deepGem

It's quite funny how Gwynne Shotwell is able to steer SpaceX despite Musk while this new Twitter CEO is borderline toothless. There is just no way she would have allowed this to play out, given that advertising and branding is her core competency. So am guessing she was overruled.

There is a clear distinction between edge and a region. That distinction is defined by the distance and by relation latency. Edge locations are not heavy regions or data centres, per my understanding and hence limited by compute/storage capability. They are also high in number. Cloudflare has 100 odd edge locations all around the world.

Cloudflare and Akamai both have the concept of edge workers - light weight application APIs that can serve modest capacity. Cloudflare also has a cache on the edge.

https://blog.cloudflare.com/introducing-cloudflare-workers/ https://developers.cloudflare.com/workers/learning/how-the-c...

I think even talking about the draft gives you an idea about the severity of the problem. My experience has been the same. If the problem is severe enough a crappy draft will do. Conversely if someone is asking for multiple refinements even to arrive at basic decisions, I begin to write off the problem as not critical.

Pretty much all my interviews except in Meta. This is in India though. Meta was the most organized but even that interview went over 2 weeks. My Walmart interview lasted 2+ months. Round1 week1, round 2 week 3 etc etc.

This is for Staff+ level positions.

EDIT: Just to state, companies do give enough time to prepare for the interview. I am only talking about the interviewing duration, not the prep time.

I often think it's a lot better to do an overall interview over a full day than to do disjointed rounds of 4-5 interviews over a few months.

Some advantages - Interviewer and the interviewee are at ease. There is no rush to solve a problem. - You can easily spend 90 minutes to 2 hours on System design, Spend 2-3 hours coding and another 2 hours in behaviour/leadership what not. - The interviews can be progressive, meaning you don't make it through the first 2 hours - good bye. - This can be done remotely as well as in person. Of course, in person would be better, hosting expenses etc.

End of the day, decision is made and you can verbally convey an offer/reject.

This calls for a lot of discipline and commitment from the companies and their interview panel, I mean so be it. Dedicate 1 month for hiring and be done, at least for senior positions. Just like you allocate time for your projects, allocate dedicated time for interviewing every 6 months, every quarter whatever.

My younger self was loathe to programming. It was just a job for me, clock in clock out as you say. I was always in this zone of product managers and managers are superior as they were always in some meetings and it gave me the impression that important work gets done in meetings. I hated being a lowly engineer right from the start.

Only when I got into those meeting rooms, did I have a rude awakening. I realized the value of shipping software. I rolled up my sleeves and started getting back into building software, reluctantly initially but then I fell in love with the craft eventually.

Today if I find anyone who is getting disillusioned as I was, I try to educate them and help them find the joy in shipping software. Am not sure if it has really helped anyone, but I think it's the right thing to do.

Anyone who had to deal with retraining models understands the pain of retraining. What prompts offer as a feature while at the surface may seem superficial, they are groundbreaking, as they completely remove re-training and the entire ML-Ops deployment lifecycle.

Yeah, you have a much better understanding of this topic than me for sure :)

Nothing proprietary in Spanner but I believe no other vendor has an atomic clock similar to what Google has. Hence they are not able to implement the paxos global transaction lock which requires that all servers participating in the transaction be perfectly synchronized. This is what one of the comments above refers to I think.

So while the spanner paper is open to be implemented by any vendor, they don't have the proprietary advantage that Google has - the atomic clock. So Yugabyte, CockroachDB don't rely on atomic clocks. I tried to get to the ground level basics of this, but I haven't understood this matter completely yet.

The atomic clock is not in the VM, it’s a service from what I understand similar to NTP but obviously much more localised.

I in fact think once you offer atomic clock as a service for any public cloud, distributed transactions become a lot easier to implement thereby disrupting Spanner.

The atomic clock is the key to enable distributed transactions and Google has a proprietary lock on their atomic clocks for Spanner.

One of the best decisions in my life was to take this course. I started riding a Buell after this and boy what a thrill it was to lean this bike in the Pacific Northwest hilly roads.

