Earlier this year, DeepSeek shocked Silicon Valley. This relatively unknown Chinese AI company released models that matched GPT-4 and Claude... Then open-sourced everything. No strings attached. Just world-class AI technology, free for anyone to use. The message was clear: proprietary moats in AI are shallow at best, and measured in months, not years. (If you're new to Ariax and our mission to democratize protein design, start with our introduction here.)

The same revolution is happening in protein design, only more dramatically. And it's about to transform who can develop the next generation of medicines.

The Open-Source Advantage in Protein Design

When Google DeepMind released AlphaFold3 for protein structure prediction, the open-source community responded within six months with Boltz-1—matching its performance while being completely free and unrestricted to use. But in protein design, open-source tools don't just match proprietary platforms. They dominate them.

Consider the tools transforming the field today: RFdiffusion, ProteinMPNN, ColabDesign, and BindCraft. All released under permissive licenses (including ColabDesign's delightful "Beer-Ware License"). While several companies tout proprietary AI platforms for designing new proteins and antibodies, none have convincingly demonstrated superiority over these open tools.

The proof? In Adaptyv Bio's international protein design competitions—the field's most rigorous public benchmarks—open-source tools consistently win. In the most recent competition, BindCraft swept the de novo protein category. Of the seven completely novel proteins that successfully bound their target, six came from BindCraft.

Why BindCraft Changes Everything: The Open-Source Protein Design Tool That Works

During my years in protein engineering, I've seen countless "revolutionary" tools come and go. Most add complexity without meaningful improvements over traditional wet-lab approaches like directed evolution with yeast or phage display. In fact, many of these earlier computational tools for in-silico protein design ultimately required the same high-throughput screening infrastructure they claimed to replace, significantly undercutting their practical utility.

BindCraft is different. It's the first design tool I've used myself that genuinely democratizes the creation of novel protein therapeutics.

Here's what makes it special:

It actually works. In the original bioRxiv preprint, 10-100% of BindCraft designs bound their targets with micromolar or better affinity. That's not a typo. Compare this to traditional approaches where you might screen thousands of candidates to find a single binder. This "few-shot" success rate means any lab, not just those with massive screening infrastructure, can now design functional proteins.

It's brilliantly automated. Traditional protein design pipelines require a specialist to manually chain together multiple tools: generate structures with RFdiffusion, optimize with ProteinMPNN, score with AlphaFold, rinse and repeat. BindCraft automates this entire workflow. Set your parameters, start the program, and it runs until you have the desired number of designs. It's like having a postdoc in computational protein design compressed into a single script.

It makes the right technical choices. Under the hood, BindCraft uses AlphaFold2 backpropagation through ColabDesign for initial design—more computationally intensive than alternatives, but with higher success rates. It then optimizes with ProteinMPNN and applies rigorous scoring with both AlphaFold and Rosetta metrics. Every decision prioritizes experimental success over computational convenience.

The BindCraft automated workflow: from target protein to filtered, validated binder designs. Source: BindCraft preprint

This is the right trade-off: computational costs are dwarfed by experimental costs. DNA synthesis and cloning of even a handful of designs easily surpasses the compute cost, not to mention the researcher time and the painful reality of failed experiments. BindCraft's approach—spend more compute upfront to dramatically increase experimental success—is exactly what the field of generative protein engineering needs.

From Command Line to Click-and-Go

But here's the catch: even with BindCraft's automation, you still need command-line expertise, a properly configured environment, and access to expensive GPUs. That's where Ariax comes in.

We've taken BindCraft and deployed it in a simple web interface. No installation. No configuration. No fighting for cluster time or navigating the labyrinth of AWS to provision a GPU instance. Just upload your target structure, set your parameters through a friendly form, and watch your designs generate in real-time. We handle all the infrastructure complexity so you can focus on the science. (Check out our BindCraft quick-start guide to see how easy it is.)

The pricing model is equally compelling. A typical BindCraft run on Ariax costs $100-500: less than the reagents for a single round of traditional antibody screening. And here's the kicker: it's actually cheaper than running it yourself on AWS, Azure, or Google Cloud. You're not paying for proprietary algorithms or platform licenses. You're just paying for the compute time, and we've endeavored to make that as affordable as possible.

The Future is Open

BindCraft represents what I believe is the future of AI-powered biotherapeutics: powerful, open-source tools that any researcher can access and use. It's the perfect embodiment of our mission at Ariax—taking the best computational biology tools and making them accessible to everyone.

And this is just the beginning. The protein design community is rapidly developing similar platforms for other therapeutic modalities: monoclonal antibodies, nanobodies, cyclic peptides. As soon as these tools prove themselves and are released openly, you can count on Ariax to offer them under the same terms. No subscriptions. No commitments. Just pay for the compute you use.

The DeepSeek moment showed us that AI's future is open. BindCraft shows us that drug discovery's future is too. At Ariax, we're here to make sure every researcher—from grad students to biotech startups—can be part of that future.

Ready to run BindCraft yourself? Create your free account on Ariax→