HNHacker News
TopNewBestAskShowJobs

deepGem

2,284 karma · joined December 30, 2010

Generalist software guy, loves investing and writing code. Multiple battle scars in startups.

Email: pradeepbs at gmail dot com

submissionscomments
deepGem··on Tesla Names Robyn Denholm as Chairman to Replace Elon Musk
I really liked how Musk handled himself. Combative and stern at times, especially when Kara Swisher said she doesn’t own an electric car :). Overall not a great interview. She kept harping on Model 3 production. I hate it when interviewers try to dissect the past. It’s more exciting to talk about the future, get Musk thinking.. but no let’s just dig about the past.
deepGem··on Neo4j nabs $80M Series E as graph database tech flourishes
Thanks Emil. I will definitely get back to you on my findings.
deepGem··on Neo4j nabs $80M Series E as graph database tech flourishes
Wow, we had used the same stack. Neo4j - gave up because of horizontal scaling issues, moved to TitanDB and then our startup folded.

Fast forward 3 years and am back in the graph world, this time dealing with money laundering and fraud rings. It stumps me as to why graphs are not used more in the financial world. Financial transactions are no different from social interactions. So concepts such as community detection etc apply to the financial realm as well and in fact we are using such concepts to determine fraud rings. I may get back to neo4j.

deepGem··on Tesla Faces Criminal Probe Over Whether It Misstated Production Figures
True and I don’t claim that musk is right always. He is human, makes mistakes, that doesn’t mean he is a criminal, which is what the shorts are implying with these smear campaigns.
deepGem··on Tesla Faces Criminal Probe Over Whether It Misstated Production Figures
It may not be the retail investors causing the swings but the algorithms trained to gauge sentiment from news sources and trade based on that sentiment. Wild guess.
deepGem··on Tesla Faces Criminal Probe Over Whether It Misstated Production Figures
Personal view again. I aM convinced that the shorts are behind these media smear campaigns. Scumbags are guaranteed to lose their bets and they are doing their best to minimise damage. They don’t care abou innovation or climate change. All thy care is about their fucking money. Sorry about the rant and yes I am long TSLA. I have had enough illogical Twitter spats with some noted shorts to come to this conclusion.
deepGem··on Tesla Third Quarter 2018 Update [pdf]
Until other companies step up and install DCFC stations everywhere

Disclaimer: I am long TSLA

I hightly doubt other car companies are gonna do this in a reaonable time frame (3-4 years). They will rely on the dealer network to build out the chargers and suck up the setup cost. Now dealers can potentially finance this with debt and get this done, but I doubt GM or any car company will be able to get consensus from the dealers and roll this ouut fast.

deepGem··on Paul Buchheit on Joining Google, How to Become a Great Engineer, and Happiness
Back in college around 1996 or so, the only Unix workstations we had were shared DEC Alpha machines. Nobody in our university knew anything about Unix. It was basically a Win95 shop. The only way to edit a file on DEC Alpha was using vi or that's what we were told. I still remember struggling for the entire exam. I think you passed if you were just able to type out the program lol. 3 years later, once I graduated and started working all of the unix guys were using vi. I had not even heard of emacs. So naturally I started using vi. 2002 some of my friends started using emacs but I was too embedded in vi to learn anything new so just didn't bother.
deepGem··on The lost art of concentration: being distracted in a digital world
But it's definitely different then when I was younger and would just tear books up.

When I was 19, I finished John Grisham's The Partner in one sitting - without even getting up for food. Something like 24-30 hours non stop reading. There's no way in hell I'll be able to do that now.

deepGem··on The lost art of concentration: being distracted in a digital world
but this is a comfortable lifestyle and I just keep on doing it. That does sound like an addiction, right?

I have similar traits that I am actively working to wear off. My only addictions are Twitter, Hacker News and Yahoo Finance. I spend at least 4 hours in a day flipping across these. I must admit Twitter is quite helpful because my feed is well curated. YF is a total waste of time - I don't know why I keep checking the stock prices when I know for a fact I won't sell anything. It satisfies some part of my brain to keep watching the tickers. What a fucking waste. Hacker News is useful too. What has helped me in the beginnings of my de-addiction is measuring how much time I am wasting and writing it down. It's kind of a slap in the face to get up in the morning and see that you wasted 4 hours the previous day.

