AI Capital Squeeze, Crypto Mortgages & The Future of Coding

This week on Software Sundays, KD breaks down major shifts shaping the future of technology, finance, and software engineering.

From the rise of new AI infrastructure players like Cerebras and what that means for NVIDIA, to Cursor’s $50B valuation and the real economics behind “vibe coding,” we explore where the market is actually headed—not just the hype.

We also dive into how crypto is entering the mortgage system, what that means for access to capital, and why increased financialization creates both opportunity and risk.

Finally, we answer key questions every builder should be thinking about:
- What secure coding actually means in practice 
- The real challenges of working in large organizations 
- How to evaluate job offers strategically 
- Observability best practices 
- How to decide between building vs buying
 
CHAPTERS: 
 
00:00 – Intro + Mission
01:00 – Cerebras IPO & The AI Infrastructure War
08:30 – Cursor’s $50B Valuation & Vibe Coding
16:30 – Crypto Mortgages & Financialization Risk
25:00 – What Are Secure Coding Practices?
30:30 – Challenges in Large Organizations
36:00 – How to Evaluate a Job Offer
41:30 – Observability Best Practices
47:00 – Build vs Buy Decisions

DISCLAIMER: This is not professional advice. The views expressed are my own or those quoted. Consult your own legal, business, or tax advisors before making decisions based on this episode.

Build Learn Impact is on a mission to help you create wealth, opportunity, and ownership through technology.