The counterintuitive turn is more pronounced in motorcycles. You are literally pushing against the turn as you lean into the turn, and to lean in you have to turn out. Even more pronounced in sport motorcycles. Maintaining the centre of gravity during a lean at those speeds is a lot of fun to master. It’s one of those heart racing moments, literally no room for error.

+1. As an Indian kid growing up in a very conservative family/neighborhood I can attest that what I have learned and have helped me in becoming successful was all through my peers and books (mostly fiction). If I had no access to those and had purely followed what my parents prescribed, I would have been a very different individual and most likely, would not have pursued computer science. More than anything, succees or otherwise, peer learning and learning on my own was and still continues to be the emblem of freedom.

There surely is a downside to this, that this self learning can veer you off in a sub optimal direction of laziness, procrastination, lost concentration etc, as experienced in today's social media. However, it's better to bet the odds on freedom of individuals than on oversight.

People in general who are not good at something or didn't find success in that field, in order to feel their salt's worth have 2 options - either build something that's far greater than what they are not good at, or disparage what they you are not good at. The latter is a lot easier.

You are right. There is no legal contract, it's an insurance policy. In practice this is as good as having a legal contract is what I meant.

You have a legal contract with FDIC which insures all deposits upto $250K is what I know. If the bank goes bust FDIC owes you your money under 250K.

The consistent response I have gotten from managers is that more or less - people fake work sitting in their homes or work multiple jobs, and we collaborate better when we are sitting in an office. They use this as a force function to pull employees back into the office.

I call BS on all of these. Employees faking work - your hiring problem. Employees working multiple jobs - again your problem, keep them occupied with meaningful work, if not bring in legality to protect your IP or whatever. AFAIK no one working multiple jobs are revealing IPs.

In person collaboration - There may be some truth to this. High profile decisions are best taken in person, that being said during covid almost all decisions were done on Zoom not just for a week or so but for 2 years. Isn't that proof enough that even government level decisions can be done on Zoom ?

Design discussions are best done on Miro/Google docs than on the whiteboard. Multiple people can collaborate efficiently on Miro, especially with an iPad. Not only this, every voice is heard instead of just 2-3 folks standing at a whiteboard and driving discussions.

So many tools that make even pair programming or production debugging a breeze.

I can pretty much negate every argument that mandates in person office presence.

What I really vouch for is in person collaboration events once or twice a month where everyone comes together for a couple of days and hash out key issues and have a totally unstructured time for collaboration.

And of course, have an office for those who want to come in to collaborate. Not everyone has a spacious home, people like to hang out with co-workers during lunch etc. However make it mandatory that all group meetings will be done on zoom, in order to provide equal participation opportunity to remote workers. These hybrid meetings with a few people in a room and a few on zoom are a total disaster.

Sadly neither of these are allowed to sell in India. It's an Amazon/Flipkart duopoly and I personally cannot stand the Flipkart shopping interface. Millions love it.

I think Amazon has an ad network of sorts, so sellers bid for the sponsored search slot. In retailers who don't have such a platform, merchandisers take the call and boost/bury products in search results.

Ads on Amazon are far more annoying than ads on other web properties, because I go to Amazon for a very specific need. I search for Tide washing liquid and I see these sponsored products on top, way above the Tide result. From a purist viewpoint, this is disgusting. I have to come to you Amazon, to just give me what I want, not what you think I want.

That said, there isn't a better alternative so I just ignore the ads and go for the product no matter how far down the page that product appears.

Yet unemployment is at the lowest levels. So the economy is doing just fine.

I have finished their build your own Redis exercise. It is very well organized and there are hints/discord groups in which you get sufficient help if you make the effort.

This model is great for someone who loses patience with all the groundwork setup.

However, I do agree that the next leg is taking this MVP/Prototype level to production and ideally sell it as a real alternative to the commercial version of Redis.

I have been exercising in some way shape or form for the last 10-15 years. What I have found to be immensely beneficial is getting into a sport that you can play almost all through your life. Most sports have spurts of high intensity so that is super beneficial for dopamine rush. Then there is the psychological confidence boost that arises from just getting things right.

