The Inefficiency of Static Chain-of-Thought Reasoning in LRMs Recent LRMs achieve top performance by using detailed CoT reasoning to solve complex tasks. However, many simple tasks they handle could be solved by smaller models with fewer tokens, making such elaborate reasoning unnecessary. This echoes human thinking, where we use fast, intuitive responses for easy problems and slower, analytical thinking for complex ones. While LRMs mimic slow, logical reasoning, they generate significantly longer outputs, thereby increasing computational cost. Current methods for reducing reasoning steps lack flexibility, limiting models to a single fixed reasoning style. There is a growing need for adaptive reasoning that adjusts effort according to task difficulty.  Limitations of Existing Training-Based and Training-Free Approaches Recent research on improving reasoning efficiency in LRMs can be categorized into two main areas: training-based and training-free methods. Training strategies often use reinforcement learning or fine-tuning to limit token usage or adjust reasoning depth, but they tend to follow fixed patterns without flexibility. Training-free approaches utilize prompt engineering or pattern detection to shorten outputs during inference; however, they also lack adaptability. More recent work focuses on variable-length reasoning, where models adjust reasoning depth based on task complexity. Others study “overthinking,” where models over-reason unnecessarily. However, few methods enable dynamic switching between quick and thorough reasoning—something this paper addresses directly.  Introducing OThink-R1: Dynamic Fast/Slow Reasoning Framework Researchers from Zhejiang University and OPPO have developed OThink-R1, a new approach that enables LRMs to switch between fast and slow thinking smartly, much like humans do. By analyzing reasoning patterns, they identified which steps are essential and which are redundant. With help from another model acting as a judge, they trained LRMs to adapt their reasoning style based on task complexity. Their method reduces unnecessary reasoning by over 23% without losing accuracy. Using a loss function and fine-tuned datasets, OThink-R1 outperforms previous models in both efficiency and performance on various math and question-answering tasks.  System Architecture: Reasoning Pruning and Dual-Reference Optimization The OThink-R1 framework helps LRMs dynamically switch between fast and slow thinking. First, it identifies when LRMs include unnecessary reasoning, like overexplaining or double-checking, versus when detailed steps are truly essential. Using this, it builds a curated training dataset by pruning redundant reasoning and retaining valuable logic. Then, during fine-tuning, a special loss function balances both reasoning styles. This dual-reference loss compares the model’s outputs with both fast and slow thinking variants, encouraging flexibility. As a result, OThink-R1 can adaptively choose the most efficient reasoning path for each problem while preserving accuracy and logical depth.  Empirical Evaluation and Comparative Performance The OThink-R1 model was tested on simpler QA and math tasks to evaluate its ability to switch between fast and slow reasoning. Using datasets like OpenBookQA, CommonsenseQA, ASDIV, and GSM8K, the model demonstrated strong performance, generating fewer tokens while maintaining or improving accuracy. Compared to baselines such as NoThinking and DualFormer, OThink-R1 demonstrated a better balance between efficiency and effectiveness. Ablation studies confirmed the importance of pruning, KL constraints, and LLM-Judge in achieving optimal results. A case study illustrated that unnecessary reasoning can lead to overthinking and reduced accuracy, highlighting OThink-R1’s strength in adaptive reasoning.  Conclusion: Towards Scalable and Efficient Hybrid Reasoning Systems In conclusion, OThink-R1 is a large reasoning model that adaptively switches between fast and slow thinking modes to improve both efficiency and performance. It addresses the issue of unnecessarily complex reasoning in large models by analyzing and classifying reasoning steps as either essential or redundant. By pruning the redundant ones while maintaining logical accuracy, OThink-R1 reduces unnecessary computation. It also introduces a dual-reference KL-divergence loss to strengthen hybrid reasoning. Tested on math and QA tasks, it cuts down reasoning redundancy by 23% without sacrificing accuracy, showing promise for building more adaptive, scalable, and efficient AI reasoning systems in the future.  Check out the Paper and GitHub Page. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 100k+ ML SubReddit and Subscribe to our Newsletter. Sana HassanSana Hassan, a consulting intern at Marktechpost and dual-degree student at IIT Madras, is passionate about applying technology and AI to address real-world challenges. With a keen interest in solving practical problems, he brings a fresh perspective to the intersection of AI and real-life solutions.Sana Hassanhttps://www.marktechpost.com/author/sana-hassan/Building AI-Powered Applications Using the Plan → Files → Code Workflow in TinyDevSana Hassanhttps://www.marktechpost.com/author/sana-hassan/MemOS: A Memory-Centric Operating System for Evolving and Adaptive Large Language ModelsSana Hassanhttps://www.marktechpost.com/author/sana-hassan/Google AI Unveils a Hybrid AI-Physics Model for Accurate Regional Climate Risk Forecasts with Better Uncertainty AssessmentSana Hassanhttps://www.marktechpost.com/author/sana-hassan/Run Multiple AI Coding Agents in Parallel with Container-Use from Dagger #othinkr1 #dualmode #reasoning #framework #cut

