Adopting Persona: Senior Technical Learning & Development (L&D) Consultant and Principal Software Architect.
Abstract
This keynote address, delivered by Frances Buontempo at Meeting C++ 2025, examines the cognitive processes of learning and teaching within the software engineering domain, specifically in the context of emerging Generative AI (GenAI) and Large Language Models (LLMs). The discourse challenges the industry's reliance on rote memorization and "copy-paste" workflows (from Stack Overflow to LLMs), arguing that true understanding requires active brain engagement, the courage to ask "stupid questions," and the exploration of counterfactual "what if" scenarios.
Buontempo posits that while LLMs function as sophisticated stochastic parrots capable of synthesizing existing knowledge, they fundamentally lack human agency and the ability to challenge worldviews or navigate subtle contradictions. The presentation provides a pedagogical framework for technical mentorship, emphasizing the importance of emotional intelligence in teaching, the value of non-linear learning via technical literature, and the necessity of maintaining human-centric critical thinking as the software development landscape evolves.
Keynote Summary: Learning, Teaching, and the Limits of AI
0:00 – 2:45 The Core Triad: Learning, Teaching, and AI: Introduction to the three investigative threads. The speaker notes that most technical professionals are never formally trained in how to learn or teach, leading to a "black box" approach to professional development.
3:08 – 4:56 The Essentiality of Learning in the AI Era: Discussion on how LLMs and the internet have not made learning obsolete. True learning requires an "extra step" beyond mere copying; cognitive engagement is mandatory for mastery.
Strategy: Developing a "question-first" mindset during technical sessions changes how information is processed and engaged.
Takeaway: Active questioning facilitates shared understanding and unpacks hidden complexities in client or technical requirements.
7:32 – 9:25 Overcoming Suboptimal Pedagogy: Analysis of learning despite "bad teachers." Learning is a resilient process that can occur even in poorly structured environments if the learner provides their own space for reflection.
10:37 – 12:27 Gamification and Rote vs. Understanding:
Detail: Teaching via play and imagination (common in childhood) remains effective for adults.
Takeaway: Rote learning (knowledge) provides a lookup table, but understanding and wisdom require the ability to recognize patterns and "bend the rules."
15:35 – 18:06 "Cheating" as a Cognitive Tool: Annotating books and using "open book" strategies are framed as brain-engagement activities. The speaker emphasizes that perfection is unnecessary; a "passing grade" in a new topic is a valid step toward mastery.
Detail: Discussion on neurodiversity (dyslexia) and past trauma.
Takeaway: Software development techniques like "Rubber Ducking," writing unit tests for simplified problems ($n=0$), and "showing your work" help bypass cognitive blocks.
30:16 – 31:18 Learning Through Games: Recommends building small, tangible projects (e.g., card games like Higher/Lower) to explore complex language features like C++ spaceship operators in a low-stakes environment.
34:55 – 36:53 Agency and AI: Utilizing William Gibson’s Agency as a reference point to discuss the lack of volition in AI.
Takeaway: Teaching is the act of encouraging free will and personal agency in the learner, which machines cannot replicate.
Detail: Tips for tackling dense material: read non-linearly, use physical bookmarks to reduce eye-bouncing, and "argue" with the text by scribbling notes.
Takeaway: Finishing a single technical book can reveal the vastness of what one doesn't know, fostering necessary humility.
42:22 – 43:30 Intelligence and Subtlety: Referencing Neil Stephenson; intelligence is the ability to handle ambiguity and contradictions. Straightforwardness should be viewed with suspicion in complex systems.
45:35 – 48:17 Challenging Perceived Wisdom: Analysis of community "rules" (e.g., "don't include .cpp files" or "no exceptions in embedded code").
Takeaway: Persistence in asking "Why not?" can lead to community-wide breakthroughs and refined idioms.
50:20 – 53:52 The Emotional Intelligence of Teaching:
Detail: Acknowledges that learning is an emotional process that challenges worldviews.
Takeaway: Teachers must remain patient when students become frustrated or "angry" at new concepts, as this signifies an active internal struggle with existing assumptions.
Detail: GenAI is often used for boilerplate, but the speaker argues for "going meta" (metaprogramming/code generation) instead to ensure correctness and sustainability.
Final Takeaway: Reclaiming the "spark of wonder" is essential for lifelong learning.