Don’t Turn Your Brain Off
This article is a curated collection of industry perspectives on staying mentally engaged in the face of LLMs.
I continue to pay attention to material that motivates me to learn and helps me improve my technical expertise. On the surface, it could seem that a lot of that is no longer necessary if AI can do it.
In a similar tone to the conversation with Jeremy Howard, Chris Lattner reiterates the importance of keeping developers’ brains engaged. He looks for individuals who have a “How hard can it be? Let’s figure it out.” attitude and are willing to tackle things that are doomed to failure.
As I see it, aiming to tackle hard problems is a great argument to motivate to grow competence and understanding. As it stands, LLMs are good at implementing or assembling different parts of well-known solutions. However, going beyond that requires deep expertise. It reminded me of what Andrej Karpathy told on Dwarkesh Patel podcast about building nanochat:
“I would say Nanochat is not an example of those [tasks AI is good at] because it’s a fairly unique repository… and it’s not boilerplate code. It’s intellectually intense code. And everything has to be very precisely arranged. And the models were always trying to… keep misunderstanding the code because they have too much memory from all the typical ways of doing things on the internet.
Andrej Karpathy on Dwarkesh Patel podcast, Oct 17, 2025
Other examples in the article also focused on how AI affects education and students. As summarized, the emerging consensus among educators is that it remains vital to learn the fundamentals. Without this foundation, students may struggle to audit AI solutions for correctness.
However, AI may be disrupting traditional pathways to learning: educating sequentially through basic problems, developing critical thinking by making mistakes, and ultimately learning from those errors. This process can lead to reaching final solutions too quickly, creating an illusion of learning.
Maybe you’re not Actually Trying
People are selectively agentic, and while they may be high performers at work, they could be completely helpless in relation to self or others. If this is the case, find the part of your life where you’re still acting like your old self, and start solving it with the same seriousness and creativity you use for important work problems
There is one theater of life where they’re not Actually Trying — where they’re approaching serious problems with the resourcefulness of a teenager, though they are now capable adults.
Warren Buffett’s last letter to shareholders
Warren Buffett’s reflection on his upbringing and life’s work. It is a sober reflection on his life’s circumstances. It points out the luck he had, especially when it came to health and longevity. It also shines a light on the people who surrounded him throughout his career.
What I found especially interesting was his outlook for Berkshire Hathaway beyond his reign:
With a little luck, Berkshire should require only five or six CEOs over the next century. It should particularly avoid those whose goal is to retire at 65, to become look-at-me rich, or to initiate a dynasty.
This is long-term thinking. I rarely hear or read about it. This holds true even when it comes to countries, let alone companies.
And a few great final thoughts:
Greatness does not come about through accumulating great amounts of money, great amounts of publicity, or great power in government. When you help someone in any of thousands of ways, you help the world. Kindness is costless but also priceless.
Choose your heroes very carefully and then emulate them. You will never be perfect, but you can always be better.
How to Enable iPad Features like MultiTasking & Stage Manager on iPhone
You can modify a system plist to control the system capabilities. This modification forces the iPhone to be treated as an iPad.
Bret Victor The Future of Programming
This is a 12-year-old conference talk that was constructed in a way to be timeless. Bret acts in this talk as if it were 1973. He goes through “recent” discoveries in the field of software engineering. He tries to foresee the trends for the future.
I learned about past projects. One example is Sketchpad, which looks incredibly ahead of its time by allowing users to draw on the screen, and the program to redraw the lines to get a perfect rectangle:
Essentially he’s created a program that draws a rectangle, but he didn’t do it by writing code, he did it by directly manipulating the data and directly applying a set of constraints to them.
Another example is a Planner system, that moves programming from procedures to goals and constraints:
You express your program in terms of the goals, the results that you want from the program… letting the computer itself figure out how to do it.
Throughout the talk, Bret has creatively demonstrated that software engineering has largely taken a different (and possibly worse path). However, the main idea was not to highlight the path we took. It was to show how disappointing it is that we stopped questioning once we got there.
Even more of a tragedy… would be if these ideas were forgotten. If anyone were ever to be shown this stuff and actually be surprised by it
The most dangerous thought that you can have as a creative person is to think that you know what you’re doing. Because once you think you know what you’re doing, you stop looking around for other ways of doing things. And you stop being able to see other ways of doing things. You become blind.
I think this talk fits perfectly into the new paradigm shift brought on by LLMs and other AI tools, or what Andrej Karpathy calls Software 2.0 and 3.0. We need to stay open and curious about experimentation. With these new tools, that should be easier, especially now that it’s widely accepted that nobody really knows the “right” way to do things yet, unlike in traditional Software 1.0 development.
As the talk revealed why we had so many exterminations to software engineering techniques as in the 60s or 70s:
It was late enough that technology had kind of got to the point where you could actually kind of do things with computers, but it was still early enough that nobody knew what programming was. Nobody knew what programming was supposed to be. And they knew they didn’t know. So they just like tried everything. Their minds were totally free… They just, you know, tried anything they could think of.
Netflix’s Engineering Culture
Netfix has high-agency, high-trust, low-bureaucracy culture with strong talent retention. This is achieved, at least in part, by historically hiring only for senior roles, maintaining a high bar, and paying top-of-market salaries. Recently, they shifted to hiring new grads and early-career talent. Stone explains this was necessary to gain new perspectives and skills (specifically “native AI” familiarity). However, the “talent density” bar remains extremely high; they look for early-career individuals who possess the same level of curiosity and ownership as senior staff.
Netflix does not operate on a top-down, command-and-control model. Innovation is expected to be driven by the teams on the ground. Engineers are given massive autonomy but are expected to be “unusually responsible” for both successes and failures. The company does not conduct traditional annual performance reviews with ratings. Instead, they rely on a culture of continuous, timely, and candid feedback.
What I didn’t know is that Netflix engineers build the infrastructure for the entire lifecycle of content. This ranges from the software used on-set during production (“Pitch”) to the encoding, promotion, and final delivery via Open Connect (“Play”). What was also surprising that approximately 1 in 5 Netflix engineers contribute to open source.
“When people ask me like, what’s the Netflix value that most resonates with me… it’s curiosity. Asking questions. Questioning whether we’re solving the right problems in the right way.”
Elizabeth Stone, Netflix’s Chief Technology Officer.
Android development, Jetpack Compose, Kotlin
I’m trying to catch up with modern Android development practices. I’m way behind on newer practices that involve Kotlin and Jetpack Compose.
I started going through Android Basics with Compose, Kotlin Tour, and watched KotlinConf’25 to catch up on the most recent developments. It’s interesting to compare the similar developments on the Swift side. However, Kotlin with JetBrains seems to be more developer-ergonomics focused and more willing to introduce new tools in their early alpha stages. Kotlin Multiplatform is intriguing. I’ve been hearing about it for years, but never tried it.
One potential benefit of AI coding tools is that choosing a single language or tech stack (for example, Kotlin for Android, iOS, and the Web) may become less important. Instead, they can lower the cost of switching between different technologies that are best suited for each situation. I subscribe to this idea, but time will tell.