Join the event trusted by enterprise leaders for nearly two decades. VB Transform brings together the people building real enterprise AI strategy. Learn more When DeepSeek released its R1 model this January, it wasn’t just another AI announcement. It was a watershed moment that sent shockwaves through the tech industry, forcing industry leaders to reconsider their fundamental approaches to AI development. What makes DeepSeek’s accomplishment remarkable isn’t that the company developed novel capabilities; rather, it was how it achieved comparable results to those delivered by tech heavyweights at a fraction of the cost. In reality, DeepSeek didn’t do anything that hadn’t been done before; its innovation stemmed from pursuing different priorities. As a result, we are now experiencing rapid-fire development along two parallel tracks: efficiency and compute.  As DeepSeek prepares to release its R2 model, and as it concurrently faces the potential of even greater chip restrictions from the U.S., it’s important to look at how it captured so much attention. Engineering around constraints DeepSeek’s arrival, as sudden and dramatic as it was, captivated us all because it showcased the capacity for innovation to thrive even under significant constraints. Faced with U.S. export controls limiting access to cutting-edge AI chips, DeepSeek was forced to find alternative pathways to AI advancement. While U.S. companies pursued performance gains through more powerful hardware, bigger models and better data, DeepSeek focused on optimizing what was available. It implemented known ideas with remarkable execution — and there is novelty in executing what’s known and doing it well. This efficiency-first mindset yielded incredibly impressive results. DeepSeek’s R1 model reportedly matches OpenAI’s capabilities at just 5 to 10% of the operating cost. According to reports, the final training run for DeepSeek’s V3 predecessor cost a mere million — which was described by former Tesla AI scientist Andrej Karpathy as “a joke of a budget” compared to the tens or hundreds of millions spent by U.S. competitors. More strikingly, while OpenAI reportedly spent million training its recent “Orion” model, DeepSeek achieved superior benchmark results for just million — less than 1.2% of OpenAI’s investment. If you get starry eyed believing these incredible results were achieved even as DeepSeek was at a severe disadvantage based on its inability to access advanced AI chips, I hate to tell you, but that narrative isn’t entirely accurate. Initial U.S. export controls focused primarily on compute capabilities, not on memory and networking — two crucial components for AI development. That means that the chips DeepSeek had access to were not poor quality chips; their networking and memory capabilities allowed DeepSeek to parallelize operations across many units, a key strategy for running their large model efficiently. This, combined with China’s national push toward controlling the entire vertical stack of AI infrastructure, resulted in accelerated innovation that many Western observers didn’t anticipate. DeepSeek’s advancements were an inevitable part of AI development, but they brought known advancements forward a few years earlier than would have been possible otherwise, and that’s pretty amazing. Pragmatism over process Beyond hardware optimization, DeepSeek’s approach to training data represents another departure from conventional Western practices. Rather than relying solely on web-scraped content, DeepSeek reportedly leveraged significant amounts of synthetic data and outputs from other proprietary models. This is a classic example of model distillation, or the ability to learn from really powerful models. Such an approach, however, raises questions about data privacy and governance that might concern Western enterprise customers. Still, it underscores DeepSeek’s overall pragmatic focus on results over process. The effective use of synthetic data is a key differentiator. Synthetic data can be very effective when it comes to training large models, but you have to be careful; some model architectures handle synthetic data better than others. For instance, transformer-based models with mixture of expertsarchitectures like DeepSeek’s tend to be more robust when incorporating synthetic data, while more traditional dense architectures like those used in early Llama models can experience performance degradation or even “model collapse” when trained on too much synthetic content. This architectural sensitivity matters because synthetic data introduces different patterns and distributions compared to real-world data. When a model architecture doesn’t handle synthetic data well, it may learn shortcuts or biases present in the synthetic data generation process rather than generalizable knowledge. This can lead to reduced performance on real-world tasks, increased hallucinations or brittleness when facing novel situations.  