SecondBrain
Ask the Brain
Index/Sourceupdated Sat May 16 2026 08:00:00 GMT+0800 (Philippine Standard Time)

Learning Software Engineering During the Era of AI (Raymond Fu, TEDxCSTU)

tedxsoftware-engineeringeducationaicareerfull-stack

Learning Software Engineering During the Era of AI — Raymond Fu (TEDxCSTU)

Talk by Raymond Fu, CSTU professor and veteran technologist, at TEDxCSTU (May 2025) on TEDx Talks. The thesis: if AI can write code, is it still worth learning software engineering? Yes — but the job has changed, and so should the curriculum. SWEs become orchestrators, system architects, and AI collaborators; "programmer" is the wrong word.

Key claims

  1. The prediction came true, just not the way it was sold. Fu's professor in 2001: "every job will be a programming job — a golden ticket to job security." Twenty-five years later, GitHub's CEO: "the future of programming is natural language." Echoes Software 3.0 (Karpathy) but Fu reframes it as a question about the worker, not the paradigm: have we lost the golden ticket?

  2. The 55/30 paradox — "55% of developers today are starting to use Copilot, but only 30% are accepting the outcome without any changes. If you're not in 55%, you're in trouble. If you're in the 30%, you may be in bigger trouble." The 30% is a cognitive-offloading rate for working engineers. (Fu doesn't cite a source for the numbers; treat as directional.)

  3. AI is a brilliant junior developer. Fast, capable, needs a human to define vision, validate results, and ensure the work serves the user. "It's up to us humans to define the vision, to validate the results and ensure what we're building is good for the society." Same mental model independently adopted by Boris Cherny and Andrej Karpathy.

  4. "AI raises the floor; software engineers raise the ceiling." (Memorable line he asked the audience to applaud for, and they did.) Direct echo of Vibe Coding vs Agentic Engineering in different language. Three reasons SWEs remain essential:

    • We understand AI under the hood (models, data pipelines, limits)
    • We use AI to build production-ready systems (scalable, reliable, maintainable), not just demos
    • We make AI better (fine-tuning, optimization, usability) — "the next generation of AI is still built by software engineers"

  5. The future-SWE curriculum. Go beyond writing code as fast as possible:

    • Master foundations (data structures, algorithms, programming concepts)
    • Think like an architect — meet senior-engineer expectations early
    • Go full-stack across disciplines (design, product, data, project management)
    • Practice communication and collaboration via team projects
    • Use AI as a creative partner — learn LLMs, generative AI, fine-tuning, RAG
    • Stay adaptable — "tools change, principles last"

  6. "Programmer" is the wrong word now. The SWE of the AI era is a visionary (defines meaningful problems), bridge-builder (connects tools, teams, disciplines), and leader (of humans AND AI).

Memorable lines

"The best engineers are not the ones who code the fastest, but the ones who think the deepest."

"AI is raising the floor, but software engineers are raising the ceiling."

"In a time when AI is everybody's assistant, engineers become the orchestrators. We remove barriers and open doors."

"The future doesn't belong to those who code the fastest. It belongs to the ones who think deeply, adapt quickly, and collaborate efficiently."

What's new vs what's in this vault

  • Reinforces — Vibe Coding and Agentic Engineering (floor/ceiling split, now in three voices: Karpathy, Boris, Fu); Code Is Free (the implementation-is-no-longer-scarce premise, here framed for educators).
  • New angle — first source in this vault aimed at CS education rather than industry practice. Fu is talking to students and parents, not staff engineers.
  • Cross-source contradiction — see Cognitive Offloading: same diagnosis as Charlie Gedeon, opposite prescription. Fu's "embrace AI" vs Gedeon's "design AI to resist."

Open follow-ups

  • The 55%/30% Copilot statistic is uncited. Likely from a GitHub or developer-survey publication — worth pinning down before quoting.
  • GitHub CEO quote "future of programming is natural language" — easy to verify (probably Thomas Dohmke); pin to a specific keynote/interview if cited.
  • Fu's full-stack-engineer claim contradicts the "T-shaped" specialist trajectory often advocated elsewhere. Worth a query if a second source pushes back.
  • The CSTU TEDx event has a small footprint (low view count, niche school). Treat the talk as one practitioner's framing, not industry consensus — but the framing itself is on-thesis with much bigger voices in this vault.

Sources