I can't even talk to other people about it because everybody's noses are in their phones everywhere. Is that weird?

Perhaps hanging out at an old fashioned bar might not be bad. Not the ones with fancy music but the ones that remain somewhat rustic, where you can just have a beer and a random conversation.

deepGem··on Ask HN: My country is getting worse. Where to go?
I don't know much but a rudimentary Google led me to this college of Rockies. There might be others like this. GOod luck !

http://www.cotr.bc.ca/programs/

deepGem··on Ask HN: My country is getting worse. Where to go?
I think you can move to Canada to purse a course in a community college. Not a degree but a 2 year course. Can you get a loan from your home contry ? If so you can move to Canada spend 2 years studying/earning and you will be set. You will have to take a loan to cover your airfare and first semester college fees. Worth exploring I suppose.
deepGem··on Rapids – Open GPU Data Science
I wonder how much of this is driven by market need. Pyarrow + Pandas is significantly fast already.

http://wesmckinney.com/blog/high-perf-arrow-to-pandas/

Also Pandas 2.0 is going to roll in a lot more utulities for parallel computing. Is there really a need for 50-100x speedups today ?

deepGem··on A fraudster got $12M out of a Canadian university
This seems quite simple to accomplish - compare the name and email address in the email header and flag based on even regex matches/mismatches. Perhaps a Chrome plugin for gmail to test out ?
deepGem··on A fraudster got $12M out of a Canadian university
Every major fraud vendor has an AML offering. Be it SAS or FICO or startups such as Feedzai. I am building a startup as well but it is hard to determine how effective these products are. Till date I haven't found data that shows how many money laundering cases were caught by these products and how many of those cases were prosecuted. It's not a very transparent industry, so it's super hard to take on the incumbents. This is just based on a few weeks of conversations I've had with some AML and strategy team members, so by no means this is a generalization.
deepGem··on A fraudster got $12M out of a Canadian university
https://globalnews.ca/news/4149818/vancouver-cautionary-tale...

This is really disturbing.

deepGem··on IQ Matters Less Than You Think
+1. The method of getting from good to best also matters a lot. A coach definitely helps. You can't become the best of anything, in isolation. Deliberate practice also helps. I was not a believer in deliberate practice but once I started it, I am seeing good results. For instance, I can retain at least 30% more of what I have studied or coded or even just casually read. However, deliberate practice takes a toll. It's excruciating to do it even for 4 hours a day. Maintaining consistency is even harder. I have probably been able to do it for a week over the last couple of months.
deepGem··on Facebook launches PyTorch 1.0
It's amazing to watch Pytorch and TF slowly converging and offering more or less the same capabilities. TF started in graph mode and are now adding eager mode. Pytorch started with eager mode (prototyping) and are now moving to graph mode (production). TF evolved from Google's production environment so it comes as no surprise that it was geared for production first deployments. Pytorch evolved from researchers which explains how it was tailor made for tinkering and prototyping. Good to see the two converging. I am guessing a year from now there won't be any marked difference between the two.
deepGem··on Elon Musk Settles SEC Fraud Charges
I think Musk technically pays the full 40M through some stock purchase.
deepGem··on The Fed stalls the creation of a bank with a novel business model
Holds true only if you convert your currency to INR. The dollar denominated deposits (FCNR) have far lower yield around 3.4% still better than what the US banks offer.

https://www.icicibank.com/nri-banking/RHStemp/rates.page

But I think you need to hold an Indian passport to open these deposits. I am not sure though.

deepGem··on Deep learning experiments in OCaml
Yeah I should have been more explicit that this comment is in reference to ML code. Training is nothing but a loop so its unlikely you pass type A in iteration 1 and type B in iteration 200. If that happens most likely your training data is messed up and type safety cannot help you as you would have compiled and tested for type A.
deepGem··on Deep learning experiments in OCaml
Yep agreed. Though you can communicate design intentions in comments, no ? There is another argument that programmers might not follow the comments, so strict enforcement by types helps - I don't believe in such a philosophy though. Most programmers will do the right thing and mistakes are not intentional.
deepGem··on Deep learning experiments in OCaml
The type safety argument is total BS. First of all the training script will fail for the very first time if there is a type error. You'd be a moron to pass an argument of a different type 'a couple of hours' into the training. No sane programmer writes such code. What kind of nonsensical argument is this.

