Not so long ago, industry research meant exploring ideas that were typically anywhere from 3 to 10 years out from actual product implementation (even faster in open source). If you’ve forgotten that definition, you’re excused. The AI era has found researchers building express lanes and wormholes throughout the research-to-production pipeline. 

Going faster doesn’t have to mean cutting corners, however. On the contrary, I bring up this disruption because while putting this issue together I was struck by the diversity of research taking an open source approach precisely because we’re all invested in making AI legitimately usable by doing things right. The Summer 2026 edition of RHRQ highlights many ways industry engages with research: Red Hat partnerships with individual universities, advanced PhD research guided in part by Red Hat engineers, cross-industry cooperation within an international benchmarking consortium, and open ecosystems like the AI Alliance. The key is open collaboration, breaking down walls between university researchers, industry engineers, and users so that everyone can move faster. 

North Carolina State University Provost Jim Pfaendtner described the academic  version of open source as maximizing “the velocity of dissemination.” In his interview with Senior Director of Engineering for Red Hat Open Shift AI Sherard Griffin, Pfaendtner talks about how university-industry partnerships are a critical factor in keeping a collective foot on the accelerator. From preparing the next generation of engineers to transferring technology via university-founded startups or leveraging AI to transform engineering design, Pfaendtner describes a compelling vision for making a large land-grant university agile enough to work at the speed of AI. Read “Stop talking, start doing: how universities rise to the challenge of the AI revolution” to learn more.

In September 2025, we announced a collaboration among Red Hat, IBM Research, the AI Alliance, and the Mass Open Cloud to support several short-term (less than one year) research projects as part of the NAIRR pilot deep partnerships, and now they are bearing fruit. I’m excited to publish our first two articles about these ambitious projects in this issue, both led by PIs from Rensselaer Polytechnic Institute. Read “CRANE: teaching code models to think without breaking their tools” to discover a model-merging approach that transfers reasoning improvements from thinking models while preserving the reliable tool-use patterns of instruct models. If you’re using an agent for software-engineering tasks that depend on reliable tool use, you understand why this matters. Or, consider the development of a Time-Series Data Agent: an agentic system that combines a numerically focused time-series foundation model with a semantically focused large language model. This next frontier in forecasting models is the topic of “Beyond the leaderboard: rethinking time-series foundation models.” 

Also in this issue, we’re spotlighting one of five papers from members of the Red Hat AI Innovation Team presented at NeurIPS 2025. In the RHRQ article “Particle filtering for better LLM reasoning,” co-author Guangxuan Xu explains the method introduced in “Rollout roulette: a probabilistic inference approach to inference-time scaling of LLMs using particle-based Monte Carlo methods.” In May 2026, a team of Red Hat performance engineers and a researcher at the Illinois Institute of Technology won the Best Industry Paper award for “Evaluating Kubernetes performance for GenAI inference: from automatic speech recognition to LLM summarization” at the SPEC International Conference on Performance Engineering; they share that work in the article “Kubernetes meets GenAI: evaluating performance for AI inference workloads.” We’ve long followed work at Karlstad University focused on building the next generation of programmable networking—powered by Linux. Check out “Monitoring latency within the end host network stack” to get an early look at their latest development, netstacklat, an advanced latency monitoring tool the team will present at the ACM Internet Measurement Conference in October 2026. 

In his column, “Building shoulders for giants to stand on,” research engineering manager Robby Stahl observes that research in AI and AI-based research—in technical or systems problems, healthcare, or other domains—“is so much more than any person could accomplish in a lifetime.” The infinite potential he describes is one of the things that makes research such an exciting place to be. At a time when the pace of change can be dizzying, research grounded in strong partnerships and communities provides wayfinding and guardrails to ensure that even as we’re going faster and faster, we’re still moving in the right direction.