Transcript

SPEAKER_03 0:00

Welcome to Software Sundays Builders. This is a space for high-level conversations about technology and the impact that it has in our community. We make sure you can walk away with tools that allow you to grow your income, become an owner inside of the economy, and help shape what's next. If this is your first time tuning in, you are in the right place. And if you've been rocking with us for a minute, thank you for making it back. Quick disclaimer before we get started: Software Sundays is for informational purposes only and is not professional advice. The views of individuals quoted and may not represent Build Learn Impact or their affiliates. The topics discussed may or may not apply to your specific business situation. So please consult your own legal business professional or tax advisors before making any decisions based upon information you found in this show. With that being said, let's jump in. So first on the news for this week was Cerebra's IPO is placing high-stakes fights for public market capital on AI infrastructure. All in all, that just means that Cerebra's is an AI chipmaking company that is currently seeking to go public or creating their initial public offering, meaning they want more or they want to raise money from the public markets for expansion acquisitions and growth. And this is at the same time where there are other companies like Anthropic and OpenAI and even SpaceX that are also seeking to go IPO or go public over the next year or so, or within the next year. So they recently released their investor Prospectus, which is basically one of the steps that you have to take in order to start the process with the SEC and make sure you cannot start trading your uh your stocks or shares for your company on the public markets. So a quick background about Cerebras is that they are directly competing with NVIDIA for a take or a piece of the cake of the AI inference layer. Right now, AI from an infrastructure perspective is like a lot of it is split between training models and inferring models. Or when you're training models to either perform some type of task and then when you're actually running those models for those specific tasks and use cases. Those are like the big differences between AI or the big differences for how AI or hardware for AI is being used right now. GPUs are primarily used for both, but there have been re there has been research that has been done that is saying that there are certain areas where you would rather use different technologies. Cerebras is coming up with one with one example of those, of that, where instead of using a GPU for training or for inferring, you use them only for training and allow these other more specialized chips to actually run the models day to day for the operating, and that's over the long term. So Cerebras is a hardware chip provider. So they are working to create an architecture for chips that will allow these AI models to run on them faster, and we've reduced the amount of energy required to run them as well as the amount of uh the amount of memory required to run those applications. If you think about AI today, it requires a significant amount of memory in order to run an inference task, right? It needs to pull its data from uh usually distributed chips, which would usually be distributed GPUs, which are provided by NVIDIA, uh to actually make sure the output of the model is accurate. With these chips from Cerebras, you would need less chips because they would be bigger chips that could actually hold the memory and the processing and do all of that together on the same piece of silicone. So it's a difference in how we want the models to run, but from a very low-level hardware level. If you think traditionally, most people and most compute limits have been surpassed by getting transistors to be smaller and smaller. Transistors are the individual connections that are in chips or on chips. The smaller those are, the more CPU cycles you can actually fit or you can actually get done in a uh unit of time inside of a CPU or some other processing unit. GPUs are also common. But this shift that Cerebras is kind of looking for is to stop focusing on getting smaller chips and actually make the chips big enough to have all of the resources that they need onto that same piece so they don't have to actually go do the latency between moving data between one chip and another chip. So it's really a trade-off in how you want that model to process and use the data it needs while it runs. So to me, the interesting part is that their value in the market today is directly tied to the supply chain issues for high bandwidth memory. Right now, GPUs require HBMs, high bandwidth memory memory in order to actually complete any computations, right? And you need to, when you're creating the GPUs, you need to actually wrap them up and package them with the memory before you ship them out. So that means that in order to actually ship out a fully functioning GPU, you need a consistent and reliable supply chain of memory available to you, which is more difficult to come by today just because of all of these hyperscalers like basically buying out the next three years' worth of memory from a lot of the major memory suppliers. So that supply chain tightness is something that Cerebras can benefit from because they can say, if you go with our product, you don't even need to worry about memory, or you don't need to worry about memory as much as you do with the competitors in the space. So right now, today, it's probably a great opportunity for them. They they'll probably make the most money, and their IPO will be able to raise the most money based on the current market conditions. Over the long term, I don't know how I feel about this company competing with NVIDIA as an entire brand, like just brand value and just market size. NVIDIA still has an advantage because their GPUs are still best in class for training and even inference task tasks. So I don't really I don't know how easy it will be for them to pull enough market share to actually make a significant difference inside of the business. Their revenue is more directly tied to individual customers, meaning that they actually need for these customers to actually succeed for them to make it. Nvidia's customers and clients are a little bit more widespread so that they have a better chance in case any one client or customer uh is not able to meet their needs. The challenge when you're building purpose-built hardware for an industry like AI, like data centers, is that the initial CapEx cost to actually build out these materials and build out these supply chains is something that you those sunken costs don't go away. Once you invest it, you kind of have to go all the way before you can even see a return inside of that investment. So interesting part there. Another challenge that I am seeing is that NVIDIA is very much aware that they are not or their GPUs are not the best resources and tools for inference specifically, but they have made other acquisitions and purchased other companies that are more directly tied to that specific niche inside of the market. So