Transcript
Speaker: I am not the guy. who will build a good to have product. It's not me. That's not me. if I were to build products for Um...
Speaker: toolsman repair guys i think i'll I, i will i'll build a screwdriver I'll design a new screwdriver. We've been through that before, which is when we started to segregate it again. So we'll segregate it again.
Speaker: Yes. We started there. Everyone is a software engineer. Now we're like, no, no, no. You know what? You're a security engineer. You're an incident response manager. Yeah. When it says simple,
Speaker: I think simple is so much harder than making it complicated. It is. It is the hardest thing to make something simple. It's hard. I don't think customers would Be the best.
Speaker: to tell you what the experience should be. And my my suggestion to this was. um Uh...
Speaker: Oh. lately i've also understood that i'm a very Obsessed about... Problem solving. Oh, that's a spike alert. sorry believe it or not that's a spike that's nice that's interesting i'm keeping it Sorry.
Speaker: Um...
Speaker: Alright, so we are back. at the root cause We have another guest today. And. Thank you for tuning in again. This is the fifth.
Speaker: episode of The Root Curse. And. I'm very excited the way Everybody is responding to it. and I hope The...
Speaker: expectation that I had to bring in people in front of you. to bring in the talks in front of you. is being fulfilled ah Do let me know if...
Speaker: I need to bring in somebody that you want. and I'll try to get them. on the show So... That being said, let's introduce our guest for the day.
Speaker: We have Gothic! Founder of SpikeSH. Gossip! Would you like to introduce yourself a bit? Oh. ah Yeah, sure. I have... ah I...
Speaker: I'm... A software developer turned. Founder and then now I'm still navigating and learning business fundamentals and building spike and ah understanding the markets and changing my narrative.
Speaker: off. how how everything works when you build a product because building a product seems so exciting What happens after you build a product? And there are so many elements that goes in it.
Speaker: and i'm slowly uncovering and discovering it I've been writing a lot about all of this, but generally I have been... ah I have learned a few things about myself. very product first person. very customer support service obsessed person and lately i've also understood that i'm a very Obsessed about...
Speaker: Problem solving. Oh, that's a spike alert sorry believe it or not that's a spike that's nice that's interesting i'm keeping it Sorry. um it is the dnd bypass alert so Um,
Speaker: And. Yeah, so... I have previously I was working with Wengify. um I'm bootstrapping Spike. Because... I worked at a very successfully bootstrap company at Bankify. A lot of the ethos and a lot of my thinking kind of arises from everything that I learned at Bankify. being exposed in that culture.
Speaker: And it's a personal choice about bootstrapping it. Right? And. i am also we we are a very small team but we've gotten we've got about um I've got roughly around... 215th nearly 267 engineering teams across the world.
Speaker: ah using spike engineering teams operations teams everyone different types of people different types of solutions that they provide to their customers. and you know one common solution that kind of binds them all.
Speaker: which is Spike. and very interesting journey. Yeah. Interesting. Interesting, Kashik. Thank you for taking out the time. and welcome to the root cause.
Speaker: Thank you. Thank you so much for having me here. Thank you. All right. We were talking Kaushik earlier And... you redesigned and shipped.
Speaker: Spike is rich. very recently. Right? and You have been arguing about... Just a single row of UI and and the thought process that goes behind it.
Speaker: What? What do you think? Why? Did you spend so much time in... working with the interfaces in Putting in the thought process and the effort.
Speaker: for Something white like... people might not even be using or interacting with they might be using it on a phone
Speaker: I think. Um, when when we built spike ah So it was just me. uh for the for the for for nearly nine months of building the product it was just me during covet and I... was not very well versed with, front end as much.
Speaker: putting it putting HTML CSS together great Amen. using React or anything was not really my cup of tea at the time. and i did not really wanted to spend time mastering all of that So I went ahead and just kept building it. but Would yours?
Speaker: ah all that design that has just like claimed. on top of each other. And. Oh. you know, Ooh. we underestimate the Kind of. usage or the way people use your product. We think that people will use it this way. and a lot of them do.
Speaker: Um... but also a huge number of customers because now you see variations when the number the moment you cross like 100 customers i think you start to see like this massive variation of people using they like 10 % of the people have spikes but spike open in browser all the time.
Speaker: Right. And that are like. I think Up until last year, when before we had to rebuild the entire mobile app and everything we did not even see a lot of people installing the mobile app as much.
Speaker: and know And we had to redo a bunch of things and bring more discoverability. At the time, our inference was that everyone... is interacting with spike across slack because that becomes it.
Speaker: Incident Command Center. and our Microsoft Teams. And they're also just interacting across. Phone Calls Let's play. and but that understanding change the moment we start.
Speaker: themes lot lot more And we now realize. so many customer requests and everything starts to come in And it would typically be. about the user experience.
Speaker: you know. And. it's it's it's It's a large reason why we changed it. and Ew. we were We are so obsessed. with this This time we've gotten like really obsessed about it.
Speaker: to a point that we we sat down and wrote all the problems customers started to face. and we wanted to bring a more keyboard centered design. But the fact is that The reason why it is the way it is, is because
Speaker: We've not come across anyone who's had to to any internal training session. inside their company. when they adopt Spike. So, that simplicity of incident response or incident management Simplicity is absolute key and that's our absolute focus and When it says simple,
Speaker: I think simple is so much harder than making it complicated. It is. It is the hardest thing to make something simple. It's hard. and And because we've debated about...
Speaker: do you really want to add responders on the table as an example right so and an incident Incident row is a synonym of your command center. And if you write down. you there are about nine actions or about 12 actions if And if I go really, really.
Speaker: big, there are 16 actions you can run on an incident. And can you embed all of them? We've we've seen how Big Panda would do it. Now. we've seen how 911 operators would do it and if you google the consoles can get as complicated if you want to present all of the action items.
Speaker: But we wanted to keep it simple. So those discussions and arguments have been very I want to say riveting You know? We've learned so much about ourselves.
