Ema
Ema at Beyond Market Intelligence is a file of 4 stories. The newest of them: “Sampling Smarter: How EVE Online Tames 20 Years of Telemetry Data”, “AI-Powered Spreadsheets Empower Enterprises, Ema Secures $77M”, and “What one GPU reveals about training a text-to-image diffusion transformer”. Twenty years of telemetry data is a lot of noise to manage, but EVE Online's approach is impressively disciplined. Ema just closed a $77 million round, bringing its total funding to $140 million, and it already counts Google and Microsoft among more than 50 enterprise customers. Acme AI is the next-generation, AI-powered spreadsheet platform built to replace Excel and redefine how analysts, data scientists, and enterprise teams work… The list below is every Ema story on Beyond Market Intelligence, newest first.
Sampling Smarter: How EVE Online Tames 20 Years of Telemetry Data
Twenty years of telemetry data is a lot of noise to manage, but EVE Online's approach is impressively disciplined. Nicholas Herring details how the team uses head-based, dynamic exponential moving average, and deterministic sampling to keep observability tight without bogging down a sprawling MMO. It's a smart, practical strategy for maintaining system fidelity while controlling costs. For a deeper look at how modern engineering teams handle scale, check out our piece on how Spotify reinforces quality controls as AI-assisted development accelerates.
AI-Powered Spreadsheets Empower Enterprises, Ema Secures $77M
Ema just closed a $77 million round, bringing its total funding to $140 million, and it already counts Google and Microsoft among more than 50 enterprise customers. That kind of traction signals something real: businesses are ready to trust AI-native tools with serious data work. We've long argued that spreadsheets shouldn't be static, and Ema is proving the shift is happening now. If you're exploring how AI can reshape your own workflows, our piece on verifying AI understanding offers a practical starting point.
What one GPU reveals about training a text-to-image diffusion transformer
Training a 210M text-to-image DiT from scratch on a single GPU is a bold experiment, and the numbers are worth reading closely. The finding that two learned null attention slots absorb roughly 90% of cross-attention mass at mid-noise is striking, especially since it quietly sidelines the usual EOS sink. That is the kind of concrete detail most write-ups skip. Flow-matching loss tracks training health, not output quality.
Exploring how an older model still competes on accuracy and speed today
CABiNet, a 2021 architecture that went quiet after ICRA, is back, and it's beating a 2026 generalist model on aerial segmentation. On UAVid, CABiNet-L hits 67.14 mIoU at 4.44 ms; YOLO26x-sem trails at 64.41 and takes nearly three times longer. The gap is mostly small classes: people and vehicles. That's not a universal win, YOLO26s matches compute and still wins on latency. But at the high-accuracy end, purpose-built still has teeth. For a deeper look at how training recipes shape these comparisons,