Still, DeepSeek’s engineering teams reportedly designed their model architecture specifically with synthetic data integration in mind from the earliest planning stages. This allowed the company to leverage the cost benefits of synthetic data without sacrificing performance. Market reverberations Why does all of this matter? Stock market aside, DeepSeek’s emergence has triggered substantive strategic shifts among industry leaders. Case in point: OpenAI. Sam Altman recently announced plans to release the company’s first “open-weight” language model since 2019. This is a pretty notable pivot for a company that built its business on proprietary systems. It seems DeepSeek’s rise, on top of Llama’s success, has hit OpenAI’s leader hard. Just a month after DeepSeek arrived on the scene, Altman admitted that OpenAI had been “on the wrong side of history” regarding open-source AI.  With OpenAI reportedly spending to 8 billion annually on operations, the economic pressure from efficient alternatives like DeepSeek has become impossible to ignore. As AI scholar Kai-Fu Lee bluntly put it: “You’re spending billion or billion a year, making a massive loss, and here you have a competitor coming in with an open-source model that’s for free.” This necessitates change. This economic reality prompted OpenAI to pursue a massive billion funding round that valued the company at an unprecedented billion. But even with a war chest of funds at its disposal, the fundamental challenge remains: OpenAI’s approach is dramatically more resource-intensive than DeepSeek’s. Beyond model training Another significant trend accelerated by DeepSeek is the shift toward “test-time compute”. As major AI labs have now trained their models on much of the available public data on the internet, data scarcity is slowing further improvements in pre-training. To get around this, DeepSeek announced a collaboration with Tsinghua University to enable “self-principled critique tuning”. This approach trains AI to develop its own rules for judging content and then uses those rules to provide detailed critiques. The system includes a built-in “judge” that evaluates the AI’s answers in real-time, comparing responses against core rules and quality standards. The development is part of a movement towards autonomous self-evaluation and improvement in AI systems in which models use inference time to improve results, rather than simply making models larger during training. DeepSeek calls its system “DeepSeek-GRM”. But, as with its model distillation approach, this could be considered a mix of promise and risk. For example, if the AI develops its own judging criteria, there’s a risk those principles diverge from human values, ethics or context. The rules could end up being overly rigid or biased, optimizing for style over substance, and/or reinforce incorrect assumptions or hallucinations. Additionally, without a human in the loop, issues could arise if the “judge” is flawed or misaligned. It’s a kind of AI talking to itself, without robust external grounding. On top of this, users and developers may not understand why the AI reached a certain conclusion — which feeds into a bigger concern: Should an AI be allowed to decide what is “good” or “correct” based solely on its own logic? These risks shouldn’t be discounted. At the same time, this approach is gaining traction, as again DeepSeek builds on the body of work of othersto create what is likely the first full-stack application of SPCT in a commercial effort. This could mark a powerful shift in AI autonomy, but there still is a need for rigorous auditing, transparency and safeguards. It’s not just about models getting smarter, but that they remain aligned, interpretable, and trustworthy as they begin critiquing themselves without human guardrails. Moving into the future So, taking all of this into account, the rise of DeepSeek signals a broader shift in the AI industry toward parallel innovation tracks. While companies continue building more powerful compute clusters for next-generation capabilities, there will also be intense focus on finding efficiency gains through software engineering and model architecture improvements to offset the challenges of AI energy consumption, which far outpaces power generation capacity.  