Speaker: what kind of product we we we want the send look The decision making framework is simple. spike has to be simple for everybody And. the ah the the question always is Is it going to be simple or not? So,
Speaker: There are points of interaction. that we did not agree with. But it's there. Right? like the fact that you want to change the status of an incident but do you want to change the status of incident for like all 100 incidents as an example if there are all 100 incidents but then the question is who has 100 open incidents
Speaker: Right, but we have we've seen teams with a who use it properly as alerts, right? So it's it's just alerts. So there are 14 ,000 open alerts. None of them are resolved. And you you have to also... bring those elements into consideration because we've we've understood.
Speaker: There are there are like lots of small teams. and there are lots of medium -sized teams We use spoon. Incident Management. or Alert Management then they would be like if this is a real incident then we'll think about it but all of the alerts just kind of stuck. up.
Speaker: so that is that user and then there is a user who or Let me press... let me put it this way, there is a theme Who would have their entire open incidents clean at all times?
Speaker: from the get -like. one incident a month if if it's critical or not. And that's something they keep arguing amongst themselves. It's the textbook method of how you would do incidents. and That's why a lot of the decision making went in.
Speaker: ah especially as someone who is who has not been an SRE. I've never been an SRE until... too. Look at it from the outside. and to see what kind of teams behave differently. and what they need and how they've architected reliability inside their teams.
Speaker: and how Spike fits in. Mint. We still want to continue to keep it simple. so people can just instantly expand they can onboard like 20 people 10 people how much server and they know exactly how it works.
Speaker: Another 4 years you will see another few revisions i think that's that's never ending i think it should settle sometime The... The thing you talked about... in terms of ux is the main part if you have cracked it then Rest everything is aesthetics.
Speaker: So you can polish it as many times as you want. Yeah. and you keep rebranding their logos even if it looks good so it's just a fresh feel to the same thing the batman has not changed their logo 100 years they just keep updating the logo right to the modern things.
Speaker: you you've seen one of those like backpanel logos in 1930s and 50s and 70s and 80s and 90s there is but there is shadow in some there is no shadow there is more metallic color This black and yellow mixed.
Speaker: not mixed. uh so i think that's like part of that identity and everything just gotten really well done I think a lot of lot of times it comes to the brand recall as well.
Speaker: Um, Bye. I think, again, this this is a, like... question of aesthetics and expression. have from a company standpoint.
Speaker: how you want to express the product. and The UX is something which The Customer's Drive yeah The experience is what the customers drive.
Speaker: But do they? Don't they? It would be nice coming from you because you have been a... Customer focused. uh Man, and...
Speaker: the the whole product that you built is around customer experience Yeah. So... and And the way like I spoke to you. You you interact with a lot of customers.
Speaker: on day -to -day basis right so Don't you think? it The experience is something which should be driven by customers. Yeah. I don't think... customers would Be the best.
Speaker: to tell you what the experience should be. And my my suggestion to this was. Um... Um...
Speaker: I've - I've - I've... I've done? onboarding many many times. Right And you... ease Sometimes, whenever there's a new product or something, right? You just like sign up instantly, experience it on boarding.
Speaker: And there is an onboarding feature where you just land up on the dashboard. and then you tell them that this is where it is this is where it is set up your first escalation set up your integration everything and then there is that is a user experience.
Speaker: so the onboarding experience that we we've changed and this is like like Pretty sure this is my fourth or fifth iteration. At the time, I remember I had written four iterations. and this onboarding experience is super guided.
Speaker: Before you reach the dashboard Like sorry. When you reach the dashboard, you will get your first incident. And you will get a phone call from Spike.
Speaker: Is it like... uh The games that we play Super Mario, let's say it doesn't give you a tutorial how to play the game but the gameplay on the first uh Like Faze itself is that tutorial.
Speaker: Yeah. Yeah. Yeah, pretty much. You can say that. that's that's that's ah That's a really nice way to put it. You know? and You can do that. By the time you actually come to spikes dashboard ah Most of our customers most of our customers today and even leads everyone who signs up. they would have gotten A.
Speaker: Alert! that says this is a sample incident on your newly created integration and they can see the escalation. so at the time as very I understand that. no matter how many people I would have spoken to I would have never landed on this conclusion.
Speaker: because the metric was The value of your product, like how how long does it take? for you to get to deriving the value. And the problem was, I'll have to wait for an incident.
Speaker: it's it's That's a big problem. Now you are also waiting for them to integrate with their AWS or Grafana or whatever. right. But up until that doesn't happen. they're not gonna get The alert. And what's to say that they don't have other priorities?
Speaker: You want an incident. You want to view on the alerts. I've understood what you want. He won the alerts. So, we will simulate an incident. you want you want You want to manage incident. you want to You want to resolve an incident. You want to you want to see when things break.
Speaker: Here's a test simulation. Thanks. and And that's an experience we had to design. There are down from my team. has debated.
Speaker: endlessly because i i would have made the onboarding longer i made i would have made the onboarding shorter But we've not. That's another alert. We've not seen.
Speaker: I'll just acknowledge or escalate it.
Speaker: Okay. Is it serious alert? Uh... It's... I mean, we know the incident now.
Speaker: It's okay. For now, it's okay. I have acknowledged it. It will time out in 20 minutes. if it's not automatically resolved. Um, You have to design experiences a lot of the times.
Speaker: I don't think people Not everyone will know what they want. They will know only from experiences. Say for example the whole idea of This acknowledge time, but you acknowledge the incident and in 20 minutes if you don't resolve it, we've set it to 20 minutes if If you don't resolve that incident in 20 minutes, then it will go back to triggered and it will resume the escalation process. the scene.
Speaker: that Um... came as a user feedback because they had seen that in PagerDuty. Hey. so experience. Is it derived?
Speaker: um nature or an experience could have been derived by somewhere else but I don't really think that you could do and experience like no one at the time did.
Speaker: an incident simulation to a point that when you sign up and you instantly just start getting these sample alerts and there is that aha moment that we're always fighting for right So I don't know Customers can only give you feedback.