they are they already understand the competition that they're up against, and so those companies that they've merged and acquired will have more resources available to actually go invest inside of the space that this this little gap that Cerebras is actually you know looking to penetrate in, penetrate further into, I should say. So it's something again, it's a very tight competition that they're going to need to be very mindful about over the next six to twelve months. But over the long term, I just don't know if they'll have the capacity to really compete the way they need to. There's a lot of risk in there. But you know, if they're if they believe in it, they started a few years ago, at least a decade ago, so they are at least aware that of the technical uh challenges related to actually building out the solution that they're proposing, or not even proposing, that they've proposed and that they're just scaling on right now. So I'm sure they understand the challenge, but it's something that I would keep in mind um from an investor perspective. And even right now, with the number of IPOs that are expected, and some really high-profile IPOs, I also see a challenge in them raising enough money to actually do what they need to do to compete. Right now, they're going to need to hire a significant amount of talent in order to operate in the different regions that they want to operate in, in order to comply with uh different laws and regulations. That talent is going to require them to have some cash on hand. That cash on hand is something that they could get from public markets, but when you have OpenAI and Anthropic and SpaceX also going public at the same time, that same dollar that could go to a company like Cerebras is more than likely gonna go to one of these other companies because they just have a larger brand name. So that's gonna also be a challenge when they have to figure out how much money, how much of the pie are you actually going to get access to when you do go public? And there's definitely an opportunity for the software product layer to continue to you know pull in as much investment as possible. Uh there are billions of dollars being raised right now for the product level of AI. People are still getting to know and use AI from a user perspective. And so we're starting, we're still starting to see actual adoption of these tools. So the those products have not actually seen their full potential in terms of revenue growth and user growth. So it's very interesting to see and keep an eye on that to see you know where is the excitement focusing, especially when we start getting closer to the IPO markets for these companies. If we separate the software layer from the like the like the AI software from SpaceX, which is very interesting with SpaceX because XAI, which is a model AI model provider, foundational model provider, they merged with SpaceX in order to make the company a little bit more valuable and attractive to investors. So you have an AI company with a space company, which is a totally different beast than an OpenAI and an NVIDIA or some other company, right? Like that combination of products and services that you're getting from SpaceX is totally different. You're getting telecoms, you're getting military. Well, they all kind of have a little bit of military exposure, but telecoms, hardware, space travel, like that sector is totally different. And what you do with AI in there, I don't even know at this point. I don't know enough about space and SpaceX and you know space debris and cleaning space centers and transportation for uh these different um, what do you call it? Satellites and space stations, like that entire market is open to SpaceX. And if they have some of that funding that can come from AI right now, but just because of the excitement and hype inside of the market, they can get some of that and pull that and send it into the space sector and get some like because they're one of the major, well, they're one of the only major competitors inside of the space market or space sector. So if they can pull the investment today and then go do something with it in space, that's very exciting for SpaceX investors. So that's the interesting part to me. Today is like, wow, I don't know if this is the right time for a company like Cerebrus to be trying to go public. They tried to do it a few years ago, I think back in 2024, 2025. They started the process of going public and ended up stopping. That was likely due to all of the turmoil happening inside of the stock market already at the time. Just companies didn't feel it probably wasn't the best time to raise anyway. Now we're seeing a little bit more appetite for equities and for these assets that could allow them to raise more money, but the competition is here now. So you're gonna see, you're gonna we'll we'll see what they do with the opportunity as we get closer to the town. Uh, additionally, in terms of raising money, cursor is raising 50 or cursor's new valuation of 50 billion dollars is signaling a shift in how developers are building software. This is another flag and show of support for vibe coding as the wave forward. There's a lot of excitement from coders and non-coders alike for not um vibe coding tools. Do you think cursor, lovable, uh cloud code, codex, all of these tools help make it easier for us to deploy software at scale. Make it faster, make it cheaper, make it arguably more reliable. Although, if you don't have the skills, you will likely not see a significant increase in reliability. But basically, right now, Cursor is looking to raise, I think another 2 billion, or they probably just raised 2 billion, and but they're being valued at 50 billion. But it's showing that the interest in getting development costs to decrease has not gone away. We are still, even in the age of AI, even with more products being built faster, we still see a challenge with actually getting products shipped. So all of these tools make it again, make it easier for you to ship. One of the backers of this fundraise is Andreessen Howlowitz or A16Z, uh a venture capital fund, where they already have significant investments inside of AI, inside of uh crypto, and in different companies. And I think the reason why they're even backing this company or Cursor in this fundraise is that they want to ensure that the tools that their startups need in order to scale faster, cheaper, and more reliably make sure those tools exist as early and as often as possible. So I think it makes sense for them to be investing in cursor. I think it's also important or interesting that NVIDIA is also investing in cursor. They are cursor is another use case for NVIDIA chips. So they're basically putting money into one of their customers to help make sure that there is a consistent future demand for their chips. That makes sense. Now, for everyone else, I don't know how valuable of an investment Cursor is to you because they don't make any money as a product, right? So their product is amazing. If you actually use Cursor to code, you can see that they have done all of the necessary work to add AI into your workflow. So as a developer, getting