Integration with AlphaFoldDB (will auto fetch results). Drag and drop results from AF3-server or ColabFold for interactive experience! (1/4)

110331

Is 3D dragging you down? Wish you could instead use the 2D ColabFold representation for all your work? 🤓 Introducing: py2Dmol 🧬 (feedback, suggestions, requests are welcome)

1294

Sergey Ovchinnikov @sokrypton.org · 28/10/2025

Working on the protein-hunter-chai google colab notebook. 😈 @yehlincho.bsky.social

0335

Sergey Ovchinnikov @sokrypton.org · 27/10/2025

Will it bind? A little worried about all the "TTTTTTT" 🧐 But looks cool 😎

1171

Sergey Ovchinnikov @sokrypton.org · 11/09/2025

Looks like someone has already tried to replace me with an AI agent 🫣

1220

We find this to be true across a number of targets. Where method used to pair sequences and filter them makes a big difference. This we find to be important when trying to disentangle paralogs from orthologs (4/4).

260

The big difference is in the pairing. The MMseqs2 server pairs sequences based on species, while our old HHblits MSAs were paired based on genome proximity (number of genes apart). Working w/ @milot.bsky.social and @martinsteinegger.bsky.social we implemented the proximity filtering in server (3/4)

1113

Side story: While working on the Google Colab notebook for MSA pairformer. We encountered a problem: The MMseqs2 ColabFold MSA did not show any contacts at protein interfaces, while our old HHblits alignments showed clear contacts 🫥... (2/4)

1133

Excited to re-share work from @yoakiyama.bsky.social and Zhidian Zhang on MSA pairformer. (1/4)

1265

Little more action (if you use "creative" mode), though protein gets a bit unrealistic.

230

PS, I found Vidu to work better for interprolation between images. Example attached:

121

Since you asked, tried to retrace my steps:

230