Speaker: on the touch points that you've decorated along the way. There is a difference in a car buying experience. between all the different car segments.
Speaker: And for that, you will have to go to the car showroom and understand it. it The car buying experience is designed by the seller. Right? And that bit is important.
Speaker: So I don't really think that we will need to wait on customer feedback every time. Customer feedbacks are important where? Customer feedback is important where we now... We have a team who say that Um,
Speaker: they they they won They won't. In on -call, they would like to have an on -call override experience where they don't I don't know, right? with a request for an override That's their culture.
Speaker: Hey, you can't just go in and say Priyank is gonna do it. I've discussed with Priyank You know, you are a remote workspace. We all work in different time zones. And I can just say that I'm not i probably going to be off my kid's birthday next week.
Speaker: Right? for like a day. And if someone can cover me for a day. Great. That's the feedback. they're not gonna be able to tell you if you if you need to be able to click this here or to tell you that It would be really nice if this this has to be on the mobile app as well.
Speaker: We had this We have out of office and spike. admins can add out of office for others and we had to Like... Plenty of admins give us that feedback. I know Zach is out of office, but...
Speaker: He's on call. We can add an override, but it would be nice to have an ability where I can add. out of office on his behalf. That is a feedback, but them coming and telling you that we need to have out of office, that's going to be hard. For that, you need really good customer relationships.
Speaker: And that's a segment that you need to invest. and that's an angle you need to come to a point that Is is that point coming off absolutely valid or not Makes sense.
Speaker: I get it. so Maybe the... use cases the a problem statements is something which a customer feedback What's best on?
Speaker: Um, Maybe on the UX and UI possibly That's something. Deep. Product should narrate that.
Speaker: um in a language that is more accessible maybe familiarity is something which would be suited for some product some sometimes it is a totally new concept then you and may have to because nowadays it's ai right ah the question that is very prevalent in ui nowadays is what is the best interface for an AI driven or agent tech Uh,
Speaker: UI, right? we had communication the conversation view we have chat right but this is something which is Just one aspect of it. What could be the next? iteration of UI this is the conversation being Do you have them? Point on that, do you have?
Speaker: An idea? I don't. I don't. All I know is we will go through many many iterations. until the world settles on something. And we will all have to be kind of okay with it.
Speaker: there are There are... just because it's conversational. It's not useful was also a point. I remember meeting an investor in melbourne and when we were having coffee he was one of the things that he was discussing two years ago is that i don't really think that conversation is the right platform for most Maybe, maybe not.
Speaker: You know? Um, but i i don't I don't know what is the right output. But it it it will also consistently keep varying. This is where Brian Chesky is right about.
Speaker: what is There. Travel. the best use of AI in travel is not going to be a chat interface. and hear strong opinions on that. I agree with that.
Speaker: but I. As I use superhuman so, so much. I think, and I use AI inside Superhuman. tremendously. Again, it's a different interface for a completely different product you know so each product will i think largely we'll have to figure out How do you want to place that? The user experience of it is...
Speaker: you do something in the air so put an output If you have to brainstorm with AI, Then I also have come to this conclusion in one sense. is that people don't want to have that discussion on your product i think people would just want to discuss with their ai on their ui and everything Because they can have history be it. on Everyone has to release MCPs.
Speaker: at the end the central operating system is either going to be clawed or Open AIR anyone. But everyone will have to release their MCPs and just have the entire world connected. at one place.
Speaker: So, it could be chat then. so to speak. So the end result could be chat, but maybe cloud will probably sit down and say it's not the end result Night. But.
Speaker: this is what i think i think everyone has to figure out what is the right way for example status pages I... Status Pages Osmotoms of an Incident the idea that you will have to Talk to Spike's AI to generate a postmortem from your recent incident.
Speaker: wouldn't really be the best usage. A really good experience in my opinion. Let not... not usage experience here is yes i can give you like an automatic rundown and a timeline and everything for your post model But if you want to make edits you can just have spike sense to be connected to cloud and go there and you can change However, Andrew X.
Speaker: Yeah, or Rubik's Cube, right? Exactly. I think that bit is also super super important. we will all have to under we're not we don't We don't live. in a product space where the average like daily active users.
Speaker: it's not really a metric we like to track right so it doesn't really matter if they're the what matters is they are able to solve the problem. And that's the experience that you will have to give.
Speaker: Anyone who's building a product, they'll have to obsessively think about if they're solving. The problem? more deeply. not does not really matter if they are solving the problem with AI or without AI.
Speaker: to solve the problem. And... everything will start to fit. in really really well. Nice. That's nice. Products are either must -have,
Speaker: or good to have. I'm a very firm believer of only build Good to have, must have products. I am not the guy. who will build a good to have product.
Speaker: It's not me. That's not me. if I were to build products for Um... toolsman repair guys i think i'll I,
Speaker: i will i'll build a screwdriver I'll design a new screwdriver. Um, There are too many companies doing screwdriver. that's where the whole analysis about okay what do you do after you've built the product how do you think about these things that part has come about much later, but it's it's the fun part.
Speaker: I get it, where you're coming from. is ah The good to have. Vitamins is something which doesn't inspire you much If you are actually making an impact by solving an actual problem.
Speaker: a very real problem that is a pain point. that's where your interest lies. Right? Yeah. That makes sense. Makes sense. You talked about... a must -have product versus a good to have product Do you think this AISRE category or what I like to call site reliability intelligence category Is a good to have or a must have?
Speaker: it's I think it's in the transition phase. It's in the transition phase because We're all... we're all getting more things done with AI. until Until... will come to a point that that becomes our basic expectation of any product.
Speaker: We're almost there? but that almost there is only for for those who mingle with any form of technology everyday. If you don't do a if if you wanted to a point that any engineer or a developer in even in vip pro has to be able to use AI left, right and center.