a change from task ticket to PR and deployed so much smoother, so much easier, so much faster using a tool like Cursor. But when you look into their financials as a product, they are spending a significant amount of money on providing their product or providing their service. Those token costs that they used in order to make sure the model is making the right change and reviewing the code base and doing all these things, that's very expensive. So they're spending more money to deliver the service and product than they are getting from the customer's subscription cost. So something inside of that equilibrium in that model is going to have to change. Either we're gonna see an increase in the cost of the tools themselves, which could potentially decrease adoption, or we're gonna start having to see that they need to bring the token cost down. And they have gotten a little better at this just with the introduction of different Chinese models like Kimi, which allow you to run you know pretty good models, pretty, you know, get pretty decent LLM results with a fraction of the cost. That opportunity is definitely there, but I don't know from a equity perspective. I think that a company like Anthropic, who has Claude Code, which is a pretty good brand, uh they have their own models, right? So they don't have to pay the they're paying wholesale prices for the tokens themselves in order to provide the service, that AI assisted development service, and they have deeper partnerships with the cloud providers, they might be able to get better numbers on their actual cloud computing cost. That's something else that I think in terms of an advantage, OpenAI and Anthropic, they they kind of have that already. And even GitHub, you think of Copilot as your well, I wouldn't actually compare copilot and cursor. The AI completion, code completion is similar, but in terms of the product and how it actually makes changes inside of the code base, I think there is a significant difference in cursor and how it works and GitHub Copilot. It doesn't do exactly the same, but you can get some significant benefits from AI. And automation inside of GitHub and their ecosystem, also. So it's definitely a few things to keep in mind from a product perspective, but I don't know. Like I I definitely see the value inside of Cursor, but again, I don't know if it's as valuable as of some of the other products just because the numbers don't make sense right now. For anyone that has not used cursor or claud code, I definitely recommend that you do look into these tools. Claud code is great at making large changes across your code base. It will do a really good job in understanding the just the overall hierarchy of different project files and understanding the like the rules that you have for your code base and making sure the quality of that code base are kept up to date based on any AI changes that you actually request. And it's fully terminal based, so you can actually go from chat and have code updated without having to even understand and look at any of those PRs. That's great. But I think from a traditional development perspective, cursor and being able to see what those changes are and see the exact thought process that your model, that your assistant is using when making those changes, I just think is more valuable and much easier to get used to if you have the skills and expertise yourself. Right? Like Claude is going to make a lot of decisions for you if you're just saying telling Claude, hey, make this change. But Cursor, it's more like if you already know what you need to build, cursor just makes it much easier for you to get that thing out the door. And that's an entirely different mindset shift for when you're building a product or building a service or scaling a product or a service. But the major thing to keep in mind with the development and the growth of Cursor is that the tools developers use still matter. As a developer, you have the skill and the experience and the knowledge to build systems. That could be a distributed system, that could be a mobile app, that could be a software product with a lot of workflow automation. Vive coding lowers the barrier of entry to getting a change done or doing some specific thing. But it doesn't make it easier to it doesn't replace the need for software engineering expertise. As a professional, you still have to understand that the tools do something. If you as develop as a developer don't understand what the tool is designed to do, what the exact flow is for you and your software development lifecycle, then no matter what tool you use, you're going to not be able to compete as well with the other people that actually understand the tools and understand the processes for this industry. So look into either your cursor or your Claude codes and figure out how to actually be more productive, not by just giving your assistant or your agent a task and having them do it. Figure out the best practices and the flows for going from zero to one and deploying something. Because that's the thing that many people are too busy giving AI the keys and just saying go do this, but not understanding why it is doing this with the key, if there's a better way. A lot of models have not figured out the best way of engineering, and there is a best way, there are best practices. If you can combine the speed of these AI assisted platforms and these tools with the best practices that you pick up from industry and from regul from regulatory bodies and just experience, then that edge you get is going to be something that you just it can't be, it can't be faked, it can't be uh replicated very easily. And that's what you want if you're really trying to stand out in the market. And it's some interesting news in the housing industry, always interesting with technology, AI, real estate. So basically, asset backed down payment assistant programs are allowing you to come to the table with less cash because you can basically use your Bitcoin inside of a Coinbase wallet as collateral to get the down payment. Like that's your one loan for a down payment, and then use that down payment loan to actually close on the property. So it's two layers of loans, and the benefit to the buyer is that you get to keep an appreciating asset, the Bitcoin, while also being able to purchase the home, right? So that a lot of people are I won't say against, but a lot of people that own Bitcoin don't want to sell their Bitcoin, right? You want to keep it forever because you know it's going to uh appreciate over the long term. So you don't want to pull it out into fiat currency, and then now you have to pay taxes and you have to pay, you know, a bunch of fees, transfer fees, and then you lose out on that future uh growth. That's one of the reasons people don't pull out their Bitcoin. But if you're able to go and say, I'm gonna take a take my Bitcoin and get a loan so I can buy the prop the property that I want, that allows you to be basically in both race races, which is pretty cool. That's a benefit. The risk that I'm seeing is that now you have two loans you have to keep up with. It's not like a like a normal