Data quality is only part of the equation. When AI is being trained and evaluated, there are a few questions worth asking: - Can each contribution be traced? - Who contributed the data? - Was their domain expertise verified? - Is there a verifiable audit trail of how it was produced and validated? For high-stakes AI, “high-quality” data isn’t enough. It needs to be expert-validated, traceable, and auditable.

Automating data quality with AI transformed my work. Discover how prompt engineering mastery helped me catch 99% of errors and save 20+ hours weekly.  — @meisshaily #ArtificialIntelligence# #TechNews# #Tech# #Technology# #Prompt#

🦾 Why Data, Not Models, Is the Real Moat in Embodied AI The timing of this question is hard to miss. This week, the World Humanoid Robot Games released a 2,500+ hour dataset covering 12 scenario categories, 44 operations, and more than 10,000 tasks. Crucially, it also includes failures and edge cases. But raw hours tell only part of the story. Zhihu contributor 于超, an assistant professor at Tsinghua Shenzhen International Graduate School, shares his team’s view on why data has become embodied AI’s hardest-to-replicate advantage. 1️⃣ Robot scaling is fundamentally asymmetric Like language models, robot policies appear to benefit from more data, larger models, and greater compute. But these three inputs do not scale equally. Model architectures can be studied and reproduced quickly. General-purpose compute can, in principle, be purchased. High-quality robot data is different. It must be accumulated through physical interaction, and competitors cannot recreate it overnight. That asymmetry is what turns data into a moat. 2️⃣ Robot data must be manufactured LLMs inherited decades of internet data. Robots did not. Every useful trajectory must be produced through physical interaction. Even a simple cup-grasping task changes with the object, lighting, environment, camera angle, and robot body. So raw hours are not enough. What matters is the diversity of embodiments, tasks, objects, failures, and recoveries. Open X-Embodiment needed more than 20 institutions and 22 robot platforms to collect over one million trajectories. DROID used 50 collectors for a year, producing only 350 hours of data. New methods such as UMI and egocentric recording make collection easier. But every hour still requires real people, equipment, and time. 3️⃣ A successful trajectory can still be bad data Robot data quality is more complicated than whether a task was completed. On the hardware side, camera accuracy, encoder readings, force sensors, calibration, communication latency, and synchronization across modalities can all corrupt a trajectory. The human operator adds another source of noise. Teleoperating a robot is not the same as performing the action directly. Operators hesitate, pause, readjust, and develop habits for compensating for the control system. A task may succeed even when parts of the demonstration should never be imitated. Success is therefore only the coarsest possible label. One trajectory can contain both excellent behavior and inefficient or misleading actions. Training on the entire trajectory without distinction effectively tells the robot to learn both. 4️⃣ The next challenge is information density Collecting more trajectories is only half the problem. Teams must also identify which parts are worth learning from. Yu Chao’s team developed STEAM to detect local progress within a trajectory without frame-by-frame annotation or manually designed rewards. It separates useful progress from hesitation, failure, and recovery. The key question is shifting from “How many trajectories do we have?” to “How much useful information does each trajectory contain?” 5️⃣ Embodied data is physically expensive Text can be copied. Videos can be downloaded. Robot data requires a physical production process. Collecting one hour may involve a robot, sensors, teleoperation equipment, an operator, a suitable environment, task materials, and engineers who maintain and calibrate the system. Real factories, stores, and homes add even more complexity. And pressing the record button is only the beginning. Transmission, cleaning, governance, and storage can cost more than collection itself. The author offers a rough calculation. If a company wants one million hours of real-world data and reduces the combined collection and management cost to RMB 200 per hour, the total still reaches RMB 200 million. 🔑 The real moat compounds over time Quantity, quality, and cost explain why embodied AI data cannot be replicated through a short burst of spending. Large, diverse, high-quality datasets require physical infrastructure, operational discipline, and years of accumulation. As robot policies continue to benefit from scaling, the durable advantage will belong to teams that can repeatedly: 🔹 Collect broader real-world experience 🔹 Identify the most informative behavior 🔹 Preserve failures and recovery signals 🔹 Turn noisy trajectories into useful learning data In embodied AI, having data and knowing how to use it are becoming two very different capabilities. And the second may ultimately matter even more than the first. 🔗 Full analysis: #EmbodiedAI# #Robotics# #PhysicalAI# #RobotLearning# #AIData# #ScalingLaw# #Tsinghua#

