deepGem

I don't use TikTok but even in my mid 40s fell victim to Instagram reels. What's astonishing to me is that I have a level of self control that's way more than the average individual (for instance I have taken cold showers for ever).Even with such an otherwise resilience I fell victim to Instagram reels and their infinite supply of mindless entertainment. It took a week for me to unshackle and get rid of this imbecile nonsense. I still have the app on my phone but am actively not opening it. My hypothesis is that if you can do this for a continued duration, then the existence of the app won't bother you. You can't go around deleting everything from your life. You lead your life on your terms, despite the distractions.

So the addictive forces are quite powerful and for the average individual, it's really hard to get out of this addiction.

I was just having a chat with my buddy today about how Apple, Microsoft and to a large extent Google are largely the best positioned to profit from any upstream applications. While the likes of Facebook, Instagram and TikTok fight for viewership, these underlying platforms are just counting profits regardless of who reigns supreme on the top. I don't know if Jobs saw this future unfold but here it is.

The gas stations of the digital era are just doing their job, ensuring that there is enough gas supply. What's ironic is that the biggest gas company also created the first automobiles, and basically told the market, everyone is welcome to build an automobile to use my gas as long as you give me 30% of your profits. Oh BTW, your automobile can run only using my gas. Boy, imagine this in the physical world.

To understand how difficult distributed computing is, at the container level of abstraction, one has to try Docker swarm.

My first interview out of college in 1999-2000 was something like this. A full day interview about designing the components of a TCP-IP stack interspersed with some personality/behaviour interviews. End of the day, you either knew you had an offer or you didn't. I had an offer. I would kill for such an interview experience now.

The last interview I attended stretched across a month or may be more. Glad I got the job but boy what a nightmare.

If someone told me today that I would finish all rounds of interviews in one day and this became a standard, even figuring out for hedging I would take a week off, schedule 5 interviews and be done.

Even fundraising for a startup doesn't take 2 months these days, whereas an interview will. That should tell you how broken the process is.

I must have clarified, what I mean by strong consistency is that you have transactional support (2 phase commit). MongoDb is atomic, meaning every single document is strongly consistent. However, I stand corrected. They have introduced transactions as of 4.2.

What I mean is that the data model in the NoSQL world is tightly coupled with the query pattern. So you define the query pattern and then tailor the data model to that query pattern. In the relational world, you typically choose the index based on your query pattern. Not tailor the data model to your query pattern. You follow normalization principles.

Of course, every data store needs a data model. No debate there :).

Not sure what you mean by relational databases are relational layers on top of key-value stores. InnoDB has a 16KB page data as it's fundamental data structure.

https://dev.mysql.com/doc/internals/en/innodb-page-structure...

True. Spanner and the likes of Spanner, CockroachDB, YugaByte all are strongly consistent and scalable dbs. The greatest advantage IMO is the ability to just use SQL without having to worry about carefully designing a data model. What bothers me however is that these data stores are not truly relational data stores. They spin a relational layer on top of a scalable key-value data store.

Is it necessary to use a strongly consistent transactional data store if your needs don't demand transactions, by transactions I mean 2PC. IMO you are still better off with DynamoDB/Cosmos/MongoDB for eventual consistency use cases. The reason being, you have to resort to a data model if you don't need the relational layer in YugaByte at least, not sure about Spanner. So why bother with Yugabyte if am resorting to a data model. Might as well stick with DynamoDB.

15”/13” MBPro with the Touch Bar and that horrendous keyboard.

You don't have to bend to every VCs demand. WhatsApp ran very lean and perhaps had 50 engineers tops. Every big tech company ran very lean until they hit scale Series D or so. So the headcount growth is not just a function of VC, but a function of people's desire to build empires. Even in many social gatherings managers boast of the number of engineers they are managing. So it's another flaunt metric for a bulk of founders and managers. Don't just blame the VCs.