margin loan where if the value of the asset decreases, it changes the loan terms where you have to maintain some type of like um you know loan to value ratio with your stocks or or with the value of the Bitcoin to the value of the loan. That's a thing that like whatever your entry price is, it's there as long as you continue to make your payments, you get to keep at control of the asset no matter what the price swings and movements are. So that's not the risk, but it's the fact that now you have to maintain two different debt payments or debt service payments, which basically increases your exposure for just risk, right? Because now instead of just saying, Oh, I have to make my home payment, my one mortgage payment for my house, you have to make your one mortgage payment from the house to keep the house, but you have to make your other mortgage payment for the down payment to keep your Bitcoin. And that is cool until you lose your job. That's cool until your expenses increase because something broke in your house and you don't have the cash flow to make both payments, right? And then now not only are you in danger of losing your property, your home, you're also in danger of losing the assets, the Bitcoin that you use as collateral for the home. So it's a very tricky situation to be in if you don't plan and prepare accordingly in the right way. But it is exciting because it gives people more access to credit and capital that they can use to then go invest. But again, we all know that debt is a trap. Um, increasing your exposure to volatile assets using debt is a way to make money quickly, but it's also a way to go down very quickly, too, if things go wrong. And that's not to say that you know things are going to go wrong, but if you don't prepare for it and you have everything in credit, you financialize your life, you expanded your access to credit, and everything is kind of going in that direction for you, then it just it's more complexity. More complexity is always going to be more challenging to deal with when the time comes. But the very cool part to me is that we're seeing crypto, Bitcoin specifically, becoming more and more synced into the financial system. Like there was a time, and I keep saying it, it was a time where we wanted Bitcoin to have nothing to do with the government and fiat and the you know centralization. That was a thing. That was a thing that people really were screaming about for years. But as soon as we started seeing real money come from the you know ETFs, the Bitcoin ETFs, when we started seeing institutions picking up Bitcoin in lows, and that appreciation started hitting, I don't think people even care about decentralization anymore. Or care about you know disconnecting their finances from the system. So it's very interesting to me just to see, like I already am confident it's Bitcoin to the moon, but and I say Bitcoin to the moon, not cryptocurrency to the moon, Bitcoin to the moon. It's only Bitcoin, everything else is whatever it is. Um but it's very interesting how deeply integrated we're seeing these systems become, and especially the different players, right? A few months ago I mentioned, or it could have been weeks ago, weeks ago now, uh the I can't remember the name of the exchange. Can't remember the name of the exchange right now, but they basically have they they have an account with the financial reserve. Like the Federal the Fed Federal Reserve, excuse me. They have an account with the Federal Reserve, meaning they are a central, not a central bank, but they are if a central bank is top numerous numerous a number two with the JP Morgan's with those large institutional banks, they have an account with the Federal Reserve. Like that's not a thing everybody could say. We've got that from a just a banking perspective, they are their own bank directly today. Then you have Coinbase creating their own just lending system, like they are powering a mortgage that is tied into Fannie Mae and the federal housing authority. Like, that's different. That's suit I say different, but these companies are still new, these are babies, and they are going in and literally reshaping the systems, resit, reshaping the policies that actually drive the entire US economy. Like this is different. Uh, so the exciting part to me is that crypto platforms and technology companies win when the technologists that are building the platforms and the infrastructure, when they come together with operators in the space, whether you're coming together with the government who understands and has the authority and the access to just make things and push things forward, or an operator that just understands the industry or the cert, the sector better than anyone else. And when you put those two uh those capabilities together and you would let them just grow and partner and just you know really nourish new solutions, that's when we start to see policies and programs like this. And it's very exciting. This is just for real estate, but there are chain like we could see this same access inside of the auto industry. I'll say the auto industry might be next for it. We already are seeing you know deeper digitization of the find like the auto-buying process from getting insurance to to finding your car and researching your car online. Like a lot of those tools have been built and are being built, even with Amazon putting cars available for sale on Amazon, like their search. So that's that's already there. But and then I would even say I think I think it's Discover or Capital One, not I don't know if it's either of those, but Coinbase definitely has a credit card that you can get where it uses your Bitcoin as the you know the means of paying for things. So I can imagine you being able to get an auto loan and instead of using fiat currency, you use your Bitcoin, your digital currency. So that's all coming. The challenge isn't even the technology, I would say. A lot of the time it's going to be adoption and getting the regulatory bodies to actually say this is okay to be done. We're clearly seeing that the current administration is pro-crypto, uh pro technology, and so they are letting a lot of these changes happen. That's exciting. We'll see what else happens over the long term and how they actually integrate AI with these cryptocurrencies, but definitely a lot going on and worth keeping in mind. Um, so definitely stay tuned. I will continue to monitor this monitor the situation and keep you guys informed and updated. Uh, but if you're serious about joining and building the digital and serious about building in a digital economy, you know, please join BLI University. That's where we go deeper into these shifts and just do more than just talk about it. We share tools that help you lead the future instead of just watching it happen in front of you. So the link will be in the description in the bio. So check it out and you know, stay tuned with us. We're gonna jump into our questions for this week. Uh, the questions are all designed to make sure you understand how to grow as an engineer and as a professional inside of the environments that we are going into. So let's get started.

