Harold

@harolddkim

Rebuilding my life at 30! Prev — Msft for Startups, AWS Startups, 3+ Startups

Los Angeles, CA

Joined September 2018

129 Following

82 Followers

389 Posts

Harold

@harolddkim

about 2 months ago

@WeWorkJP

Harold

@harolddkim

about 2 months ago

Will forever cherish my time working with YC founders both at AWS and Azure

britton

@Winterrose

about 2 months ago

one reason i enjoyed doing solution architecture with yc founders is that most of the time the interface for delivering information to them could just be verbal. it could just be a conversation. i didn't have to sit there and do some super detailed email follow-up with them afterwards to explain all the recommendations that were made. the email that got sent out afterwards generally was just a conversation and i would just drop links to all the documentation for all the recommendations. I would say "here's the links to the documentation I shared in chat during the meeting" + "hey let me know if you need anything else" and generally, they didn't because for the most part yc founders are exceptionally smart and have solid memories. i think this was possible partially because it matched the pattern they experience in yc office hours, this process where they're going and talking to partners and generally walking out and executing against the most important immediate action. it's also just partially just my communication preference. (ngl I’m a yapper. I’d rather take a phone call than an email, i have never got to the point of enjoying email) i have a lot of colleagues who really loved whiteboarding. i think for a lot of people, whiteboarding is more of a way to get all the context out of your mind and hold it and look at it. i never really needed to whiteboard very often, except on very large backend systems. i enjoy whiteboarding very very large backend system problems. its fun. whiteboarding with customers and deep dive cost-calculator excel spreadsheets, yeah! those were so much fun to build. "let's figure out how much this is going to cost." "let's build a financial model and figure out how much this is going to cost to run at all sorts of scale before we ever run it and build it." because you could cost model it a lot faster than a software engineer can build it and also you know then there are also scaling based rate limits considerations and trade offs - those things take a little bit more calculative exploration. these models rarely needed at early stage though. occasionally i would prepare and send an email just because i knew a specific founder would need it but the reality was if i was sending an email to every single one of them recapping what we talked about in the conversation with all the same links in the conversation - that meant there was another customer i wouldn't have time to meet with. another founder that i wouldn't have the chance to help because our team was small and we basically at capacity. when I started on the AWS YC team the YC Summer 2019 cohort had just wrapped up I was the third SA hired to the team to work with John & Jordan. there were only 8 or 9 SAs that covered all the startups that AWS interacted with at the time, including the 3 of us, so there wasn’t a lot of bandwidth to spare. that Summer cohort had 176 startups. W19 had 229 startups (a 30% jump) which John and Jordan mostly handled between the two of them as I was still onboarding. John moved to another team, Harold joined as his backfill, Leda (my daughter) was born and spent 45 days in the NICU, so Jordan hero carried us through S20 which, lucky for us, dipped to 208 startups… but that winter the craziness began to accelerate. YC W21 had 336 startups (+62% cohort/cohort), S21 had 391 (+16% c/c), and W22 peaked at 398 (+2% c/c). our team hovered around 2-3 SAs covering it full-time as people joined the team or left for other opportunities. our coverage ratio was between ~130 startups to 1 SA during that peak. around 80% of YC companies chose to use AWS back then so there was a lot of work to be done. so we just adapted. my approach - get all the information out, make the meeting the deliverable. the reason i did it that way was because generally the thing that helped the founders most was to have the information they needed right that very second and to not have to wait for me to go away and come back with an answer after the meeting. founders hate being blocked and they hate to wait. the coolest and best SAs / sales engineers / forward deployed engineers, they can just fix it right there; they’re allowed to write code that ships in the customers environments (SAs at AWS were forbidden from doing this). they just know exactly how to fix it. and if they don't, it's because it's a hard problem. and generally what it means is they need to go talk to somebody else who does know. not writing code for the customer was actually a feature, not a bug for us though. our customers were generally exceptionally talented software engineers, many with years or even decades of experience. they were faster than us at building, they just needed help traversing a very steep learning curve and avoiding mistakes they’d never considered or encountered before what else made this possible? something else i was good at was figuring out and mapping out the customers architecture and business in my mind during the conversation and holding that map in my mind. i’d try and explore/understand the founders / engineers strengths of who was really talented at what and identify where there was need for more information. without that context i’d lack the judgement to apply email sparingly. knowing my strengths was important as well. AI/ML was my focus area, but i knew exactly who i'd go to for crazy database questions, for ai questions outside my domain, or for weird and obscure networking and security questions. that, and I knew who would respond fastest to any given problem. this context allowed me to efficently close the loop when i needed to get back to a customer. there’s a lot of really great people working in these cloud companies. a lot of my good friends. a lot of knowledge that founders can tap into. (shoutout to awesome SAs that worked on YC with me. Jordan, John, Harold, Kevin, Andy and all the ones who stepped in to help during the high demand weeks) overall, the most important thing always seemed to be always seemed to be: get the founder of the information as fast as humanly possible. fastest generally meant right there in the meeting, in real time, moving to synchronous email only when needed, and it worked! I had an incredible csat score, an immense amount of good feedback. founders kept asking for more meetings on new topics & challenges. if a founder came back and needed to ask more questions on the same topic, i saw that as though I hadn’t done enough, so I watched for that. founders generally would only come back and ask me questions when they had already started growing and had new problems. but there's almost never like, you know, they always, i always tell them like, if you have any more questions, you can send me a follow up email and we can, and if it's something i can answer via email, i will. if it's not something i can answer via email, you can hop on another call. i'm also just extraordinarily slow at typing. i hate writing emails. so you can know that if you ever see a tweet / post this long from me, there's a good chance i voice dictated it. i’ll always be greatful for those four years of learning and growing with and serving some of the greatest founders on earth. there’s no way i would be anywhere near the angel investor i am today if it wasn’t for ~ 500+ cumulative hours spent doing technical deep diving with them and the thousands of hours of research and learning it took to keep up.

