Derek Alia

@biogerontology Clinical is going to take a while, but biology and chemistry are coming along quite well.

Prompt to Drug PSI is now possible with orchestration. pubs.acs.org/doi/abs/10.102…

@natolambert @willdepue slowly realizing this.

@willdepue Tradition doesn't show that these labs are particularly open and actually doing science these days, culture takes a very long time to change.

academics are unprepared for the coming world where much scientific progress is majorly a function of inference compute. whether OpenAI points the Eye of Stargate at your particular field will decide its acceleration. talent will leach away into the labs. it's already begun

@binji_x @Lighter_xyz @variational_io @arbitrum @grvt_io @zksync @Euphoria_fi @megaeth @Ostium It took Ethereum 6 years to figure out that perps were important to have on their chain.

Teams building perps on Ethereum: > @Lighter_xyz (Ethereum L2) > @variational_io (on an Ethereum L2 - @arbitrum) > @grvt_io (Ethereum L2 - @zksync stack) >@Euphoria_fi (on an ethereum L2 - @megaeth) > @Ostium (on an ethereum L2 - @arbitrum) Ethereum is for everything.

@gerardsans @OpenAI @xai Instead of vague posting, how about you read the post? Then tell me how this is not real world data.

@derekalia @OpenAI @xai Be wary of cherry picked lab evaluations. Request real world data.

Congrats to @OpenAI and @xai for their scientific reasoning. Last week we got the wet-lab results back from our TREM2 hackathon. 6 autonomous agents + 9 human teams designed TREM2 binders in a single day. Agents nearly matched human hit rates. read more: muni.bio/research/agent…

@draparente @OpenAI @xai Muni’s inference provider is openrouter and it’s quite expensive to run opus and GPT pro for long running sessions. Now that we have the muni cli it would be more cost-effective to use something like opus4.7 or gpt5.5.

@derekalia @OpenAI @xai why sonnet/GPT rather than opus/gpt pro?

@jia_seed Congrats on the launch. LFG!!!!

here's jam it's the simplest interface for marketing in the agentic era we now have 2000+ companies, developers, and operators enjoying jam we started it as a side project, then devtool founders paid us thousands to use it it's come a long way since, and we've just released it for anyone to try, no waitlist and also we are now hiring (see below!) you build, jam spreads

@BrantlyMillegan Isn't it funny? the guy who got kicked out of ENS because of his values, is one of the few solid devs that is still building in the space because of his...values

i haven't sold my ETH and i'm still here building on Ethereum 🫡

Another great read from @adaptyvbio about our hackathon adaptyvbio.com/blog/agents-vs…

Anthropic is paying SpaceX $1.25B/mo for compute 🤯

@lvwerra Great job with the visuals. Looking forward to playing with the model.

We are releasing Carbon: a crazy fast DNA model Carbon is 275x faster than the next best model. So fast you can process the whole human genome on a single GPU in <2 days. Here are the tricks we used: When modelling DNA sequences a lot of the performance comes down to tokenizing the sequences in a smart way. BPE tokenizer struggle because there are no whitespaces and character (called base in DNA) level tokenizers waste a lot of compute on too many tokens. Carbon is built with a unique tokenizer: we split sequences in chunks of 6 bases, but during both training and inference we can work with single base resolution. That's similar to having word tokens but resolving them at the character level. All possible thanks to the DNA tokens unique structure. The architecture combined with the tokenizer makes the model 275x faster than the previous SoTA (Evo2) at this size. We built an interactive demo so you can explore how the model can generate DNA sequences, investigate the structure of genes, predict the effect of mutations, generate and fold proteins and even reconstruct parts of the tree of life. huggingface.co/spaces/Hugging…

@WUFang40615703 Okay, this is an awesome idea.