Speaker: That's it. the level of expectation that i'd say is that when they would say It would have been really nice. If. Yeah, I would have triaged this to a point that I would not have gotten a false alert at like 9 p .m during my dinner break as an example right That would be really nice. And that is the point when people will instantly be like, there has to be an AI which automatically triages these incidents.
Speaker: And that's when they start searching. And I think that that transition phase is happening right now, but across. In my opinion, not... not 12 months is still like an eternity in today's days. But I'd say like across a six month cycle, we'll see a huge shift.
Speaker: happening across where people are searching instantly for Why do I have to deal with these incidents? Why is there not an AI? It's just the expectation that is pending.
Speaker: Product. and Tech is already there. a lot of companies are actually Utilizing it very well um lot of products that i have seen are doing it very good um I try to be on those lines.
Speaker: um But adoption is something which still I feel SREs and operators are in a Dilemma. whether now is it that trustworthy okay i can put it in production systems are.
Speaker: Should it be allowed enough? Actions. Enough. privileges to take actions on my behalf. should i be making the judgment should ai be making so these are a lot of Questions that keep you know uh disturbing lot of operator operators and I think it will settle down after some time It has to be infectiously spread.
Speaker: the the contagious nature of We do it with AI and you are not. That. that has to be infectiously spread and that will That will.
Speaker: um we're not we're not we're not about three or four years ago where We will sit on question. EI? we are at a point where Yeah, I think if AI is saying it.
Speaker: And if it's doing it... i kind of believe it I kind of trust it. So we're at that phase, a huge part of it. is The problem is just with the trust, right? Because trust is, as I said in my previous... episodes as well Trust is something which has to be earned.
Speaker: You cannot enforce trust. on somebody But is that hard to earn? It is. hard. this the thing is kaushik if you see lot of stories and news about ai deleting database tables it becomes hard and difficult.
Speaker: because then you have a spectrum where AI is... you know wasting production and then you have a spectrum where ai is doing very well in production system yeah Sad.
Speaker: this This is where the challenge comes. But isn't that like... I... if If you're going to deploy something like AI in production, wouldn't you be really cautious about what kind of...
Speaker: harnesses and what how do you gatekeep it? what kind of harnesses what kind of what kind of like at least the gatekeeping how how What? Wouldn't you... give AI access the same way that you would give any Let's see.
Speaker: um intern or a junior yeah intern yeah I didn't want to say intern, but sure. intern because i don't want to production and then intern but sure right Wouldn't we be... doing the same thing like a perhaps a read -only or you know That's true. That's the exact approach. uh people do follow nowadays and uh So there is an education missing. of how to deploy this in production.
Speaker: there is education missing there is a lot of experimentation being done in terms of what is the correct guardrails and correct uh Like. the reins that we need to put around AI.
Speaker: First, So this is where once we finalize that then I think it will be much easier. but we haven't finalized nobody else It can't be finalized. It will just keep evolving.
Speaker: it yes it will it will just keep you on or the models would Keep getting better. Yeah. Even if they do, I think As long as the models can just follow instructions to the T. which I'm pretty pretty sure they can.
Speaker: right. then I think it... is good. there There are going to be open source guardrails. about how do you deploy this in production, right? and Once you do that,
Speaker: ah We've all learned that that trust seems to be pretty easily won. with air. Like you said. Rubik's Cube can tell that you need to fine tune your alert. There is no possible root cause. Just fine tune it.
Speaker: you don't need for this threshold change the threshold threshold That's clever. You do this, you read this three times and then you change the threshold, you know you're starting to trust it now. So time. that's Yeah, that's exactly. I think the earning trust metaphor comes into place. yeah You have to spend time with it and see the...
Speaker: actionable it is or doing if If you validate that, If you are comfortable with that, you start trusting it. Hmm. I feel very confident. I don't really doubt that people...
Speaker: Will not trust trust couldn't At least in the early adopter phase across the crossing the chasm. diagram that you see Your early adopters i don't really think trust is going to be an issue For you.
Speaker: or Rubik's Cube or any other AISRE people. Trust is not going to be an issue. I don't really think that could be like a blocker. The blocker is genuine. if people questioning it or People coming to a point that is This is absolutely going to be useful.
Speaker: or also just sitting down and always questioning this is that But... I can do this with cloud. right and that question mark comes in so that differentiating factor and we're we're fighting against ah Claude. with specific product. a whole different discussion. but I think.
Speaker: Can AI actually help me? Find the root cause. triage the incident. Trust is not going to be a blocker. I highly doubt it. I highly doubt it. Got it. i I partially agree to that.
Speaker: But... I have seen... People who operators who are very reluctant to adopt. And I have also seen a spectrum of people who are very willing to was talking to randy who is a like veteran and legend in the operator community and he is 100 % AI first.
Speaker: Yeah. So... It's... Same with John. in in my last conversation is the same approach of people who have a domain expertise they know Like they can control it.
Speaker: they know they have a mechanism to guard rail it properly they know how to put in the safety measures. to ensure they get expected output and expected behavior they are very confident about. ai They're full on AI. Well, who...
Speaker: lack that education And. do not have that skill set or expertise are the ones who are a bit reluctant What is a... How big is the...
Speaker: the people who have Not been for it. How? What are their roles? And in the team size or in the companies that they work, how big are those companies? How many people?
Speaker: teams I've seen the whole variety. Oh really? Yes. I was at SRE day just a while back and I was talking to folks my Talk. was about the fear paradox of SREs.
Speaker: Yeah. And the paradox is like this. You have... a fear of using ai in production systems or not using it. What are fears?
Speaker: If you don't use AI in productions, ah Yeah. generally don't use ai you are getting left behind behind developers have adopted it fully and you are just getting left behind so you have that fear okay i want to learn it i want to understand and do it use it properly And if you use it, then you have a lot of problems, hallucinations, cost.
Speaker: uh guard railing loss of control and everything else so both are the fears that that is stopping you from actually starting to utilize it.
Speaker: Right? Yeah. The first fear about Using AI and all the lists that i i say nine ten majorly issues that were there. or people are afraid of these pointers.