Winterrose's tweet photo. one reason i enjoyed doing solution architecture with yc founders is that most of the time the interface for delivering information to them could just be verbal. it could just be a conversation. i didn't have to sit there and do some super detailed email follow-up with them afterwards to explain all the recommendations that were made.

the email that got sent out afterwards generally was just a conversation and i would just drop links to all the documentation for all the recommendations. I would say "here's the links to the documentation I shared in chat during the meeting" + "hey let me know if you need anything else" and generally, they didn't because for the most part yc founders are exceptionally smart and have solid memories.

i think this was possible partially because it matched the pattern they experience in yc office hours, this process where they're going and talking to partners and generally walking out and executing against the most important immediate action. it's also just partially just my communication preference.

(ngl I’m a yapper. I’d rather take a phone call than an email, i have never got to the point of enjoying email)

i have a lot of colleagues who really loved whiteboarding. i think for a lot of people, whiteboarding is more of a way to get all the context out of your mind and hold it and look at it. i never really needed to whiteboard very often, except on very large backend systems. i enjoy whiteboarding very very large backend system problems. its fun.

whiteboarding with customers and deep dive cost-calculator excel spreadsheets, yeah! those were so much fun to build. "let's figure out how much this is going to cost." "let's build a financial model and figure out how much this is going to cost to run at all sorts of scale before we ever run it and build it." because you could cost model it a lot faster than a software engineer can build it and also you know then there are also scaling based rate limits considerations and trade offs - those things take a little bit more calculative exploration. these models rarely needed at early stage though.

occasionally i would prepare and send an email just because i knew a specific founder would need it but the reality was if i was sending an email to every single one of them recapping what we talked about in the conversation with all the same links in the conversation - that meant there was another customer i wouldn't have time to meet with. another founder that i wouldn't have the chance to help because our team was small and we basically at capacity.

when I started on the AWS YC team the YC Summer 2019 cohort had just wrapped up I was the third SA hired to the team to work with John & Jordan. there were only 8 or 9 SAs that covered all the startups that AWS interacted with at the time, including the 3 of us, so there wasn’t a lot of bandwidth to spare. that Summer cohort had 176 startups. W19 had 229 startups (a 30% jump) which John and Jordan mostly handled between the two of them as I was still onboarding. John moved to another team, Harold joined as his backfill, Leda (my daughter) was born and spent 45 days in the NICU, so Jordan hero carried us through S20 which, lucky for us, dipped to 208 startups… but that winter the craziness began to accelerate.