SPEAKER_00 37:42

What are secure coding practices?

SPEAKER_03 37:48

So, secure coding practices are the practices that you use to make sure that when you're designing, writing, or maintaining software, you understand and assume that your software is going to be breached, or not to be breached, someone is going to try to uh breach your software. You have to assume that your application that you are putting out into the public, whether it's on mobile, whether it's on the web, whether it's on the edge inside of some device. Assume that there's going to be someone, some hacker or some adversary that is going to try to access the data or take control of the systems that you are deploying. And then you use best practices to actually combat those potential risks. If you know that your application is going to be holding a lot of data or a lot of private data, you have to make certain that the data that you hold and process is being protected to the best of your ability. So that means ensuring proper access control with least privileges. Um, that means using encryption whenever possible for data in rest or data in transit and data at rest. And that could also mean not collecting some data because you don't want to have some certain privacy violations or privacy risks involved inside of your company. So when you think about secure coding, it's a way for you as a developer to understand that, hey, there are some things I need to keep in mind when I'm building this app so that I do not put my customers at risk or put my company and business at risk. Because all software is going to have some type of vulnerability. It's almost impossible to build software and not have some type of vulnerability, whether at the supply chain level, like a dependency you use, and it can be something as simple as the curl version of the uh the package that you use inside of your API. It's using some version that has a vulnerability that can be exploited. That can be something simple, but it could also be that you just didn't do proper input validation. So you're trusting the info that you got from the client on a non-private or public uh client and sending that to your back end and saying that I trust whatever they sent me, which is not a great idea.

SPEAKER_01 40:32

Like it.