Speaker: The revelation is Kaushik These are the Specs These are the PRDs. ah you know that need to be developed. because if you are afraid of using ai in production systems or any working life system and you are afraid of hallucination you you need to build something which hallucinates less if you want.
Speaker: more control then you need to have proper guardrails. So things like that This actually is a spec of what is the requirement for an operator's AI.
Speaker: It has to be pretty easy to sell in one sense that This is significantly going to bring down your MTTR. the mean time to resolution right that alone is a metric good enough if If you're running a bank,
Speaker: There is a critical incident. and we can reduce the time to resolution from say 2 hours which is plenty. it's It's a lot, two hours is a lot, but still two hours.
Speaker: down to about 20 minutes. You know? You can have all the money in the world. Let's do it. Whatever it takes. Whatever it takes. As long as there is the significant bet that it is definitely going to reduce the MTTR.
Speaker: right by a huge percentage I mean, Look. I'm... We know of teams who would Invest. in just reducing it by 20 % and 30 % margins.
Speaker: But I think. what we are discussing is much significantly larger than that could could be. could be significantly larger than that not for all incidents but But... a chunk of them. And I think...
Speaker: If that is demonstrated... tradable. then everything else Uh... The universe will find a way. Everyone will find a way. of how to We often talk about blameless postmortems, right?
Speaker: but In live production systems, there is just one thing that you need to... do wrong Yeah. yeah right and the operator is questioned why Why did you allow AI to do it?
Speaker: even if everything if ai wasn't there then also this could have happened but still there will be the question that's the fear people are more operators are more afraid about but having said that in the talk this is what i talked about like these are the specs these are the product requirement document that you need to follow and build an ai uh a harness a better thing around it The second where Kaushik is also A paradox, it's...
Speaker: It's a... The revelation there is, SREs? ah If they don't learn AI. right the time is very less the clock is ticking right and it's it's going to The gap is going to be huge if they don't do it today. Hmm.
Speaker: But the revelation here as well is SREs are very good with deadlines. Right? Really? Yeah, they they love being in chaos, right? You have to fix this issue in, let's say, one hour.
Speaker: yeah they are the ones who are brought into those yeah situations right so they are good with timelines they are good with deadlines so they can learn fast They can adopt very fast. And you have a PRD that you need to develop. these this breaks the illusion of the fear that you had.
Speaker: And the talk I gave. initially everybody was nodding their head yes i'm afraid i'll not put it in production and by the end they exactly knew and walked out with a prd with a proper a schema and spec of what needs to be built or bot. right if somebody else is doing it So, this was very nice and eye -opening talk that i did and i hope uh it got conveyed the way i wanted to
Speaker: Oh yeah. people there It does, it does. You are at a really good position. Two. walk people through this.
Speaker: on how do you make that transition. It's a really interesting. piece of content that you can talk about and I think it's a game changer for Rubik's cube.
Speaker: if It is helping a lot. The whole... concept of the root cause is helping a lot. and The people I'm... you know talking to I don't know about my audience.
Speaker: But I personally am like getting a lot of insights a lot of industry experiences being transferred to me by these veterans by you by john by randy and everybody is talking about why we should be like full on focusing on adoption.
Speaker: full -on focused on utilizing and implementing ai in the ways that can that it can help us and things will go grow from here Right? Things. It's...
Speaker: I think it's the worst phase that AI is in today. And from here on it's going to get better and better and better. It's only going to get better. It's only going to get better.
Speaker: There's this Honesty thing. with AI. That is. A challenge. if you ask it a question It answers you. Okay.
Speaker: Claude, I am designing this UI like this is it good. Yes, it's good. And he has three reasons why it's good. Yeah, he has three reasons why it's good. Now you ask... k Claude Be a devil's advocate and tell me why it is bad.
Speaker: here are the three reasons why it is planing Right? So that's the honesty gap. in ai or lms today that they do not Say no.
Speaker: Hmm. One thing that we trained Rubik's Cube, it's not a model by the way, it's a harness, but we trained it to say no when it doesn't know the answer. That's the biggest trust factor that you can bring in.
Speaker: We spoke about honesty in our last Like. Conversation Kaushik You learned something when you... I had a visit at New Zealand.
Speaker: Right would you Would you like to touch up on the story? Oh, yeah, yeah. I think. so in my um Am. In my earlier, in a few years ago,
Speaker: Um, we were we weren't we we still did not have these many customers we had a very small set of customers and I was ah visiting Uh...
Speaker: New Zealand with my family. Mm -hmm. And at the time... Uh... A lot of our plans have gone up and down. Last minute changes because of...
Speaker: We're done. mainly lots of There's a storm? And fires broke out. Yes, Fires Broker. Correct. and we're going to milford sound it's this is amazing you should absolutely google milford sound Okay.
Speaker: So, we were going to Milford Sound. we're gonna go the other route. That's fine. We could not. We reached Queenstown. And when we were in Queenstown,
Speaker: There are already kind of locked down much of the places. Um, End. Towards Milford Sound. Now, you should know that the incident happened.
Speaker: Um, Like a day ago. and the next day I'm like 60 kilometers away from Queenstown. And we're visiting this little river. I think it was a lake.
Speaker: and At that point, like the tourist point. You know how they've got these directions. go left and here and here and here and they finally did like this is the beautiful place that you walked all the way.
Speaker: to see. And we got there. my first intuition. is perhaps to, there's just Put your feet in the water. That's that's me. That's typically me. Right.
Speaker: But there was a board there, really nicely polished. brand new board. suggesting a notice that you should not get into the water This might have been contaminated because of the...
Speaker: Amen. fires. Yeah. because of the fires. Right? and I read that and I did not go into the water. We also had kids.
Speaker: So, Really handy. And. I was surprised. by the response. I mean, You could have just put it on an A4 print sheet, but no, this is preparedness.