YC W21 had 336 startups (+62% cohort/cohort), S21 had 391 (+16% c/c), and W22 peaked at 398 (+2% c/c). our team hovered around 2-3 SAs covering it full-time as people joined the team or left for other opportunities. our coverage ratio was between ~130 startups to 1 SA during that peak.

around 80% of YC companies chose to use AWS back then so there was a lot of work to be done.
so we just adapted. my approach - get all the information out, make the meeting the deliverable. the reason i did it that way was because generally the thing that helped the founders most was to have the information they needed right that very second and to not have to wait for me to go away and come back with an answer after the meeting.

founders hate being blocked and they hate to wait. the coolest and best SAs / sales engineers / forward deployed engineers, they can just fix it right there; they’re allowed to write code that ships in the customers environments (SAs at AWS were forbidden from doing this). they just know exactly how to fix it. and if they don't, it's because it's a hard problem. and generally what it means is they need to go talk to somebody else who does know.

not writing code for the customer was actually a feature, not a bug for us though. our customers were generally exceptionally talented software engineers, many with years or even decades of experience. they were faster than us at building, they just needed help traversing a very steep learning curve and avoiding mistakes they’d never considered or encountered before

what else made this possible? something else i was good at was figuring out and mapping out the customers architecture and business in my mind during the conversation and holding that map in my mind. i’d try and explore/understand the founders / engineers strengths of who was really talented at what and identify where there was need for more information. without that context i’d lack the judgement to apply email sparingly.

knowing my strengths was important as well. AI/ML was my focus area, but i knew exactly who i'd go to for crazy database questions, for ai questions outside my domain, or for weird and obscure networking and security questions. that, and I knew who would respond fastest to any given problem. this context allowed me to efficently close the loop when i needed to get back to a customer.

there’s a lot of really great people working in these cloud companies. a lot of my good friends. a lot of knowledge that founders can tap into.
(shoutout to awesome SAs that worked on YC with me. Jordan, John, Harold, Kevin, Andy and all the ones who stepped in to help during the high demand weeks)

overall, the most important thing always seemed to be always seemed to be: get the founder of the information as fast as humanly possible. fastest generally meant right there in the meeting, in real time, moving to synchronous email only when needed, and it worked! I had an incredible csat score, an immense amount of good feedback. founders kept asking for more meetings on new topics & challenges. if a founder came back and needed to ask more questions on the same topic, i saw that as though I hadn’t done enough, so I watched for that.

founders generally would only come back and ask me questions when they had already started growing and had new problems. but there's almost never like, you know, they always, i always tell them like, if you have any more questions, you can send me a follow up email and we can, and if it's something i can answer via email, i will. if it's not something i can answer via email, you can hop on another call. i'm also just extraordinarily slow at typing. i hate writing emails. so you can know that if you ever see a tweet / post this long from me, there's a good chance i voice dictated it.

i’ll always be greatful for those four years of learning and growing with and serving some of the greatest founders on earth. there’s no way i would be anywhere near the angel investor i am today if it wasn’t for ~ 500+ cumulative hours spent doing technical deep diving with them and the thousands of hours of research and learning it took to keep up.

542

Harold

@harolddkim

about 2 months ago

@Winterrose I must say it’s an honor to be mentioned by name in your post 🥺🫡

Harold

@harolddkim

2 months ago

Looking to connect with people in Tokyo 🇯🇵

246

Who to follow

Number #1 Taiwan Supporter. Falun Dafa co-founder

Mark Lynch

@marklynchdev

Node / React / React Native dev and Typescript enthusiast.

Harold

@harolddkim

5 months ago

@Winterrose holy shit.. the weight loss in insane

Harold

@harolddkim

6 months ago

We launched Anymorph last month. There are 100s of SaaS tools out there to help you appear better on AI Search. They show you the visibility data, which prompts you show up in, where you get cited. They provide recommendations on how to improve your visibility score. But that's where most of the tools stop. We take it a step further and help you with the entire IMPLEMENTATION workflow. Anymorph generates fully functional on-brand pages for your website that are SEO/GEO optimized. No more waiting on content writers, designers, or web folks. Just click Generate -> Publish and watch your AI Visibility score go up. This is a short demo we showed to @HubSpot last month. Enjoy!!! https://t.co/JqvSvHcAvt