Feynman's Explanations
Notes from the Richard Feynman lectures on The Character of Physical Law(1964)
I recently came across this series of lectures by Richard Feynman. I’m going to assume you are familiar with him and know his work. I was particularly interested in two meta things about the video rather than just the content. His presentation of deep topics to laymen and the art of discovering new knowledge.
The following are my first reflections on the video, formatted a little. How should this apply to Product Development? It feels like an important piece in working with AI and communicating what we discover with the world.
Presenting Deep Topics
I get told off by my Product Manager all the time about this. It’s why I’m making explicit notes on the the topic. The tendency of the Engineer is to drive directly to the heart of the matter or the first roadblock.
Feynman would probably nerd out with you on it in private, but would certainly not approach it the same way. It’s hard to tell from the lectures whether he does this implicitly or it’s a practiced thing. Regardless, the strategies he employs are somewhat timeless in the application of sharing deep topics.
Admit an unavoidable depth
“It is my purpose…to explain really why I cannot satisfy you if you do not understand mathematics…mathematics is not just a language, it is a language plus reasoning.”
Be upfront about the complexity inside your topic; honesty builds trust faster than hand-waving.
A good analogy is the deep dive. Deep dive over what?
Imagine it as a cliff. Avoid the tendency to start talking about what there is on the way down before even telling your audience that there’s a cliff ahead.
Use layered analogies
When asked “How big is a foot?” he answers: “Six feet tall is one hundred and seventy thousand million hydrogen atoms high.”
Anchor abstract metrics (e.g., semantic similarity) to everyday scales your users already feel.
Like the cliff analogy, putting your topic into everyday, visual terms means you can get to the core of the idea faster adding on only the necessary details to carry on your point.
This strategy also forces you to understand the topic better and comes with a learning technique named after Feynman.
Stretch imagination—then hold it
“Our imagination is stretched to the utmost just to comprehend those things which are there.”
Good demos create an “aha” moment that widens mental models without hiding the weirdness.
Explicitly design where those moments are as you descend into the deep dive. Deep topics are where all the interesting things are after all.
Discovering New Knowledge
Feynman’s playbook for discovery translates neatly into product experimentation with AI. As backlogs shrink and product development shifts from pre-planned execution to real-time discovery, these principles feel more essential than ever.
In the future, engineers may spend less time pulling tickets and more time running and reviewing experiments. That means getting much closer to users, observing real behavior, and enabling systems that allow for rapid iteration and structured feedback loops.
Below are some of the core behaviors and a quick lens for applying them.
The Three Step Method
“First we guess it; then we compute the consequences… and then we compare with experiment… If it disagrees with experiment, it’s wrong. That simple statement is the key to science.”
Discovery is an iterative cycle of bold conjecture and merciless testing. Theory has no privileged status.
He got a good laugh out of the audience when he said to guess and I don’t think that was one of the aha moments he intended. It’s hard to quantify all the inputs that should be used in a decision and I think this all reduces to one of the core components that the AI Industry keeps harping on, taste.
Taste will remain when all else is done by AI SWEs.
For me, taste is guessing well. Which includes the followthrough on the experiments.
It goes something like this:
Guess → craft a hypothesis: “If we auto‑summarise support tickets, CS handle‑time falls 15 %.”
Compute → prototype or simulate; estimate the impact with offline data.
Compare → A/B‑test live; delete or double‑down strictly on the numbers.
This loop will accelerate and we’ll be able to do it on smaller pieces of our stacks.
Fail Fast
“We’re trying to prove ourselves wrong as quickly as possible, because only in that way do we find progress.”
Progress speeds up when scientists design experiments aimed at breaking their own ideas.
We should be designing our software with this in mind. In fact, that’s the core thesis behind the MVP.
Steering the team toward this requires constant effort and also takes a healthy dose of the next one.
Beware the unfalsifiable
“If the process of computing the consequences is indefinite, then with a little skill any experimental result can be made to look like an expected consequence.”
Vague models insulate themselves from refutation; clarity is a virtue because it allows a theory to fail.
Pretty much the definition of a SMART goal.
Product cue:
Define success metrics before you peer at the dashboards.
Time‑box experiments and commit to rollback thresholds.
Keep dashboards public; collective scrutiny resists narrative‑driven data fishing.
Closing thought
A theme right at the end stuck with me. I feel a constant tug in Engineering that if only we stuck with pure CS principles and good Engineering practice, we could do all and explain all.
It’s far from the truth—and to quote Feynman:
“An understanding of the physical laws doesn’t give an understanding of significance… the details of real experience are very far off from the fundamental laws.”
That’s as true as anything else he said in the lecture.
So, I’m left thinking about Product Engineering in a state of constant discovery as the industry shifts again underneath us with AI.
As ever, our job is to make these discoveries through real experience—by rapidly proving and disproving hypotheses in Product, and finding the clearest way to communicate what we learn.
Our role is to bridge the gap between fundamental laws (data, algorithms) and their significance in the world. Feynman’s twin arts of clear exposition and relentless inquiry remain the most durable guide I’ve found.