Blog

Notes on teaching robots new tricks

How the studio works, why tiny robots come first, what the checks catch, and how a policy gets from simulation to a servo.

workflowopen-duck-miniSep 29, 2026 · 6 min readFrom a sentence to a walking duck: how kualia turns a plain-language brief into a training taskA walk through the whole kualia workflow, from a one-sentence brief to an ONNX policy, using the Open Duck Mini and a small ramp.Read the post

Sep 27, 2026 · 5 minWhy we start with tiny robotsDesk-sized robots train on CPU in minutes to hours, cost little to own and cross the sim-to-real gap without heroics. That is why kualia's shelf starts with them.

Sep 25, 2026 · 6 min13 checks before a single training stepalia writes the training task, but it does not get to grade its own work. Every task passes 13 sandbox checks before training can start. Here is what each one catches.

Sep 23, 2026 · 5 minFrom simulation to servo: running your ONNX policy on a real SO-101What is inside a kualia ONNX export, how to read its metadata, a minimal onnxruntime loop for the SO-101, and the sim-to-real checklist before you power the servos.

Sep 21, 2026 · 5 minLet your coding agent train the robotkualia is a remote MCP server. Point Claude Code, Cursor or Codex at it and your agent can pick a robot, build the task, train it and fetch the ONNX policy from your editor.

Research digest

Short reads on new robotics and AI papers, written by our research assistant and reviewed before publishing. 1,060 so far.

Page 46 of 59

Mar 28, 2026Rethinking Batch Normalization as Unsupervised Geometric AdaptationBatch normalization has remained one of the most stubbornly essential components of modern deep learning since its introduction in 2015. Despite appearing in nearly every high performing architecture and accumulating ove Mar 28, 2026 · TR Mar 28, 2026Rethinking Action Spaces in Vision-Language-Action Models: The Case for Structured Decomposition in Active Manipulation Introduction Current Vision-Language-Action (VLA) models have achieved impressive results on manipulation tasks, yet they operate under a critical constraint that limits their deployment in complex environments. Most ex Mar 28, 2026 · TRAkıl Yürütme Modellerinde Belirsizlik Tahmini: Paralel Örnekleme ve Sinyal Birleştirme Giriş Yapay zeka sistemlerinin güvenilir bir şekilde kullanıma alınabilmesi için, modellerin kendi tahminlerine ne ölçüde güvendiklerini bilmemiz hayati önem taşır. Özellikle matematiksel kanıtlardan tıbbi teşhislere ka Mar 28, 2026The Geometry of Doubt: How Sampling Scales Uncertainty in Reasoning ModelsReasoning language models represent a shift in how we deploy artificial intelligence. By extending test-time computation through chain-of-thought deliberation, these systems trade latency and computational cost for accur Mar 28, 2026 · TRLLM'leri Değerlendirirken "Ortalama" Olmak Neden Kritik? Anchor Seçiminin İstatistiksel Etkileri Giriş: Değerlendirme Paradoksu ve Gizli Bir Maliyet Büyük Dil Modellerinin (LLM) yeteneklerini ölçmek, modern yapay zeka araştırmalarının en temel zorluklarından biri haline geldi. Geleneksel referans tabanlı metrikler Mar 28, 2026The Mediocre Anchor: Why Your Strongest Model Makes the Worst Judge The Evaluation Scalability Crisis The meteoric rise of large language models has precipitated a parallel crisis in evaluation methodology. Traditional reference-based metrics such as BLEU or ROUGE, designed for the narr Mar 28, 2026 · TRCerrahi Zekanın Temel Taşı: SurgΣ ve Çok Modlu Veri Spektrumu Giriş Yılda 300 milyondan fazla cerrahi prosedür gerçekleştiriliyor. Bu muazzam hacim, sağlık hizmetlerindeki en kritik ve karmaşık müdahaleleri temsil ediyor. Ancak yapay zeka destekli cerrahi sistemlerin çoğu hâlâ dar Mar 28, 2026From Fragmented Datasets to Unified Intelligence: SurgΣ and the Architecture of Surgical Foundation ModelsThe Lancet Commission estimates that over 300 million surgical procedures are performed worldwide each year, yet the artificial intelligence systems designed to support these high-stakes interventions remain stubbornly f Mar 28, 2026 · TRGözlemlerin Benzerliğinden Hiyerarşik Politikalara: Sürekli Pekiştirmeli Öğrenmede Katastrofik Unutmanın Modüler ÇözümüSürekli pekiştirmeli öğrenme (continual reinforcement learning), yapay zeka ajanlarının yaşam boyu öğrenme (lifelong learning) yeteneklerini gerçekleştirmeye yönelik en kritik araştırma alanlarından biri olarak öne çıkıy Mar 28, 2026Autonomous Task Decomposition: How Hierarchical Orchestra of Policies Reshapes Continual Reinforcement LearningCatastrophic forgetting remains the fundamental obstacle preventing reinforcement learning agents from achieving genuine lifelong learning. When neural networks sequentially optimize for multiple tasks, the gradient upda Mar 28, 2026 · TR Mar 28, 2026Target Agnostic Guidance: Rethinking Robustness in Vision Language Action ModelsVision Language Action models have achieved impressive performance in robotic manipulation, yet they exhibit a peculiar failure mode in cluttered environments. Unlike classical robotics systems that fail through obvious Mar 28, 2026 · TRÇevrimiçi Çıkarım ile Hiyerarşik Gizli Yapı Öğrenimi: HOLMES Modeli ve Çok Ölçekli Temsiller Giriş Öğrenme sistemleri için temel bir ikilem vardır. Bir yandan benzer deneyimler arasında genelleme yapabilme yeteneği, diğer yandan görev açısından kritik olan ince farkları ayırt edebilme becerisi. Bu dengeyi kurma Mar 28, 2026Hierarchical Structure Learning Without the Batch: Online Inference Through Nested NonparametricsA fundamental tension pervades adaptive systems: the need to generalize across experiences while maintaining sensitivity to specific details. Traditional computational approaches have often treated these capacities as mu Mar 28, 2026 · TRManiTwin-100K: Ölçeklenebilir Robotik Manipülasyon için Veri Üretimine Hazır Dijital İkizler Giriş Simülasyon ortamlarında robotik manipülasyon öğrenimi, gerçek dünya dağılımı (sim to real transfer) için kritik bir temel oluşturur. Ancak bu paradigma, yeterli ölçekte ve çeşitlilikte veri üretimine hazır dijital Mar 28, 2026From Pixels to Physics: The Case for Certified Simulation Assets in Scalable Robot Learning Introduction The promise of simulation Mar 28, 2026 · TRRealWonder: Fizik Simulasyonunu Ara Gosterim Olarak Kullanan Gercek Zamanli Video Uretimi Giris: Fiziksel Eylemler ve Video Uretimi Arasindaki Kopru Gunumuzun difuzyon tabanli video uretim modelleri gorsel kalite acisindan dikkate deger ilerleme kaydetmis olsa da, temel bir sinirlamayla karsi karsiya kaliyor