Speaker: Like, these guys have come prepared. With time stamps and everything, right? Someone actually just went and then they they got this entire thing made, planks made everywhere and they put up all the way to Milford Sound across so many tourist destinations.
Speaker: that don't get into the water or here's a warning system or anything That's preparedness. I mean, I absolutely loved the transparency as well. Like it's it's super transparent. They're like, look, this happened.
Speaker: is what it is. Don't do this. Don't get into it. And... When I when i get got back home... Oh. we We weren't using status page.
Speaker: as much at the time. Okay. But. It... i I like... I truly understood how impactful status page... Could be. which is also why we built status pages and everything after that.
Speaker: Was that? Now. we we We run spike. with absolute honesty and transparency if there is an incident we put it up on our status page first and For the first year, there will be so much debate about, are we going to declare this publicly? Are we not going to declare this publicly?
Speaker: And it's like, oh, no, you know what? if if we if we declare this publicly wouldn't wouldn't our customers Aren't we losing their trust? they'd like Wait a second, you guys. Are you kidding me with this incident? is this is this the kind of world -class software that you're aiming to build i don't trust in you guys at all You know?
Speaker: And... These questions keep coming. into our minds. I don't know if this happens with everyone. I don't. i Did it ever happen to you when you were declaring public status wage incidents in the past or anywhere?
Speaker: It did did not. See, the I don't know. It's just me or... everybody that I have known were always willing to be transparent. We see Claude. gemini everybody has having their own status pages Uh,
Speaker: Being transparent. and being accurate, honest about your issues, your problems. and when things go down is It's getting you more trust actually. It's the other way around.
Speaker: but This is what we learnt. over a period of time. This is what we learnt. we we We felt like ballpark, it would. I think customers largely would Appreciate it.
Speaker: Um. but It. did not cast away from the fear that, hmm, should you? And you should also understand, I think.
Speaker: one One of the things that I really appreciate about all of the... AI labs is that they're so transparent about Just telling that It's not available. and right which is going through like some sort of a partial outage or anything like that.
Speaker: but i don't really think that we saw this much open honesty for this we had to subscribe to github status page linear status page and just keep an eye on how much they're updating and jesus guys they do They're consistently out there, right? And that really motivated us.
Speaker: And. Ever since then, we've been we've been absolutely transparent about everything that happens. But largely I want to give. a huge amount of credit to Everyone in New Zealand, their rescue operations teams, the firefighting departments, everybody.
Speaker: i'm pretty sure as a small country they would have so many volunteers coming in together as well the way they pulled it off. miraculously. Let's go. Pretty freaking awesome. the preparedness that goes into it because for us Oh.
Speaker: I think until you've not Prepared? It just feels theory and unnecessary. Yeah. Incident preparedness. but huge kudos we came back and that kind of changed the way we look at incidents and we built status pages and now we've got so many customers using status pages Now we've found a way to connect it. do it properly. You can run the entire workflow from Slack.
Speaker: ah makes it much much much easier. Yes. i I was very fortunate to be working with companies who... already had this Culture integrated within their systems and this this is what i derived when i started rubik's cube the first thing after building the product was to put up a status page. Because...
Speaker: I... I wanted to know it and I wanted my customers to know when it fails. Unfortunately, I haven't seen any failures. yeah And I'll tell you the reason for it. And that's very interesting. Rubik's cube runs on Rubik's cube.
Speaker: It's eating your own dog food, right? so it runs on rubik's cube and we have enabled in production that that's what we propagate we you know educate people about so unless we do it it's it's not gonna work right so we have seen lot of incidents coming in getting auto resolved with Rubik's Cube and.
Speaker: I am getting informed. in form of an RCA and a postmortem. This broke. This happened. This is how I fixed it. And I'm surprised a lot of times okay This is exceptional. I don't have to touch my production system at all.
Speaker: Yeah, yeah, yeah. That's pretty cool actually. do you think the gtm strategy has changed over the years now Yeah. It always changes.
Speaker: It always changes about where people hang out. What's... What's the latest playbook for GTM today?
Speaker: Depends on the product. Depends on the product. but Um... If you are... i'm I'm also the worst person to speak up. gtm width i must say this okay but Um,
Speaker: I think. fundamentally Just understand. Where are your BIOS? slash users. And number two. What are they asking?
Speaker: And... Just answer that. and be there. Be where they are. be in the same room? being the same conference. And be there.
Speaker: when people ask these questions. So basically just the way people are moving with the new approaches It used to be let's say an event that happened.
Speaker: five years back. and that's the event that you would want to go today it's a lot of uh social media Podcast could be one of the things, right?
Speaker: so This is where people hang out and this is where you should be present on the comment section on engagement fronts. and So. The methodology remains same, it's just the channel changes. there's just the channel changes 100%.
Speaker: 100%. People... people in the in the 19 when When Vijay Mallya took it from Vithan Mallya, he... He started. Perhaps one of English... bangalore's first pops Right?
Speaker: So he gathered people. Right. to this one place but at least he knew where the people are going to go sell them Sell the PO. And before that, they knew where people are going to sell the beer.
Speaker: Right? So, ah the interest in beer remains the same. Just the channel has changed and the product has remained the same. Right. in Enough.
Speaker: And that's that's the most important part. the the GTM part of it is Just remember the virality is not a GTM. Playbook. Hoping for virality across something is is a hard game to play.
Speaker: wouldn't be Wouldn't it be good for a new... Startup. It's always good. everyone wants to get viral but it can't be a part of the playbook it's It's a...
Speaker: Like. a glucose pill that's all Give you instant rush and that's it. You don't even know if it's going to get viral. But you're gonna spend all the efforts there.
Speaker: yeah Yeah. you know what i mean it's basically saying that i'm gonna start boxing and i'm gonna go fight Muhammad Ali the first match. so i can get viral right Do the do fundamentals.
Speaker: Throw punch first. Learn to throw punch first. That part. I think the GTM gtm playbook is... It keeps evolving, it keeps changing. Yes?