What I have found static typing to be really useful for is in remembering what I have coded. It's quite hard to remember a dynamic type while you are writing code, given the number of variables you are dealing with. Seeing that type definition next to your variable name is a handy reference. I find it helpful to speed up coding a bit and being able to remember a lot more clearly what I have done.

deepGem··on GoPro Hero 7
I've found the keyboard shortcut to be a lot less annoying Esc. I just do it automatically these days on any popup. These damn website notifications aren't going away anywhere, they are getting all the more pervasive. So I've just learned to live with them and be less annoyed.
deepGem··on Ask HN: Solutions for collaborative code review, pair programming?
You had me at the shared debug session. I am amazed at how VSCode has evolved in such a short span of time. Technically it's not an IDE but it has most of the required features of the IDE.

The liveshare feature does require you to be signed in to Github or Microsoft. So the sharing might not work with other proprietary SC or even Bitbucket or Gitlab.

deepGem··on Entrepreneur raises millions of bugs to tackle food waste problem (2017)
Looks like the protein can be consumed by humans as well, since many people eat roaches. Can't the protein powder be used like whey isolate powder? If yes, that's a massive industry to tap.
deepGem··on All Things Sales: Mini-lessons for startup founders
Loved the 6th video - How, when and who of sales. I was pleasantly surprised how the generic advice to find a corporate sales guy to lead your early stage startup sales is so ineffective.

A similar analogy to engineering also applies I guess. One who functions well in a corporate engineering department - with a ton of support might not do well in an early stage startup and vice versa.

deepGem··on Modi’s Cash Crackdown Failed, Indian Bank Data Shows
I absolutely detest government intervention in any aspect of the society. In fact, I suffered personally like a billion others because of this crap that the government unleashed on us. I am just seeing the brighter side of things.

I can't comment on the convenience of digital wallets for all sections of the society. For the urban middle class I can say with conviction, that digital wallets trump cash. The biggest obstacle for digital wallet adoption were the mom and pop shops. Now even a juice vendor accepts PayTM. I don't think the wallets are as harsh as the credit card guys on the merchants. So I see the adoption as a win-win, at least for the urban middle class.

All of this doesn't negate the endless suffering ulneashed by the cash crackdown though.

deepGem··on Modi’s Cash Crackdown Failed, Indian Bank Data Shows
Perhaps, but on the bright side, I see an increase in digital transactions, particularly wallets. Almost every nook and corner shop in India now accepts payment by Paytm or an equivalent wallet. It'll be interesting to see the uptick in digital payments since the crackdown on cash.

I don't necessarily think wallets are a better way to pay but they are a hell lot convenient than carrying petty cash and having to hit the ATM every now and then.

deepGem··on How to Hire Your First Engineer
As engineer no 1 at an early stage startup, you don't need ths same skills that are needed at FAANG. I may be wrong, but you are constantly iterating and mostly building software that has a shelf life of may be 3-6 months. You are not building for reliability or scale. You are building mostly ad-hoc and to fulfil an immediate market need (which may or may not exist 2 months from now). Where as at FAANG you are by design building for reliability and scale from day 1. Very different skills.

I can't comment which skill is more valuable. They are somewhat orthogonal to each other right ? The better you get at software engineering the more process oriented you become. Hopefully an early startup gives you that kind of progression, but very unlikely, since the progression is directly correlated to the startup's scale and success.

If someone desires to put good software engineering into practice from day 1, then engineer #1 at an early stage startup is not the right choice.

This is all based on anecdotal experiences so am happy to stand corrected.

← PreviousPage 13 of 32Next →