This is what beats the heck outta me. A company that is sitting on nearly 150B in cash somehow feels the need to pinch 30% commissions from developers till date. I understand this as an initial business model. I mean a 15% reduction in app revenues is not even a rounding error in Apple's P&L. What the hell are they thinking. The goodwill that they'd earn from devs will go a long long way and if they signal that their share will eventually go to zero over a period of 10+ years, that'll get more devs to embrace iOS. I fail to understand the current leadership at Apple.

I donno. I keep thinking about this all the time. I might also be blinded by the intricacies of where I live, Bangalore. Cancellations by Uber drivers are rampant here and it's very hard to find the black equivalent. The UberXL service which counts as the black equivalent has absolutely zero predictability. My guess is that there are at least 10,000 - 20,000 people in Bangalore, a city of 10M willing to pay a 2 - 3x premium for predictability and good QoS. There simply is no such option for non airport rides.

There are those who pay for business and first class though. It’s a matter of figuring out the price per km and commission dynamics. With the appropriate incentives and commissions supply should ease out. Obviously am trivialising this approach. It will be great to understand why Uber Black couldn’t scale.

I think the market is ripe for a real Uber Black. A super premium service that guarantees a cab no matter where you are within a minute. No cancellations. Your own personal on demand limousine. People are so tired of cancellations and not finding rides, I bet some will pay a premium just for the predictability that you can go anywhere whenever you desire.

I usually don't catch satire that easily but somehow this headline screams satire and the first line was a dead giveaway.

It may not be preparation for that particular interview or task but there is preparation elsewhere that's paying off here in the interviews. Perhaps even sub conscious preparation.

One such preparation happens at school when you prep for debates and speeches. This may not be apparent at that time but I see those who are very coherent in interviews have had some debate prep in their schools or colleges. Just quoting an example. Any public speaking prep actually helps in stitching together multiple ideas as you are talking and as you converse and do it in a coherent way.

I really wish many people had these skills. Most of the meetings I sit in, people take 30 mins to convey what should ideally take like 10 mins. 20 mins is just blabber.

The other counter intuitive prep that personally helped me was GMAT verbal. What started as pure hatred turned into a treasure trove of brevity and coherence. Especially the reading comprehension aspects. A month or two you spend through that exercise really helps in the long run. You get pretty good at communicating your ideas coherently.

Even better luck debugging in the enterprise without logging. You won’t even know which machine your code is running on and in which machine Splunk is spitting out the logs. Without printing time stamps and debug worthy information to the logs, you are fucked, period.

Different time. They did this mostly because they knew they wouldn’t be able to sustain the demand as storage was still expensive back then.

GMail was also in beta for a really long time like > 2 years.

Another side effect is that you have to now think of separating the expensive parts of the code from the non expensive parts because that code has to be offloaded from the blockchain itself into side chains or something else.

Security is another aspect that has to be brutally evaluated, even at the OPCODE level. This is really strange.

TL;DR it's almost like writing assembly level code with a 1000% emphasis on security, despite writing code in a high level language.

There was a major hue and cry by founders, esp in Bangalore about how ungrateful the prospective employees were, when they ghosted the founders after accepting offers. These founders never really tried to understand the simple market dynamics and that they would do something similar to employees, should the market go down.

The playing God shit by founders has to end and this move by Coinbase gives a huge boost to employees to hedge their offers. I won't be surprised if employees accept 2 offers, join one and after a a couple of days say no to the other or ghost them.

This is exactly how I approached Leetcode. TBH the grind has had satisfactory effects even though the immediate practical impact is just for interviewing. I don’t use it as a tool for interviewing, just as a tool of list of things to get done. I should just pour more energy into it and do it every day.

I also am stealing this principle ‘My scheduling principle is to do the thing I hate most on my to-do list’

The moment you have to work on a large code base either in Java or C++ you almost never have enough resources. 64GB RAM is fair, 32GB bare minimum to work on such codebases. Otherwise everything slows down, IntelliJ will take 2 seconds to open up etc.

