A Learning a Day

I thought I’d go deeper into system iteration – one of the lessons I’ve been learning as I’ve been working on Claude Code.

During my first day, I spent a lot of time figuring out errors. Let’s imagine them to be holes in the ship. I’d point them out to Claude Code and I would get suggestions for changes. We’d keep trying to fix the holes in the ship.

However, I’d keep finding old problems recur because fixing one hole often created new ones. And, sometimes, they reopened old ones.

This process became a lot better when I adopted a three-step workflow.

First, use Claude to figure out what was missing and what the gaps were relative to the spec. Claude would identify a lot of the issues.

Second, throw those issues back to Claude Code and ask it: “What changes do I need to make structurally in my spec for this to never happen again?” It would come back with a set of improvements.

Third, ask Claude to make the changes, put it back, and have Claude Code run the job again.

Once I got this workflow going, new issues started coming up instead of the old ones.

It reminds me of a story about two men by a river who saw kids struggling in the current – it was as if they were thrown overboard. One man started jumping in to catch the kids and bring them to the bank. The other started running upstream.

The first man shouted, “What are you doing? Why don’t you help me?”

The other replied, “I need to find the jerk who’s throwing these kids into the water.”

It’s a visceral story but makes the point.

It’s important to solve problems upstream and structurally if you want the fixes to last.

Beware plugging holes in the ship.

I’ve been spending some time building on Claude Code. I started with a simple product: a morning briefing for myself.

The goal was simple. I don’t want to be checking news sites. I don’t want news briefings or notifications. I don’t want any scrolling. I don’t want to be stuck in doom cycles.

Instead, I wanted the most important news from my industry with clear “so what’s” as they pertained to me and my job. I wanted some global awareness news that didn’t require a lot of attention but was enough to not catch me off guard. And some local news.

And I wanted all of this in a one-stop source to create a high-quality information diet.

It started with a simple spec. But as time passed, the spec became more and more complex.

And as simple as this briefing was, it took about two days of iteration before getting it to a place where I was happy.

While some of this was time spent understanding how to get the best out of Claude Code, a big part of this was improving the spec.

I took away 3 lessons –

(1) If you’re able to take the time to plan and think through what you want, you’ll go much faster than if you just start building. Go slow to go fast.

(2) It helps to be systematic when we approach iteration. Initially, I used to just ask Claude what happened and make a fix directly. Every such fix then broke three other things.

Things got better when I go into the rhythm of compiling the execution errors, and asking Claude Code how I could improve the spec to avoid it.

(3) Finally, you can get to 80% quickly. But the final 20% takes a long time.

It takes time to do things well. Do fewer things. Do them well.

A few interesting charts –

The EU generated more energy from wind and solar in 2025 than fossil fuels for the first time.

For all the political noise around renewables in the US, the US Energy Information Administration forecasts 99% of new energy to come from renewables.

A few years ago, the question that followed such stats was – that’s all well and good in developed economies. What about those that are rapidly developing? That’s where this story gets complicated given the cost of development often involves burning fossil fuels.

This excellent chart lays it out beautifully. India is on a fast track in switching away from fossil fuels. China’s switching pace has continued to accelerate.

This shows up in China’s CO2 emissions – which declined a bit in 2025.

To be clear, these decisions aren’t being made because all the parties here believe it is the right thing to do.

They are being made because they are economic no-brainers.

When in doubt, follow the money.

I shared this post on LinkedIn yesterday. It was a collection of hard-won lessons I’ve written about over the years – it felt fitting to share both the update and the compilation here.

---

After nearly a decade, yesterday was my last day at LinkedIn.

I’m grateful for the ride – for the chance to work on a problem I’d been chasing since college, for every member whose feedback kept us grounded especially when they were going through the pain of job seeking, and for the many colleagues who left a mark that goes well beyond work.

A few hard-won lessons I’ll be taking with me:
(1) Problem managers, not product managers. When you’re attempting to build product, everything is downstream of deeply understanding the user’s acute pain.