Companies are taking note. Microsoft, for example, has halted data center development in multiple regions globally, recalibrating toward a more distributed, efficient infrastructure approach. While still planning to invest approximately billion in AI infrastructure this fiscal year, the company is reallocating resources in response to the efficiency gains DeepSeek introduced to the market. Meta has also responded, With so much movement in such a short time, it becomes somewhat ironic that the U.S. sanctions designed to maintain American AI dominance may have instead accelerated the very innovation they sought to contain. By constraining access to materials, DeepSeek was forced to blaze a new trail. Moving forward, as the industry continues to evolve globally, adaptability for all players will be key. Policies, people and market reactions will continue to shift the ground rules — whether it’s eliminating the AI diffusion rule, a new ban on technology purchases or something else entirely. It’s what we learn from one another and how we respond that will be worth watching. Jae Lee is CEO and co-founder of TwelveLabs. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured. #rethinking #deepseeks #playbook #shakes #highspend

Autoregressive image generation has been shaped by advances in sequential modeling, originally seen in natural language processing. This field focuses on generating images one token at a time, similar to how sentences are constructed in language models. The appeal of this approach lies in its ability to maintain structural coherence across the image while allowing for high levels of control during the generation process. As researchers began to apply these techniques to visual data, they found that structured prediction not only preserved spatial integrity but also supported tasks like image manipulation and multimodal translation effectively. Despite these benefits, generating high-resolution images remains computationally expensive and slow. A primary issue is the number of tokens needed to represent complex visuals. Raster-scan methods that flatten 2D images into linear sequences require thousands of tokens for detailed images, resulting in long inference times and high memory consumption. Models like Infinity need over 10,000 tokens for a 1024×1024 image. This becomes unsustainable for real-time applications or when scaling to more extensive datasets. Reducing the token burden while preserving or improving output quality has become a pressing challenge. Efforts to mitigate token inflation have led to innovations like next-scale prediction seen in VAR and FlexVAR. These models create images by predicting progressively finer scales, which imitates the human tendency to sketch rough outlines before adding detail. However, they still rely on hundreds of tokens—680 in the case of VAR and FlexVAR for 256×256 images. Moreover, approaches like TiTok and FlexTok use 1D tokenization to compress spatial redundancy, but they often fail to scale efficiently. For example, FlexTok’s gFID increases from 1.9 at 32 tokens to 2.5 at 256 tokens, highlighting a degradation in output quality as the token count grows. Researchers from ByteDance introduced DetailFlow, a 1D autoregressive image generation framework. This method arranges token sequences from global to fine detail using a process called next-detail prediction. Unlike traditional 2D raster-scan or scale-based techniques, DetailFlow employs a 1D tokenizer trained on progressively degraded images. This design allows the model to prioritize foundational image structures before refining visual details. By mapping tokens directly to resolution levels, DetailFlow significantly reduces token requirements, enabling images to be generated in a semantically ordered, coarse-to-fine manner. The mechanism in DetailFlow centers on a 1D latent space where each token contributes incrementally more detail. Earlier tokens encode global features, while later tokens refine specific visual aspects. To train this, the researchers created a resolution mapping function that links token count to target resolution. During training, the model is exposed to images of varying quality levels and learns to predict progressively higher-resolution outputs as more tokens are introduced. It also implements parallel token prediction by grouping sequences and predicting entire sets at once. Since parallel prediction can introduce sampling errors, a self-correction mechanism was integrated. This system perturbs certain tokens during training and teaches subsequent tokens to compensate, ensuring that final images maintain structural and visual integrity. The results from the experiments on the ImageNet 256×256 benchmark were noteworthy. DetailFlow achieved a gFID score of 2.96 using only 128 tokens, outperforming VAR at 3.3 and FlexVAR at 3.05, both of which used 680 tokens. Even more impressive, DetailFlow-64 reached a gFID of 2.62 using 512 tokens. In terms of speed, it delivered nearly double the inference rate of VAR and FlexVAR. A further ablation study confirmed that the self-correction training and semantic ordering of tokens substantially improved output quality. For example, enabling self-correction dropped the gFID from 4.11 to 3.68 in one setting. These metrics demonstrate both higher quality and faster generation compared to established models. By focusing on semantic structure and reducing redundancy, DetailFlow presents a viable solution to long-standing issues in autoregressive image generation. The method’s coarse-to-fine approach, efficient parallel decoding, and ability to self-correct highlight how architectural innovations can address performance and scalability limitations. Through their structured use of 1D tokens, the researchers from ByteDance have demonstrated a model that maintains high image fidelity while significantly reducing computational load, making it a valuable addition to image synthesis research. Check out the Paper and GitHub Page. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 95k+ ML SubReddit and Subscribe to our Newsletter. NikhilNikhil is an intern consultant at Marktechpost. He is pursuing an integrated dual degree in Materials at the Indian Institute of Technology, Kharagpur. Nikhil is an AI/ML enthusiast who is always researching applications in fields like biomaterials and biomedical science. With a strong background in Material Science, he is exploring new advancements and creating opportunities to contribute.Nikhilhttps://www.marktechpost.com/author/nikhil0980/Teaching AI to Say ‘I Don’t Know’: A New Dataset Mitigates Hallucinations from Reinforcement FinetuningNikhilhttps://www.marktechpost.com/author/nikhil0980/This AI Paper Introduces LLaDA-V: A Purely Diffusion-Based Multimodal Large Language Model for Visual Instruction Tuning and Multimodal ReasoningNikhilhttps://www.marktechpost.com/author/nikhil0980/NVIDIA AI Introduces Fast-dLLM: A Training-Free Framework That Brings KV Caching and Parallel Decoding to Diffusion LLMsNikhilhttps://www.marktechpost.com/author/nikhil0980/Meet NovelSeek: A Unified Multi-Agent Framework for Autonomous Scientific Research from Hypothesis Generation to Experimental Validation #bytedance #researchers #introduce #detailflow #coarsetofine

html PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN" "; A still from a demo shot created using MotionMaker, the new generative AI toolset introduced in Maya 2026.1 for roughing out movement animations. Autodesk has released Maya 2026.1, the latest version of its 3D modeling and animation software for visual effects, games and motion graphics work.The release adds MotionMaker, a new AI-based system for generating movement animations for biped and quadruped characters, especially for previs and layout work. Other changes include a new modular character rigging framework inside Bifrost for Maya, plus updates to liquid simulation, OpenPBR support and USD workflows. Autodesk has also released Maya Creative 2026.1, the corresponding update to the cut-down edition of Maya for smaller studios. MotionMaker: new generative AI tool roughs out movement animations The headline feature in Maya 2026.1 is MotionMaker: a new generative animation system.It lets users “create natural character movements in minutes instead of hours”, using a workflow more “like giving stage directions to a digital actor” than traditional animation. Users set keys for a character’s start and end positions, or create a guide path in the viewport, and MotionMaker automatically generates the motion in between. At the minute, that mainly means locomotion cycles, for both bipeds and quadrupeds, plus a few other movements, like jumping or sitting. Although MotionMaker is designed for “anyone in the animation pipeline”, the main initial use cases seem to be layout and previs rather than hero animation. Its output is also intended to be refined manually – Autodesk’s promotional material describes it as getting users “80% of the way there” for “certain types of shots”. Accordingly, MotionMaker comes with its own Editor window, which provides access to standard Maya animation editing tools. Users can layer in animation from other sources, including motion capture or keyframe animation retargeted from other characters: to add upper body movements, for example. There are a few more MotionMaker-specific controls: the video above shows speed ramping, to control the time it takes the character to travel between two points. There is also a Character Scale setting, which determines how a character’s size and weight is expressed through the animation generated. You can read more about the design and aims of MotionMaker in a Q&A with Autodesk Senior Principal Research Scientist Evan Atherton on Autodesk’s blog. According to Atherton, the AI models were trained using motion capture data “specifically collected for this tool”. That includes source data from male and female human performers, plus wolf-style dogs, although the system is “designed to support additionalstyles” in future. Bifrost: new modular character rigging framework Character artists and animators also get a new modular rigging framework in Bifrost.Autodesk has been teasing new character rigging capabilities in the node-based framework for building effects since Maya 2025.1, but this seems to be its official launch. The release is compatibility-breaking, and does not work with earlier versions of the toolset. The new Rigging Module Framework is described as a “modular, compound-based system for building … production-ready rigs”, and is “fully integrated with Maya”. Animators can “interact with module inputs and outputs directly from the Maya scene”, and rigs created with Bifrost can be converted into native Maya controls, joints and attributes. Bifrost: improvements to liquid simulation and workflow Bifrost 2.14 for Maya also features improvements to Bifrost’s existing functionality, particularly liquid simulation. The properties of collider objects, like bounciness, stickiness and roughness, can now influence liquid behavior in the same way they do particle behavior and other collisions. In addition, a new parameter controls air drag on foam and spray thrown out by a liquid. Workflow improvements include the option to convert Bifrost curves to Maya scene curves, and batch execution, to write out cache files “without the risk of accidentally overwriting them”. LookdevX: support for OpenPBR in FBX files LookdevX, Maya’s plugin for creating USD shading graphs, has also been updated. Autodesk introduced support for OpenPBR, the open material standard intended as a unified successor to the Autodesk Standard Surface and Adobe Standard Material, in 2024. To that, the latest update adds support for OpenPBR materials in FBX files, making it possible to import or export them from other applications that support OpenPBR: at the minute, 3ds Max plus some third-party renderers. LookdevX 1.8 also features a number of workflow improvements, particularly on macOS. USD for Maya: workflow improvements USD for Maya, the software’s USD plugin, also gets workflow improvements, with USD for Maya 0.32 adding support for animation curves for camera attributes in exports.Other changes include support for MaterialX documents and better representation of USD lights in the viewport. Arnold for Maya: performance improvements Maya’s integration plugin for Autodesk’s Arnold renderer has also been updated, with MtoA 5.5.2 supporting the changes in Arnold 7.4.2.They’re primarily performance improvements, especially to scene initialization times when rendering on machines with high numbers of CPU cores. Maya Creative 2026.1 also released Autodesk has also released Maya Creative 2026.1, the corresponding update to the cut-down edition of Maya aimed at smaller studios, and available on a pay-as-you-go basis.It includes most of the new features from Maya 2026.1, including MotionMaker, but does not include Bifrost for Maya. Price and system requirements Maya 2026.1 is available for Windows 10+, RHEL and Rocky Linux 8.10/9.3/9.5, and macOS 13.0+.The software is rental-only. Subscriptions cost /month or /year, up a further /month or /year since the release of Maya 2026. In many countries, artists earning under /year and working on projects valued at under /year, qualify for Maya Indie subscriptions, now priced at /year. Maya Creative is available pay-as-you-go, with prices starting at /day, and a minimum spend of /year. Read a full list of new features in Maya 2026.1 in the online documentation Have your say on this story by following CG Channel on Facebook, Instagram and X. As well as being able to comment on stories, followers of our social media accounts can see videos we don’t post on the site itself, including making-ofs for the latest VFX movies, animations, games cinematics and motion graphics projects. #autodesk #adds #animation #tool #motionmaker

When people talk about UX, it’s usually about the things they can see and interact with, like wireframes and prototypes, smart interactions, and design tools like Figma, Miro, or Maze. Some of the outputs are even glamorized, like design systems, research reports, and pixel-perfect UI designs. But here’s the truth I’ve seen again and again in over two decades of working in UX: none of that moves the needle if there is no collaboration. Great UX doesn’t happen in isolation. It happens through conversations with engineers, product managers, customer-facing teams, and the customer support teams who manage support tickets. Amazing UX ideas come alive in messy Miro sessions, cross-functional workshops, and those online chatswhere people align, adapt, and co-create. Some of the most impactful moments in my career weren’t when I was “designing” in the traditional sense. They have been gaining incredible insights when discussing problems with teammates who have varied experiences, brainstorming, and coming up with ideas that I never could have come up with on my own. As I always say, ten minds in a room will come up with ten times as many ideas as one mind. Often, many ideas are the most useful outcome. There have been times when a team has helped to reframe a problem in a workshop, taken vague and conflicting feedback, and clarified a path forward, or I’ve sat with a sales rep and heard the same user complaint show up in multiple conversations. This is when design becomes a team sport, and when your ability to capture the outcomes multiplies the UX impact. Why This Article Matters Now The reason collaboration feels so urgent now is that the way we work since COVID has changed, according to a study published by the US Department of Labor. Teams are more cross-functional, often remote, and