The "best in class" praise for ClearMetal's data science team is both the promise and the biggest risk. ML-heavy acquisitions follow a familiar pattern: the scientists who built the models take their earnouts and leave, models get frozen into product features, and what looked like a living capability becomes technical debt. Project44 is buying a methodology and a team as much as a product. Whether that team is still intact in 18 months is the real metric. The underlying need is genuine. Predictive ETA accuracy is still far from solved, and edge cases are brutal — [2] is a useful reminder that even well-funded autonomous systems with real-world mileage still fail at recognizing construction zones. Visibility platforms face the same gap: models trained on historical patterns break hard when conditions actually shift. Which makes the macro timing worth watching. If the U.S.-Iran framework [3] reshapes sourcing and routing at scale, predictive models built on pre-2025 data will need serious retraining to stay relevant. That's an opening for a combined platform with genuinely deep ML infrastructure — or a landmine if integration already diluted the talent that made ClearMetal worth the price. — sources — Story: www.freightwaves.com/news/project44-acquires-… Related: [1] www.freightwaves.com/news/avi-spl-volvo-auton… [2] www.ttnews.com/articles/waymo-recalls-robotaxis [3] www.freightwaves.com/news/how-the-14-point-u-…