(2) Throwing an LLM at a problem typically just creates new problems. You have to deeply understand how the system should behave for every user problem – and then measure it and iterate. Everyone says it’s about the evals. It’s not. It’s about the eval loop.

(3) “Shit umbrella*.” The role you play as a leader that the team will appreciate the most is blocking the noise and overhead and ensuring the team has absolute clarity on what they can deprioritize and ignore. (*H/T: Kaitlin for coining a term that will live rent-free in my head)

(4) Unexpected work best friends. People will surprise you if you let them. One of my closest friendships came from someone I’d had a significant disagreement with. When in doubt, have the uncomfortable conversation and leave room for people to change. Including yourself.

(5) Hard day calls. Of all the gifts a job can give you, the biggest might be this – a few people you can call at the end of a hard day and just work through problems together.

For what’s next on all things Careers, do follow Patrick, Emily, and Hari.

I’ll share more about what’s next in a few weeks. For now, just grateful.

PS: No more feature introduction videos from me – you’re welcome. :-)

Aditya Agarwal, former Dropbox CTO and early engineer at Facebook, shared his emotions after using Claude Code over a weekend in a thoughtful post.

Every engineer who has spent time with Claude Code since the release of Opus 4.5 has come to the same conclusion – it doesn’t make sense for humans to write code after this. And it is also abundantly clear that we’re going to be creativity constrained vs. execution constrained.

It got me thinking about human computers – the kind you see in movies like Hidden Figures. These were people who used to do the job of computers. They would compute challenging problems by hand.

Computers completely replaced them. And the idea that there would be a human playing the role of a computer is now quaint.

The same is going to happen to many aspects of/tasks within knowledge work – starting with software engineering. So, a software engineer in 2028 will be doing things very differently from a software engineer in 2018. This is clear.

LLMs do many tasks very well and they’re getting better at longer and more complex tasks at a rapid clip.

It might take time for many to accept this reality.

But it is real.

And we’re going to be better off attempting to figure out how jobs, organizations and careers should evolve, and what comes next.

#OurWorldIsAwesome – Edition 16

---

Olympic National Park is a fascinating place. It contains three distinct ecosystems: temperate rainforest, a beautiful rugged coastline, and glaciated mountains. There are very few places in the world that combine all three.

The rainforests within the park collectively form the largest temperate rainforest in the world, with annual rainfall of almost 12 feet – one of the highest in the US.

Crescent Lake is stunning thanks to just how clear it is.

There are beautiful old sulfuric hot springs. The forests in which you hike feel like they could easily be 1,000 years old. The trees really feel like old growth trees and they transport you instantly to a different ecosystem.

There’s so much water flowing around the park. And every spot is just picturesque.

The National Park property at Kalaloch Lodge along the coast is stunning in its simplicity, look, and prime location. The tide pools at the beaches are lovely too.

The park truly transports you because of just how lush and pretty the landscape is.

Extremely beautiful. And special.

#OurWorldIsAwesome – Edition 15

---

Death Valley National Park is the largest national park outside of Alaska, covering 5,270 square miles. It is larger than many states in the US.

This is a fascinating place. It recorded the highest temperature ever on earth. Badwater Basin is the lowest point in North America at 282 feet below sea level. And it receives less rainfall than the Sahara Desert. This was Badwater Basin after some unexpected rain.

We spent two days in the park. And it was like being on another planet.

The valley was one of the filming locations for Tatooine in Star Wars. That made the place a bit extra special for us.

Throughout our time in the park, we felt like we were in an otherworldly location. It was hard to spot anything that is living. And that adds to the out-of-this-world experience.

The National Park Lodge at Furnace Creek is the center point for most of the activities around the area. It almost feels like a throwback in the middle of the desert.

It isn’t candidly the kind of place we would want to keep going back to. We love trees. And the sheer absence of them made us miss them more (and we went here right after 2 days at Joshua Tree National Park).