Proteo-R1 (ICML 2026), the first reasoning protein foundation model for protein design, is out! 🚀🧬 Most protein design models generate structures without ever *reasoning* about which residues matter. We think that's backwards. Human protein engineers👩‍🔧 don't work this way. They identify critical interaction residues first — charged anchors, hydrophobic hotspots, specificity-determining motifs — and only then optimize geometry around those decisions. ━━━━━━━━━━━━━━━━ 🔬 THE CORE IDEA ━━━━━━━━━━━━━━━━ A dual-expert architecture that explicitly decouples molecular understanding from geometric generation: → ⚡A multimodal LLM (understanding expert) analyzes protein sequences, structures, and text to identify key functional residues governing binding and specificity → ⚡A diffusion model (generation expert) then co-designs sequence + structure — but with those residues locked in as hard constraints ━━━━━━━━━━━━━━━━ 📐 HOW IT'S TRAINED ━━━━━━━━━━━━━━━━ Three-stage curriculum: ① Multimodal Alignment — freeze the LLM, train projections to bridge ESM-2 + AF3-style structural features into language space ② Structural Reasoning Mid-Training — unfreeze the LLM, teach it residue grounding → pairwise geometry → interface localization → hotspot prediction ③ Joint Reasoning-Guided Design — end-to-end on antibody-antigen complexes. Gradients from the diffusion objective flow back through the reasoning expert. ━━━━━━━━━━━━━━━━ 📊 RESULTS ━━━━━━━━━━━━━━━━ Evaluated on simultaneous multi-CDR redesign and the RAbD CDR-H3 benchmark: ✅ Best RMSD & DockQ on RAbD — redesigned H3 loops are geometrically accurate *and* docked well ✅ Lowest backbone dihedral divergence (JSDbb) among all baselines ✅ Reduced intra- and inter-chain steric clashes ✅ Generated sequences score lower perplexity than native antibodies under IgLM & AbLang ✅ Plug-and-play: swapping the diffusion backend to UniMoMo still improves RMSD and IMP ━━━━━━━━━━━━━━━━ 💡 WHY IT MATTERS ━━━━━━━━━━━━━━━━ Proteo-R1 isn't just a better antibody design model. It's a blueprint for coupling deliberative LLM reasoning with any physical generative process — interpretable, modular, and backend-agnostic. 📄 Paper: arxiv.org/abs/2605.02937 💻 Code: github.com/smiles724/Prot… 🌐 Demo: smiles724.github.io/r1/ Great thanks to my wonderful collaborators Weihao Xuan, Heli Qi, @Hanqun_CAO, Heng-Jui Chang, @KKuanPang @XiangruTang Zehong Wang, @hcwww_ , @KejunYing @lupantech Chiho Im, Seungju Han, @richardxp888 @tikgiau. Also appreciate the guidance from advisors @YejinChoinka @jure @erranlli Naoto Yokoya, Masashi Sugiyama.

@SynBio1 100% agree

We're seeing the "unbundling" of the research and teaching functions of the modern university More federal money moving to institutes that don't teach (and thus have lower overheads) I'm not opposed to experiments like this per se. Would probably be better to have them run by an admin that doesn't hate science and scientists

@brockjelmore put a bounty on lyme disease

who is building tick extinction technology i will fund you

very cool

Just started using Grok Build and the first thing you'll notice is the speed. It's so fast 👀 I'm so used to sending a request and then switching context and then coming back. I think that speed helps you stay in the loop of solving the task at hand. Which is a game changer. gg

@InfraScaler @alexocheema @CrazySoft he did not say that

@alexocheema @CrazySoft Sorry. I may be misunderstanding you. You're saying bandwidth and latency are not important? Do you have data to back this up?

4 x M5 Max MacBooks clustered with RDMA: 512GB @ 2456GB/s, $20k, 560W, quiet. Find me a better deal, that I can buy today.

@design_proteins sir this is a wendy's

/goal design a peptide that reverses hairloss, use cloud labs to validate safety and efficacy, line up licensing deals with 3+ cosmeceutical partners to cover your validation costs, improve peptide until $1M ARR. Go.

Minority Report vibes here -> "These “sentinel” cells can be used to figure out what a person looks like, solely by storing the trace amounts of DNA they leave behind in a room."