increasingly complex. Silos are easier to fall into, due to distance or lack of face-to-face contact, and yet alignment has never been more important. We can’t afford to see collaboration as a “nice to have” anymore. It’s a core skill, especially in UX, where our work touches so many parts of an organisation. Let’s break down what collaboration in UX really means, and why it deserves way more attention than it gets. What Is Collaboration In UX, Really? Let’s start by clearing up a misconception. Collaboration is not the same as cooperation. Cooperation: “You do your thing, I’ll do mine, and we’ll check in later.” Collaboration: “Let’s figure this out together and co-own the outcome.” Collaboration, as defined in the book Communication Concepts, published by Deakin University, involves working with others to produce outputs and/or achieve shared goals. The outcome of collaboration is typically a tangible product or a measurable achievement, such as solving a problem or making a decision. Here’s an example from a recent project: Recently, I worked on a fraud alert platform for a fintech business. It was a six-month project, and we had zero access to users, as the product had not yet hit the market. Also, the users were highly specialised in the B2B finance space and were difficult to find. Additionally, the team members I needed to collaborate with were based in Malaysia and Melbourne, while I am located in Sydney. Instead of treating that as a dead end, we turned inward: collaborating with subject matter experts, professional services consultants, compliance specialists, and customer support team members who had deep knowledge of fraud patterns and customer pain points. Through bi-weekly workshops using a Miro board, iterative feedback loops, and sketching sessions, we worked on design solution options. I even asked them to present their own design version as part of the process. After months of iterating on the fraud investigation platform through these collaboration sessions, I ended up with two different design frameworks for the investigator’s dashboard. Instead of just presenting the “best one” and hoping for buy-in, I ran a voting exercise with PMs, engineers, SMEs, and customer support. Everyone had a voice. The winning design was created and validated with the input of the team, resulting in an outcome that solved many problems for the end user and was owned by the entire team. That’s collaboration! It is definitely one of the most satisfying projects of my career. On the other hand, I recently caught up with an old colleague who now serves as a product owner. Her story was a cautionary tale: the design team had gone ahead with a major redesign of an app without looping her in until late in the game. Not surprisingly, the new design missed several key product constraints and business goals. It had to be scrapped and redone, with her now at the table. That experience reinforced what we all know deep down: your best work rarely happens in isolation. As illustrated in my experience, true collaboration can span many roles. It’s not just between designers and PMs. It can also include QA testers who identify real-world issues, content strategists who ensure our language is clear and inclusive, sales representatives who interact with customers on a daily basis, marketers who understand the brand’s voice, and, of course, customer support agents who are often the first to hear when something goes wrong. The best outcomes arrive when we’re open to different perspectives and inputs. Why Collaboration Is So Overlooked? If collaboration is so powerful, why don’t we talk about it more? In my experience, one reason is the myth of the “lone UX hero”. Many of us entered the field inspired by stories of design geniuses revolutionising products on their own. Our portfolios often reflect that as well. We showcase our solo work, our processes, and our wins. Job descriptions often reinforce the idea of the solo UX designer, listing tool proficiency and deliverables more than soft skills and team dynamics. And then there’s the team culture within many organisations of “just get the work done”, which often leads to fewer meetings and tighter deadlines. As a result, a sense of collaboration is inefficient and wasted. I have also experienced working with some designers where perfectionism and territoriality creep in — “This is my design” — which kills the open, communal spirit that collaboration needs. When Collaboration Is The User Research In an ideal world, we’d always have direct access to users. But let’s be real. Sometimes that just doesn’t happen. Whether it’s due to budget constraints, time limitations, or layers of bureaucracy, talking to end users isn’t always possible. That’s where collaboration with team members becomes even more crucial. The next best thing to talking to users? Talking to the people who talk to users. Sales teams, customer success reps, tech support, and field engineers. They’re all user researchers in disguise! On another B2C project, the end users were having trouble completing the key task. My role was to redesign the