A model is only as good as its data, and we’ve long since exhausted the internet. From here on out, model progress is gated by data production. @mercor’s @BrendanFoody joined us at our Sovereign AI event to talk about how RL environments get built, and why your data might be your real moat: 00:00 Introduction 00:47 A short history of the data market: crowdsourcing to agentic data 02:29 What an RL environment is: worlds, apps, tasks 03:57 Why only humans can measure the frontier 05:35 Building verifiers is the hard part 06:44 Walkthrough: a real legal RL environment 08:18 Leaderboards — and what open weights change 09:45 Post-training results on Apex Agents 11:17 Three ways companies buy data 12:49 Q&A: How do you price data? 14:17 Q&A: What "data quality" actually means 16:42 Q&A: The misunderstanding about synthetic data 18:17 Q&A: Why RL environments now — and what comes after 21:20 Q&A: Can you scale rubric generation with models? 23:00 Q&A: RL environments for cyber defense 25:33 Q&A: Build data in-house or partner?

Kled has just signed a $12 million non-exclusive data deal over the next 2 years making the marketplace officially profitable. We’re now heavily expanding our team to bring Kled to android, support every global currency / payment method, increase our daily uploads and registrations on Trace by 2x in 1 month, and roll out our next fraud detection model for better data quality. We want the best engineers in the world. All open careers can be found here:

Every robot on the Fleet marketplace has passed setup validation and meets PrismaX data quality standards before it lists. Piper, TOK2, and YAM are live now, with setup and onboarding included. What else shipped in Q2:

You can't have trusted enterprise AI without absolute visibility into the data feeding it. 🔍 At #SnowflakeSummit#, @montecarlodata's Ethan Post breaks down how their native integration creates an end-to-end observability foundation, tracking data quality straight to AI agent performance 👉🏻

41 production dbt models migrated in ~20 hours this week — including full data quality validation against legacy outputs before every cutover. And yes — I completely burned through my Cursor monthly quota doing it. Here's the workflow that made it possible. 🧵 ➢ The stack: @cursor_ai + @claudeai + @dune MCP + a living SOP Every model follows the same loop: 1. Author in Cursor with Claude Cursor keeps the SOP in-context across the entire session. Claude writes the SQL. Known Trino failure patterns (integer division, USING join resolution, sequence() 10k limit) are in-context before they become bugs — because the SOP is always there. I've tested other models. Claude still produces the fewest SQL bugs on Trino-specific edge cases. Not close. 2. Validate on Dune FULL OUTER JOIN new vs legacy. Every metric. Every day. The gate runs before cutover — not after. 3. Update the SOP Every edge case becomes a guardrail. The next migration starts from a better baseline. ➢ The SOP isn't documentation. It's a feedback loop. Bug → rule → Cursor keeps it in Claude's context → fewer bugs per model. By model 47, the failure modes in the SOP are all real — each traced to a dated incident with root cause and fix. The document became the system's memory. ➢ Why this matters — and why Dune made it possible Complex DeFi financials (per-user balances, supply indices, borrow rates across chains) used to refresh once a day at best. Dune's dbt + Trino stack makes hourly incremental jobs on this data actually viable — at sustainable cost. That's not a small thing. Hourly on-chain financial data, production-grade, open to the ecosystem. Thanks to @dune and the team ( @onchain_ben @fr0zensun @kdotkrisp @Maria) for building infrastructure that makes this real. 🙏 ➢ Why I care this much about accuracy @sparkfinance is one of the most data-driven teams I've worked with. Data here isn't a dashboard for show — it directly informs protocol decisions. The team asks hard questions. Numbers need to be right before they're trusted, not after. That standard is what pushed me to build the validation gate in the first place. Working with a team that respects data rigor pushes you to build better systems, not just better dashboards. I'm grateful for that.

The Q4 Roadmap is out. Oct brings synthetic gold data and a reinforcement learning loop that continously improves data quality. Nov scales up production and adds structured, statistical artifacts that will allow us to judge the credibility of empirical evidence. Dec will show a rapidly growing knowledge graph and benchmarks for how much Claims improves agentic accuracy in research tasks compared to unstructured text. Based on this, we will be able to launch the commercial product and start generating revenue.