Speaker: um But the fundamental. Good luck. is always going to remain the same Stick to your fundamentals. aware people are. See what they want. and solve that.
Speaker: What problems are they facing? not just what they want. See what problems they have. That's the hard part. When Coffee metzki Met Kale is a book about jobs to be done.
Speaker: And the jobs to be done framework is a fantastic framework. What job? is your user. hiring your product. to do. and why should they switch from an existing solution.
Speaker: And how are they solving it now? That's it. Yakuza 3 has failed door. um, Good is filled with their mini fridge. And they are hoping that people would.
Speaker: like it was affordable so they they hope they were hoping that the segment would use it to store milk. and vegetables. so they don't go stale.
Speaker: but then After the product failed when they were interviewing their users, they learned that people would just end up boiling their milk two times a day. and they'll buy fresh fresh vegetables nearly every day. You know?
Speaker: ah So they every product is hired to do a job. and focus on the problem, focus on the solution and then focus where these users potentially are and just go there.
Speaker: And that's it. That's your entire fundamental. about. the gtm after Yeah, I want to switch a bit. We we have been, you know. talking a lot about Operation stuff.
Speaker: this is the is the uh Sure. till now regarding um operations about infrastructure About products.
Speaker: Yeah. That... Cater these. operators. and
Speaker: Lot of times, lot of things that we talk about. is around those things We uncovered UI today. Which is a very... interesting fact here interesting uh approach to begin the show with we spoke about gtm we spoke about product And this is what I wanted.
Speaker: From you Kaushik. Uh, because It gets monotonous if we keep continuing on the same. 30, right? um bringing these concepts out from a founder who has been into the weeds for very long.
Speaker: Gibson. good enough insight and motivation to the younger folks who are trying out new products who want to build a product right So at least they know what not to do.
Speaker: but kohika advice to what what to do right so this this helps a lot But I would like to switch it a bit. um There's another. Think you were built.
Speaker: ah three months back. Yeah. Right. You became a father in the middle of building a company. congratulations thank you thank you and how How is your experience post that?
Speaker: If I start a review. to the baby. Pfizer review. ah All across, I think. ah I... I think. it's It's... ah The role of a I don't know Three months is such a short time also to discover so many things.
Speaker: All I know right now is that everyone is healthy, which is really nice. And uh... the the brain develops very slowly. You know? it takes the the cognitive development i think that is what is cognitive development right like you're learning like i think now after at four months they're like oh this is my hand I can use it to grab things.
Speaker: you know. So you can see it slowly until it can get to a phase that it will solve puzzles. So, you can't just Immediately go in and start teaching. abcd or you know it can probably so it can identify the sounds but i've also understood i've come to this understanding I don't know I've not googled it is that The first words are
Speaker: Amma, Appa or mama papa whatever right like on and mama and dada and everything because it's so easy to say they're not words they're just sounds Right?
Speaker: So, those those vocal sounds, the repetitiveness and everything. How do you do that cognitive development? Training Like training the baby? is is completely different.
Speaker: You know? but it's a fun one. I am very excited. with uh for the future it gives some It gives completely different... ah Might I say purpose?
Speaker: Right. heavy but it's fine i think yeah it kind of feels like it gives completely this is a new kind of an on -call that you have to be yeah don't i know ah
Speaker: ah but You do not have a shift rotation as well. There isn't one. Nice. Awesome. smith Very nice. Very congratulations.
Speaker: and Many, many uh congratulations to you on that and Thank you. don't treat it as engineering task so Stop dissect dissecting how the brain works. but Enjoy the moment, I think.
Speaker: That's what it is. yeah It's fun. Um, I would like to flip the mic now. uh Do you have anything for me? Yeah.
Speaker: I think i just I just wanted to ask one question. Out. Where do you see? the role of SRE in 10 year timeline.
Speaker: Interesting. in 10 year timeline. There is an incident in 10 years. Hmm. Talk about what is that incident and what happened. And 10 years is… compared to the last 2 years, 10 years since. Crazy? Yeah, it's crazy. Maybe five years is a... more relatable way to think about it.
Speaker: I asked a similar question to Reddy. uh in my podcast He has an opinion. I asked to John he also has an opinion For me...
Speaker: i think so I think things a bit different. uh with the way the explosion has happened. with the way AI is being adopted and The growth is happening.
Speaker: This is going to be a J curve. So it starts good. slows down a bit and then again extends a lot. so as i said right we are at the worst phase of ai development and then it will it is going to get better from here on I think...
Speaker: Infrastructure should
Speaker: Maintain itself. There shouldn't be any humans. operating it And that's very contrarian approach, very controversial thing to talk. uh But I'm not targeting SREs. I'm targeting the role. So today I'll give you one example. do you think when you're writing a piece of code do you think about how the compiler is going to compile it.
Speaker: Exactly. You don't, right? yeah That's the... you know, vision I have. Within 5 years, infra should be intent driven. I want to run this system. I want to run this production system. And this is how I want to do. This is the intent human should give.
Speaker: Now maintaining that, that's the Kubernetes approach of... intent -driven desired state versus real you know actual state. so this is where the reconciliation should happen the infra should self -heal itself if a problem arises or um Come to you. Obviously, it should come to you when when you are needed.
Speaker: But... infra should be the one coming to you the Boss, this is the problem I am facing. you need to fix it by this is approaches i also know the solution Right? But maybe I cannot take action. or maybe i can't take action so this this is where i'm thinking uh towards Infra should be invisible.
Speaker: Hmm. and SRE the role I think is not needed. the role will transition into something else these are brilliant folks these are one of the most intelligent folks in a company in an organization because You wake them up. at 3a Right? And at that cognitive state of mind, they can come and fix an issue that is going to impact
Speaker: ah hugely on revenue. that will that is going to impact hugely on trust of customers so that's the the capability they have. We are I think we are wasting that energy into something which is very Trivial It should be utilized.
Speaker: in a more fruitful way. that's that's my opinion on it Oh nice. That makes sense. Am I too far fetched on this? No. I don't agree 100 % with it.