onboarding experience for an online identity capture tool for end users. I was unable to schedule interviews with end users due to budget and time constraints, so I turned to the sales and tech support teams. I conducted multiple mini-workshops to identify the most common onboarding issues they had heard directly from our customers. This led to a huge “aha” moment: most users dropped off before the document capture process. They may have been struggling with a lack of instruction, not knowing the required time, or not understanding the steps involved in completing the onboarding process. That insight reframed my approach, and we ultimately redesigned the flow to prioritize orientation and clear instructions before proceeding to the setup steps. Below is an example of one of the screen designs, including some of the instructions we added. This kind of collaboration is user research. It’s not a substitute for talking to users directly, but it’s a powerful proxy when you have limited options. But What About Using AI? Glad you asked! Even AI tools, which are increasingly being used for idea generation, pattern recognition, or rapid prototyping, don’t replace collaboration; they just change the shape of it. AI can help you explore design patterns, draft user flows, or generate multiple variations of a layout in seconds. It’s fantastic for getting past creative blocks or pressure-testing your assumptions. But let’s be clear: these tools are accelerators, not oracles. As an innovation and strategy consultant Nathan Waterhouse points out, AI can point you in a direction, but it can’t tell you which direction is the right one in your specific context. That still requires human judgment, empathy, and an understanding of the messy realities of users and business goals. You still need people, especially those closest to your users, to validate, challenge, and evolve any AI-generated idea. For instance, you might use ChatGPT to brainstorm onboarding flows for a SaaS tool, but if you’re not involving customer support reps who regularly hear “I didn’t know where to start” or “I couldn’t even log in,” you’re just working with assumptions. The same applies to engineers who know what is technically feasible or PMs who understand where the business is headed. AI can generate ideas, but only collaboration turns those ideas into something usable, valuable, and real. Think of it as a powerful ingredient, but not the whole recipe. How To Strengthen Your UX Collaboration Skills? If collaboration doesn’t come naturally or hasn’t been a focus, that’s okay. Like any skill, it can be practiced and improved. Here are a few ways to level up: Cultivate curiosity about your teammates.Ask engineers what keeps them up at night. Learn what metrics your PMs care about. Understand the types of tickets the support team handles most frequently. The more you care about their challenges, the more they'll care about yours. Get comfortable facilitating.You don’t need to be a certified Design Sprint master, but learning how to run a structured conversation, align stakeholders, or synthesize different points of view is hugely valuable. Even a simple “What’s working? What’s not?” retro can be an amazing starting point in identifying where you need to focus next. Share early, share often.Don’t wait until your designs are polished to get input. Messy sketches and rough prototypes invite collaboration. When others feel like they’ve helped shape the work, they’re more invested in its success. Practice active listening.When someone critiques your work, don’t immediately defend. Pause. Ask follow-up questions. Reframe the feedback. Collaboration isn’t about consensus; it’s about finding a shared direction that can honour multiple truths. Co-own the outcome.Let go of your ego. The best UX work isn’t “your” work. It’s the result of many voices, skill sets, and conversations converging toward a solution that helps users. It’s not “I”, it’s “we” that will solve this problem together. Conclusion: UX Is A Team Sport Great design doesn’t emerge from a vacuum. It comes from open dialogue, cross-functional understanding, and a shared commitment to solving real problems for real people. If there’s one thing I wish every early-career designer knew, it’s this: Collaboration is not a side skill. It’s the engine behind every meaningful design outcome. And for seasoned professionals, it’s the superpower that turns good teams into great ones. So next time you’re tempted to go heads-down and just “crank out a design,” pause to reflect. Ask who else should be in the room. And invite them in, not just to review your work, but to help create it. Because in the end, the best UX isn’t just what you make. It’s what you make together. Further Reading On SmashingMag “Presenting UX Research And Design To Stakeholders: The Power Of Persuasion,” Victor Yocco “Transforming The Relationship Between Designers And Developers,” Chris Day “Effective Communication For Everyday Meetings,” Andrii Zhdan “Preventing Bad UX Through Integrated Design Workflows,” Ceara Crawshaw #collaboration #most #underrated #skill #one