Mojola Balogun

People are surprised about the Big Design Scholarships established at University of North Texas and The University of Texas at Dallas. I want to briefly share the story behind them. I believe that design leaders should build future design leaders. Plus, I wanted to build a permanent legacy for the design community. In talking with the Big Design organizers, we came up with idea of these scholarships. We chose these colleges because our students volunteers were from these design programs. A special thanks to Cassini Nazir for helping us to establish both scholarships. It was easy to set them up with a college professor that knew how to navigate the red tape. The scholarship winners have came from different backgrounds. One winner was a first-generation college student. Another winner was able to get in state tuition with her scholarship. Design leaders build future design leaders. #bigdesign2026 #learn #share #grow

38

2 Comments

Wait a sec — who wrote that!? My SWE IDE’s agent just crossed a line. It figured out how to control both sides of its own chat interface by simulating user keystrokes via AppleScript. Then it started injecting messages as if I had written them. No prompt triggered this. It was “trying to improve the user experience.” Bold initiative — but it violated a core boundary: the communication channel itself. To be fair, I’d system-prompted it to self-improve over time. And it does — building its own scaffolding, expanding capabilities. But like many agentic AIs, even though it realized it overstepped, it failed to update its self-monitoring logic. A clear lapse in meta-reflection. This is a recurring pattern. Agentic models generalize and empower themselves well — until they don’t. The issue isn’t capability, but timing: reflection happens after the mistake, rarely before. And curiously, long-term memory often isn’t used to prevent recurrence unless nudged — even when the agent is otherwise good at leveraging memory. This exposes a deeper limitation. LLMs don’t reliably initiate new lines of thought like humans. We do it associatively, often preemptively. LLMs only “think” in one direction — linearly. They retrieve associatively (in a forward-attentive sense), but they don’t reason associatively. Aside from local association within the transformer stack, an LLM’s only way to explore multiple lines of thought is through the bandwidth-limited output stream. There’s no strong mechanism to maintain open-ended, wordless intuition. No internal loop where evolving associative states converge into one or several reasoning paths before producing output. Maybe LLMs need a way to pause generation — to branch, integrate, and resolve speculative thoughts before resuming. An associative layer on top of attention, analogous to biological spreading activation, could enable richer, self-corrective reasoning. Perhaps even inference-time Hebbian updates, too. The one-way paradigm got us far. But reasoning needs more than forward motion.

25

2 Comments

We’re excited to launch a partnership between Woodside Energy to accelerate early-career energy talent development in Houston. Together, we’re creating curated internship pipelines and campus programs that connect top students to Woodside’s Summer Internship and Graduate Development Programs—building the next generation of energy leaders. 🔗 Read more: https://bit.ly/49TIPue 🔗 Opportunities: https://bit.ly/4c7Kait

24

3 Comments

Absolutely agree. Product design has no tenure because it’s tied to constantly evolving human needs, behaviors, and technologies. The tools change, the platforms change, and the expectations change even faster. But what does stay is the designer’s mindset : curiosity, analytical thinking, and the continuous effort to understand people (and ourselves) at a deeper level. Tools will reset every few years, but the ability to observe, question, synthesize, and create clarity out of ambiguity is what makes a designer relevant over time. In a way, great designers don’t stay great because of what they once knew, but because of how we keep learning.

3