SPEAKER_03 40:33

You should be validating everything that you get from your user to make sure they're not trying to do any type of SQL injections or any type of cross-site scripting to take control of your services. Some people are really good at this, some people are just getting lucky. Uh, one of the things to keep in mind is that hackers aren't geniuses, they aren't finding really novel and unique ways to reach your application. Most developers make very common mistakes that literally are found in almost every application. Uh, in the industry, you'll hear about the OWASP top 10, which are the top 10 vulnerabilities in web applications that get exploited by adversaries. I think they update it every year or every few years to basically keep um to keep up to date with some of the most recent attacks. But a lot of those a lot of those vulnerabilities again are gonna be tied to you know not ensuring some type of proper access control or authentication, not uh validating input, allowing SQL injections, uh like very simple mistakes that are very easy to protect against if you only just look. So, as a secure coding developer, if you want to make sure that the apps that you are building and working on are you know secure to the best of your ability. Again, we're not looking for 100% security because 100% security is probably not what the business needs, but at least making sure that you are not letting in undo unnecessary risk, doing your due diligence, at least look into these vulnerabilities and check your app against these risks. See if you can run some type of dynamic, um dynamic scanning tool, dynamic automated scanning tool to try different inputs, see what happens when you run it against your application and it puts in some type of data. Does it send results back to your user or to the client that they're not supposed to see? Does it break your application? Like these types of things are going to be just best practices and bare minimum that you can do to protect your applications and your development or whatever teams that you're in. So in small teams, you win by writing code quickly, and you can ship those out. In large organizations, in order to ship, you have to align with significant numbers of people. You have to have endless meetings to prioritize or to align priorities, and you have to have another bunch of bureaucracy tasks and just meetings and things just to get changes to go through. So these are the requirements that you need for a large organization because they mitigate. Risk. They minimize the amount of exposure and problems that happen when a change is made. But they slow execution. Right? The more time it takes, the more hoops you have to jump through, the less breaks that go into production, hopefully. Easier, like conceptually, that's true, but sometimes it's really not true because breaks still happen. Uh, but the slower it takes to actually make any real changes inside of that product. So when you're going into a large organization, think Fortune 500s or even big tech, expect to have to follow policies, right? There's a process to do anything, there's a process to release the code, there's a process to take on new tasks, there's a process to do XYZ. If you are not comfortable following that process, then you're going to have a very difficult time doing well inside of that company or in that organization. So be aware of that. Uh, you're very rarely going to be building something from scratch. Or if you're building something from scratch, it's going to be a very small component of a very large system. And, you know, you just gotta own as much of it as you can if you're just excited about doing that. Most of the production systems are going to have been running inside of production for the last 10 years. They're going to be very limited documentation for how they work. And there's going to be integrations and workflows that people kind of just know work, but they don't know how they work. These are the things that you kind of just run into. And it's not the funny part is it's it'll happen in a small organization, but usually the people that built it are still in the smaller organization. So that knowledge, that experience is all still here. So there's someone you can talk to. When you get to these large enterprises, a lot of the times the original developers have gone. They have changed teams, changed companies already. And so there's no one that can explain to you why this thing happened. So you have to keep just talking to different people and kind of building out a model of what you really see. So it's more exploratory in that sense. But you do get to learn the if you talk to the right people in enough times, you do get to learn why certain decisions were made and you start to see the best practices. Like this is what we do when you have a billion dollars coming through and being processed daily, or a trillion dollars being processed daily. If you're talking about a Capital One or a Bank of America, they have more data that they're running through their systems. Or this is what you do when you are dealing with global scalability to make sure that nothing goes down, that there is always a disaster uh recovery process in place, that you know, we always have active, active uh what do you call them, like backups? Like there are best practices that you can only see at the organization, like the enterprise organization level, that I would say are pretty cool to see, or at least to understand. And then you realize how difficult it was to actually build these systems, but you understand now that you can build them because you see how they work inside and out. So there's that beauty in having that type of visibility. But the thing I would say to you is that builders need to make sure to overcome the challenges of large organizations by understanding that there's a business purpose to everything, that there are goals that your leader has, that their leader has, that the organization has. The best part is when you're in a company, if you can figure out what that goal is, if there's not enough communication, you probably will have a difficult time. But figure out what matters at the level that you're in. Figure out what matters to your skip level manager and see how your impact can actually align with that. If that's again, if if we want more automate automation, if we want to use more of this like um auditability inside of our system, then we have to take steps to actually apply that. We have to take steps to do that, and all of that is possible if you just think about what the goals are, and so figure out how to reduce the business risks by applying the skills that you have. And sometimes you have to have conversations that you know need to be had with people that are in different parts of the organization because then you learn how to put the business goal together with the implementation uh ability of the people around you.

SPEAKER_00 49:28

What should I look for when reviewing a job offer?

SPEAKER_03 49:34

So keep in mind that a job offer is a package of incentives, constraints, and expectations for you and the work that you're about to put in. And you should understand completely what you are agreeing to before you say yes. So your compensation is not going to just be your salary, right? You get your base salary, but you also get PTO or some type of uh paid time off. You might get sick leave, you might get some other benefit like that. You might get equity, but and that equity could be through um it could be vested or immediately, or there's some other type of vesting schedule, and then you get other benefits like uh health insurance or um, you know, what else could you get? You get health insurance, you could get uh discounts for different products and services. Like there are things that you get from joining a company that all go into how valuable this company or this offer is to you. You should be balancing all of those based on what is important to you whenever you are looking at your offer. So you could look and say, I have a salary offer that's 150k, but you have no equity. And compare that to another offer that's 130k or 125k, but you have a decent amount of equity, so that you can say if this company grows and I expect it to grow, I can make more than what I was expected to make over the next year. So keep those in mind and figure out what's important to you. If you really just want the cash now and you're gonna invest it to something else, you have other investments on the side, whether you invest in crypto, whether you invest in equities or in real estate, whatever it is that you invest in, or even if you have another business that you are uh you know pushing money into. Maybe cash up front is better for you. But if you know that you are going into a company that is in a growing market and you see a competitive advantage inside of that business, maybe you say, hey, I can take a lower base salary for a little bit more upside potential, and then now you got equity inside of a high growth engine, and that's something that everyone doesn't have, right? So figure out the right balance and combination for you, and then go into negotiations with that in mind. Sometimes the offer could include a hybrid schedule versus a remote schedule versus a full-time and office schedule. That's something else to keep in mind in terms of how valuable it is to you, how much is it going to cost you. If you're getting paid more, but you have to show up in office, maybe that's okay. Because now the commute time or the commute cost can be offset by the you know, that increase in salary. Or maybe it's costing you too much to get to work, so that increase in salary kind of goes away because now, hey, it's costing me an extra two hours of my day or $100 in Uber fees or whatever it may be just to get to work. And so that might, you know, if that's daily and you're working five days a week in the office, you gotta figure out does this make sense for you? So keep that in mind. Um, and also remember when you're starting that the job description was just marketing to get you in the door. Now that they made you an offer, you have to make sure that you understand exactly what those expectations are gonna be. If they're gonna be asking you to do something, then you should know what you need to get done and you should be prepared to do that, and it should make sense based on the number that they're offering. If they're offering you a salary to do something that doesn't really align with what everyone else is doing in the market for that same task, then you should be worried. You should be saying something about, hey, I need to either get paid more, or we need to limit the scope of what you're trying to say you want me to do in the next 90 days or six months or whatever it may be. Because you don't want to say yes to something that is a bad deal for yourself. Right? That job offer is not just a paycheck, it is a contract that you're making on how you're going to invest your time and your energy over the next year or so. So you want to make sure that the return on investment, the return on skill and experience is worthwhile. So don't just say yes because you got a yes from them. Do some due diligence.