If you don't work on such code bases then yes 16GB is more than enough and 8GB you can live with on your day to day life.

"Self-belief alone is not sufficient—you also have to be able to convince other people of what you believe."

This is the true test of self belief. In the corporate world this is why many ideas don't take off or take off in a half baked manner. Just like self belief is visible, the absence of it is also visible. I can quite figure out when someone is proposing an idea or wants to make something work but they themselves don't believe in it. The corporate world is littered with such people and leaders as well.

My boss put it very well 'The moment you think what's in it for me, you've lost the plot". This question does not arise in anyone who has this utmost self belief in what they do.

I lived through the 2008 meltdown. The goddamn crisis was more than real. In front of my eyes, people lost their jobs including my own colleagues. No one had money. Even those with jobs came to work fearing it was their last day. I just hope such a day never comes again. September to December 2008 and even Q1 2009 was really bad.

Borrowing shovelware. This is gold Jerry!

25% is if the organization is ultra efficient. I think most organizations other than select FAANG/Stripe are at 75%+. I think pmarca quoted the number to be something like 1:4 which makes the efficiency at 25%, 75% dead weight. FAANG is probably at 50%.

The good big companies are overstaffed by 2x. The bad big companies are overstaffed by 4x or more.

https://twitter.com/pmarca/status/1515195878604087297

Yep highly encourage to take a breath, at some regular cadence and contemplate on all possible paths. I don't know how to be effective at this though. You need some notion of feature management to accomplish this.

I am looking to build a similar tool for engineers who are switching jobs and have multiple job offers. Essentially a financial projection of options 1, 2 and 3 taking into account inflation, location and salary growth.

Happy to brainstorm and collaborate if you are up for it.

If you decide to offer advisory services on your own, then yes there might be some regulatory hurdles. You are not going down that route. Match making should still be fine, am not sure.

Agreed about turning off existing users, but you can carefully orchestrate this to help users find appropriate advisors as a completely opt-in service. I think you can find a scalable monetization model in match making. Not sure what your plans are, but yeah unlikely you'll be able to pull it off in solo mode.

Totally impressed with what you have built.

To add, the non tech crowd doesn't really know or have the time to learn about investing. They just need someone to manage their finances. The advisory profession is alive mostly because of this. Instead of acting as a lead gen to advisors, perhaps you could add more value by doing some sort of filtering or match making based on user's data, totally at the discretion of the investor/user. This will be incredibly helpful. You could become some sort of a scaled arbitrage between investors and advisors - a huge gap that exists today.

The pain point raised was about annotations. So just providing an example framework that doesn't use annotations or any other code magic.

Java annotations are terrible to reason about. Once you get a hang of them, then it's somewhat easier. That being said, Spring takes these annotations to the next level of complexity. Spring boot just makes this more difficult to reason about with added annotations. Anyone who states that Spring annotations are just 'abstractions' and that's how software is built is welcome for a healthy argument.

There are much better web development alternatives such as Javalin.

Compare this

import io.javalin.Javalin;

public class HelloWorld { public static void main(String[] args) { Javalin app = Javalin.create().start(7070); app.get("/", ctx -> ctx.result("Hello World")); } }

to

@SpringBootApplication @RestController public class DemoApplication {

public static void main(String[] args) { SpringApplication.run(DemoApplication.class, args); }

@GetMapping("/hello") public String hello(@RequestParam(value = "name", defaultValue = "World") String name) { return String.format("Hello %s!", name); } }

However, Spring cloud, has all the necessary distributed features built in. While I haven't used these, not having to write code for these is a god send. Load balancer - Check Service mesh - Check Circuit breaker - Check

Heck, they have even included all Netflix OSS components. So this alone is worth the pain of living with annotations.

https://spring.io/projects/spring-cloud