Speaker: I think what will happen is lot of the parts will get Um, Obviously. because we will solve the same problems better with AI.
Speaker: so there are going to be new types of engineering challenges there are going to be new types of incidents there is going to be infrastructure that's self -healing and that's correct. right But infrastructure is not the only incident.
Speaker: Right? right There are business incidents. infrastructure incidents are Until and unless something is absolutely crazy, but by most defaults, I think everyone is always pivoting towards.
Speaker: What is the business impact and the what is the business incident here? Right? like say let's say for example if you're a if you're a booking if you are a OTT platform. then videos and uploading not necessary it's it's it's a it's a technical uh, And...
Speaker: And... If you have booked my show. or make my trip. the fact that we're not able to book the ticket across the airline. Right? Now that's a business incident.
Speaker: I think there are going to be more business incidents. and those are the incidents that also comes up with a big question mark. Although, they will Uh... our ai and everyone will learn to probably jump heel it or anything.
Speaker: But because we will continuously keep improving and companies continuously keep deploying new things. with or without AI. that means we are now open to open to new types of incidents we have not seen before Hmm.
Speaker: you know. Um, and which is why one thing is the similar approaches that we take today. What SRE does. right um That may not be valid. that may not be yeah That's what I said, right? the The role will transition into something else. New role will be created.
Speaker: They may not call it a SRE then. Yeah. What would do you call it?
Speaker: That's a tough one. if you had to coin it. You coined it today.
Speaker: I'm not able to think anything. Yet. That's something we'll have to see. uh down the line it's an operator that
Speaker: Intent Engineer Eh, what? intent. engineer. Hmm. Yeah. I will think. Yeah.
Speaker: But I think what will happen also is largely... Most reliability engineers. just just reliability engineers i think they they'll not be sres they'll just be res so reliability engineers will move to sdlc at not just google but they I think that that everywhere at amalgamation and, you know.
Speaker: mixing of this is the other way around people are going you know development yeah to come to production and that's the shift we already are seeing we i have seen a lot of developers who are well versed with infrastructure.
Speaker: Hmm. I think... the sri role was that right You brought in a developer in a sysadmin's room. so that they can automate stuff. that same transition will go bleed into the development side as well we often talk about shifting left a lot of you know incident that come into post -production should be handled at the development phase this is sres operate on i think this role transition will eventually merge towards
Speaker: Just engineers. who are well versed with the whole system. and Developmental production. Hmm. We've been through that before which is when we started to fabric segregate it again, so we'll segregate it again Yes.
Speaker: We started there. Everyone is a software engineer. Now we're like, no, no, no. You know what? You're a security engineer. You're an incident response manager. Then we started breaking it down and segregating. That's fine.
Speaker: But it's a fun one. It's a fun exercise. We'll see. Bye. I really liked uh our session today and would want to want to have you more often on the show yeah you guys you know,
Speaker: Parallelly, we are working on the similar objectives, similar products. um lot of times there will be overlaps uh lot of times there will be I don't want to call it conflicts, but Conflict of interests Probably.
Speaker: it won't be conflicted in this because you are trying to help a reliability engineer and so am i I think we can help you transition to a more Intelligent alert triggering and you can help us get better re alerting uh triads okay There will be a synergy. Yeah.
Speaker: so yeah i think ah people would love to know your journey as in when we progress uh Any last words for my audience? Oh no, yeah.
Speaker: Thank you. Thank you so much for listening. Thank you for. Um, thank you for well Thank you for listening. and if If you are... Anybody? who has not tried.
Speaker: Cloud code or anything Please do. It is... It is a perfect time to build. whatever you've always wanted to build because the most The most... a famous programming language right now is English.
Speaker: Okay, so build. and if you're going to build a startup there is only one thing that i would like to say there are actually two things that i would like to say Number one. is Um,
Speaker: If you are sorting today and write down what is here. not product goal but write your personal revenue goal and then reverse reverses back from there.
Speaker: That's how. it's the I think it's a better way to build. your startup, your company, your future, everything. Write it down. timeline money. Nothing wrong with it. It's going to be hard exercise.
Speaker: tell me exactly how much money you want to earn from this in 10 year timeline or a 15 year timeline it's a great way to then you will automatically sit down and weed out a lot of product ideas or solutions and do you want to raise not to second is Um,
Speaker: Despite of the fact that I am bootstrapped, I would like to say, VCs are not bad people. Just go for it. You know? not all businesses can be bootstrapped if you think that capital is going to help.
Speaker: and or a good investor partner is going to help, just go for it. Let's go for it Yeah. Nice. lot of Operators.
Speaker: in the same space who were using similar products for alerting. by Should they switch to Spike SH? It's simple.
Speaker: Spike is... Very simple. To use, Spike is incredible. incredibly simple to use and pricing we are very friendly on your pricing. Right?
Speaker: A lot of people come in because they find pricing is an attractive offer. and then they stick because Pike is absolutely simple and such a delight to use. You will like it.
Speaker: Give us a shout out. Give us a try. Come in. If you're especially coming in from Obscene or anything, um if you're coming in from Obscene and PageDuty, then... and also Zen duty because Zen duty is now acquired.
Speaker: um then we do give out. Ayutthaya offer for the first year with spike Awesome. Yes. Awesome. And I think...
Speaker: pretty soon we are having uh like spike integrated with rubik's cube as well Fingers are crossed. yes and yeah i think it will be very interesting Yeah.
Speaker: Awesome, Kaushik. Thanks a lot for taking out the time. Uh... giving such good delightful moments and Such insightful talk. and I really would love to have another session with you.
Speaker: One more thing. But yeah I think we'll stop it here today. and I hope i have people like it. People, if they have any questions, do post in the comments and we'll reach out to...
Speaker: It's goshaken. have it answered Yeah. Thanks so much for listening. Thank you so much for having me, Priyank. It was wonderful. Let's do this again sometime. Thank you, Dasik. Thanks.