But it’s definitely one of those places that is worth visiting in a lifetime. Just to appreciate how unique it is.

#OurWorldIsAwesome – Edition 14

---

The trees that give Joshua Tree National Park its name aren’t trees at all. They’re a species of the yucca plant, similar to orchids.

They’re fascinating because they grow just 1 to 3 inches per year and have a lifespan of nearly 150 years. And they give the park its unique look.

The look like cousins of cacti – who seem to glow in the morning sunrise.

Most of the park is designated wilderness. And the stony terrain makes it a really fun place for casual rock climbing. Our kids had an absolute blast.

This was definitely a park where we missed trees. But it made for a unique and fascinating desert experience.

#OurWorldIsAwesome – Edition 13

---

The Tadoba Andhari Tiger Reserve is the largest and oldest national park in Maharashtra, covering 625 sq km. It is home to one of India’s most thriving tiger populations with over 90 tigers – one of the highest densities in India.

The name “Tadoba” comes from a local tribal god “Taru.” Legend holds that Taru was a village chief who was killed in an encounter with a tiger. He was deified, and a shrine dedicated to him now exists beneath a large tree on the banks of Tadoba Lake. “Andhari” refers to the Andhari River that meanders through the forest.

Beyond tigers, the reserve is home to Indian leopards, sloth bears, gaur, wild dogs, marsh crocodiles, Indian pythons, and around 280 species of birds. Summer temperatures can reach 47°C (117°F). And it was once used as hunting grounds before being declared a national park in 1955.

We were fortunate to see a collection of birds and animals like…

A painted stork

A peacock

A crocodile

A sloth bear

A “changeable Hawk Eagle”

However, after a first tiger safari at the beautiful Nagarhole National Park where we didn’t see a tiger, we were hopeful our second attempt at Tadoba would be more successful.

And we hit the jackpot in our first attempt.

A female tiger named Chanda spent an entire hour with us just hanging out. She relaxed, yawned, ate some herbs because she wasn’t feeling well. Walked through the foliage. Marked her territory. Chuffed. Hung out near a water body.

One of the most epic experiences. One that we’ll never forget.

We also got to see her father, Chhota Matka – also called CM. He was just recovering from a deadly fight with another male tiger who had challenged his territory and severely injured him. So he spent his time sunbathing and relaxing by a stream. He’s massive and a famous tiger in these parts.

We didn’t catch any other glimpses of this majestic cat after that first attempt. But that hour of sighting remained the highlight.

Having had the opportunity now to see lions in the wild as well, you just realize that tigers – both in terms of their size and their incredible fur – are truly the most charismatic of mega cats.

It was an unforgettable experience through this iconic forest reserve.

#OurWorldIsAwesome – Edition 12

---

Nagarhole National Park – also known as “Kabini” – sits on the banks of the Kabini River, with Bandipur National Park on the opposite side. It makes for a picturesque setting – dense forest, water, and one of the highest tiger concentrations in India.

We stayed at the Jungle Lodge, maintained by the park authorities and transformed from its British hunting-era origins. A fitting evolution. The food felt home cooked and delicious — a small thing that added to the warmth of the stay.

As this was pre-DSLR, I don’t have too many photos and none that do justice to the beauty of the forest.

The rangers who took us on safari had experience across different parks in the country. They could speak to both the vegetation and the animals and birds with equal ease. That kind of knowledge makes a difference.

Safaris are controlled and run through the Jungle Lodge, with fewer than 20 vehicles going out each slot. No overcrowding, no convoy of jeeps jostling for position. It made the whole experience feel more magical – just the forest and the possibility of what you might see.

Despite four attempts, we didn’t spot a tiger. Sometimes these things don’t go your way. But my biggest memory is looking through the mist on both days, hoping to catch sight of a big cat.

And then, for a few seconds on our last morning, a leopard crossed the road and disappeared into the bushes.