I started tennis lessons last week and honestly it's one of the best lifestyle decisions I have ever made. There's a lot of counterintuitive physics at play in Tennis and so it appeals to the somewhat geekier side of me. Lessons really help because the coach gives some meaningful tips totally unrelated to the sport - like having very high mental confidence when you enter a shot. This proved to be life changing for me. Like most HN readers, I do suffer from the imposter syndrome and generally it's hard to be very high on energy and self confidence (despite what one may call a very blessed life) and such tips really help in boosting your morale and confidence.

Tennis or a sport like Tennis is helpful in tiding over failures, because it's really hard to get a shot right consistently and when you do it, it's an immense morale boost coupled by the dopamine rush from the physical exertion. For someone like me, who keeps attempting at iterating on ideas and who easily gives up, such confidence boosters are proving to be super helpful.

Something similar is applicable to competitive swimming also. Lots of counter intuitive physics at play. The human body is not designed for survival in/on water by design. So you have to fight the basic design in order to move efficiently in water. So this is also very frustrating until you get it right and when you get it right, boy the feeling of elation is unparalleled. Then you continue upping your game - increase speed, distance whatever.

Playing such a sport and especially taking lessons have been more helpful to me than lifting weights in the gym. I do enjoy lifting weights and I do hit the gym because now it's a necessity to play tennis, but on a comparative scale, playing tennis is far more gratifying.

So Indians are on the top of the emigration food chain and I wonder why Japan is not one of our top destinations.

There are like 50K Indians in Japan, mostly Tokyo. This is a surprisingly low number. Sure there is the language barrier but that barrier holds true for Germany also which has a lot more Indians.

My hunch is education - Japan does not have too many universities to attract young talent, as compared to Germany, Australia or the UK. I'm not sure if this is by design or if there are some other factors at play.

If Japan built a lot more universities and made it easier and relatively affordable for foreign students to sign up with English classes etc, perhaps there will be a lot more immigrants from countries such as India, Vietnam, Cambodia etc.

Or carry the 60W charger and turbocharge the phone, agreed it's not as portable as the brick.

This is true. In the current scheme of things, you don't need to know any math to get things going. At some level, if you really need to write a custom model from scratch, probably you need to - but not really.

Now you have the likes of HuggingFace which is another abstraction on top of Pytorch. Heck you don't even need to code a transformer from scratch. They have done it for you.

from transformers import AutoTokenizer, AutoModelForMaskedLM tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased") model = AutoModelForMaskedLM.from_pretrained("bert-base-uncased")

I was like wow, this level of abstraction is totally breaking barriers to entry and we will see many many flavours of huggingface in the years to come.

Most of the architecture diagrams that I review and sometimes write are of the second kind. Once you have planned and set for execution, the architecture rarely changes. If it did, then there was something fundamentally wrong with the first architecture and as s result this diagram will be modified in a review or a rewrite or whatever.

A live architecture document derived from code, even on a batch basis if not in real time, will be a really cool endeavour. If you can somehow look at a repo and generate a diagram describing all the different subsystems and their interactions, boy imagine that !

A simple traceability might be just through imports starting with the main program. Start with the main functions and trace them through. May be using a tool like sourcegraph.

My apologies. I didn't mean you to him, i meant it as a generic response to everyone. I will reword the reply to say one.

"and don't let the failures define you. There's always a chance for success, no matter how small it may seem."

These small successes is what one needs, to propel them back into the entrepreneurship journey. I don't know about these influencers and not sure what they sell. You just have to get lucky. I have been through so many startup failures that I have stopped counting, but I attribute my failures to my own laziness and not being in the right market.

So back to successes, what I found super helpful is a micro dosing of success like winning a point in a game of tennis, or getting 10 consecutive shots above the net , something like this, totally unrelated to the world of startups can bring in tremendous confidence and morale boost. Why I say tennis is because getting some of these shots right consistently involves some luck in the beginning. While you may not get all the 10 shots correct, that one lucky shot can bring in a high dose of morale boost. It could be shooting balls as well. I personally find that small successful acts like this, which involve your skill and some luck really help in overcoming failures, failures like what you state, which are truly a result of not just your effort but a lot of outside factors/luck.

EDIT - Changed you to one. I didn't imply this message directly to OP, was more of a generic reply.