Notes

I realized an important thing while reading Remote Inc. book. I should focus on impact or outcome, not other small stuff such as responding to Slack, number of hours worked etc. No one remembers the latter after awhile, only impact matters. Not so much as I learned but rather remembered: perfectionism sucks. After a point, people who evaluate my work won’t notice the tiny details that make the work perfect (in my mind). In many cases, getting the work done and creating impact is more important than doing it perfectly.

W4: D4: 7 on squats! 5 each on 125, 135, 145. 3 on 155, 165, 175. 2 on 185 but let’s call that the max. D3: 6 on bench. Overall, good workout on the bench today. D2: 7 on bench. Just hold the barbell tight and pull out your chest. 70% at 105 for deadlifts. 5x hangs. D1: 10 on squats. Bench felt heavy, maybe my form isn’t good enough. Need to fix that soon. W3: D3: 6 on bench. Increased bench max by 5. It’s time to add some accessories in between major sets. D2: 8 on bench. Deadlift: I increased weight to 145. D1: Legs paining from the half marathon. 10 on squats. 6 on bench, so keeping right next week too as is. (It’s possible I had an extra 10 lb plate though.) W2 D4: 6 on squats. Increasing max. Sumo: Start with 125. W2 D3: 7 on bench AMAP. Increase bench max to 95. W2 D2: 7 on bench AMAP. I should create 3 columns in weight to plate sheet that contain warm up plate count. (Too much brain to use otherwise.) Deadlift: 135. W2 D1: Check bench press form. 80 was hard. W1 D4: In at 6:45 am sharp. 9 AMAP on squats! 2x6 sumo with 120 lb. W1 D3. 12 squats in AMAP! I learnt about shoulder retraction or whatever during bench press. Need to continue doing that next time and see more videos. W1 D2. Gym at 6:45! Bench: Look at form. Why do I get something in the shoulders? Deadlift: Max of 5: 125 Done with primary 2 at 7:20, so have time for an accessory. OHP: 45 lb. 3x7. Protein shake was too cold!

Removing distractions, getting bored (instead of mindlessly checking my phone) etc. is great. But I also need to fill some of that time with some activities. I have this (new) nagging itch to start learning Rust. I should resist that and try to build something instead. I’ll probably not crack Google’s staff interview as well today and that’s fine. I should still continue interviewing regularly. For example, once a month.

https://www.gobytram.com/ Powell’s Books and http://www.everydaymusic.com/ are close by.

Long form blog posts, those that are 500+ words long. Learn to speak clearly and slowly. Real things: Food: learn cooking. Art and craft: Paint the town, maybe once a month. Quilling - we anyway have the raw material at home. Fill in a coloring book. AI models to: Optimize ETF selection for the long term (as opposed to trade stocks in short term). Software: Create open source alternatives to nice but closed source software, similar to Headscale for Tailscale.

https://www.cookwithmanali.com/instant-pot-quinoa-khichdi/#wprm-recipe-container-41394 Quantities: 3 times the measurements but same cooking time. So, multiply all numbers, including below, by 3. 2, not 2.5, cup water. 1/3 tsp turmeric & chili powder each. 1 tsp ginger garlic paste. 1, instead of 0.75, teaspoons salt. Other steps: Onions -> ginger garlic paste -> masalas -> tomatoes -> peas and other stuff. Sauté everything, before and including tomatoes, for awhile.

Clarify requirements. Highly important: understand who is going to use the system and how. Call it out explicitly and don’t make assumptions. Internal or public facing. How can you extend the system in future, but that’s far less important than solving all the use cases I am given. (While obvious, I didn’t properly do the latter in all the interviews.) Security requirements. Non-functional requirements, give numbers where possible: Availability vs consistency: what matters more? Availability, but don’t discuss if some components can have lower availability. (No one wants to design a low availability system in an interview.) Consistency: strong vs eventual. Can some operations be asynchronous? Latency. End to end latency for an operation. For e.g., if an event enters your system, how soon do you want to process it and present results to end users? Design considerations. Back-of-the-envelope estimation: only do those that make sense for the question. Examples: If there are mobile users, estimate data size. API design. High level design. Standard stuff: Telemetry (i.e. logging and metrics). Based on interview’s choice: ...

Number of seconds: 1 day: 100k 1 month: 2.5M Number of users: USA: 300M Social networks (i.e. daily active users): Facebook: 2B YouTube, WhatsApp: 1B. (YouTube hasn’t published its DAU but it’s definitely going to be more than Instagram’s, so 1B should be fine.) Instagram: 500M Snapchat, Twitter: 200M Reddit: 50M Proportion of users who create vs consume: 1% 80-20 rule: 20% things get 80% traffic. So, assume cache rate of 20%. Machine: Number of connections an average modern server can handle: 50k Latency (see more here, ignore other Google search results as the numbers there are outdated): Read 1MB from: Memory: 3 us SSD: 50 us Disk: 825 us Network round trip: Same data center: 0.5 ms Halfway across the earth: 150 ms ID: Assume number of bytes for a string = 2 * length. Assume you use [a-z, A-Z, 0-9] to generate your IDs. Then 62^8 is a huge number that should suffice for most use cases. UUIDs are 16 bytes. Bytes: KB =1e3 bytes. MB = 1e6 bytes. GB = 1e9 bytes. TB = 1e12 bytes. PB = 1e15 bytes.

Now that I’ve found moderate success in reducing distractions, thanks to the Deep Work book, I have some free time on my hand and I need to figure out what to do during that. Here are some ideas that come to mind. Finish the Designing Data-Intensive Applications book. Continue with the D2L book, but only during work hours. (Amazon should be paying me for learning a new skill.) (Finally) build something. Could be some software, machine learning model, (naive) art etc. I should probably start jotting down ideas. What I don’t want to do: ...

I recently setup a personal Algo VPN server. Once I did that, google.com started redirecting me to https://www.google.com.hk/, presumably because it thought the Oracle server was situated in Hong Kong! So, here’s the link to block such a redirect: https://www.google.com/ncr