SPEAKER_00 54:44

What are the best three practices for implementing observability inside your applications?

SPEAKER_03 54:51

So observability isn't just the tools that you use, right? I know we all know uh Datadog, we all know that you can build a dashboard with you know a bunch of different tools that can give you insights, but observability is not just the tool that you use, observability is the processes that you have in place, the resources that you have in place to decide and identify what's broken and why it's broken. So if you are building something that's being used by a million users and you want to know reliably what is happening to this subsection of your users, you need some type of observability system in place. And the system is not just again the tools, you have to have a mindset that allows you to understand this is what we're actually tracking, this is what is important to us in our business, whether it's a the service we're providing or whether it's the product that we are providing. You want to make sure that the observability platform that you create is designed to support the people that are supporting the users of that product. All right. So, first thing is you want to understand what is classified as an event inside of your system. Everything that your API does is not an event. Alright? Let's start there. Everything your API does is not an event, but you can be logging everything your API does. That does not mean you need to create an observability platform that tracks everything your API does at each step, step and layer. You should be bubbling up the most critical alerts from the API for a specific request so that you can say, all right, inside of this API, this is when we got the request, this is when it left, and this is what it left as. And then, you know, what's next, whatever's next. Having a clear way to identify what that event is and classifying it in your mind, your mental model of your system is going to be super helpful because it allows you to have conversations with the clients or your users to say, all right, this is what they're saying happened, and this is what we're seeing in the logs. Okay, this is how we fix this, or this is how we can identify when it happens again later. Right? That's one thing. Two, you want to be able to set up systems to track requests and inside of your system. Most modern day applications are distributed. You're going to have an API that picks up the request, maybe drops off a message to a Kafka topic, and that Kafka topic is being monitored by some other event-driven system to process events as they come through. You want to be able to retract any of those requests from system one to system two to system three and be able to say reliably, all right, this is where it broke inside of the system. And this is where we need to go focus our attention for the next hour or whatever, however long it takes to fix any ongoing problems or even to optimize for something. If you're seeing inconsistent results in this part of the system, then you know, all right, this is where we need to make the solution. This is what we need to fix or optimize. Uh, so having a way to track distributed requests is going to be super helpful, especially when you have different applications inside of your system running, right? You don't want to be looking and having you want to have some type of way of centralizing it, right? So that's why you might use a data dog versus just using CloudWatch uh for your logs. Because you know, CloudWatch, you could technically put all of your logs inside of CloudWatch and it'd be for multiple applications if you just keep the same log group. I don't know if you should. I would say you shouldn't personally. You should use some like centrally managed uh logging platform. It could be Datadog, but it could also be uh uh Dyna not Dynamo. Uh what's what's that? I'm gonna put it in the comments or something. But there are other systems that you you can use, definitely, when you're building dashboards and alerts. Uh and speaking of, three, make sure you build the dashboards and alerts that again make sense for the business that you are running. You don't want to be taking every event and calling it an alert just because something like uh CPU usage got high. Doesn't matter if it got high. Did that CPU the CPU usage result in a thousand other requests being dropped? Did that CPU usage result in the application being restarted and shut down? Like that is actually going to affect users. If the dashboard and the alerts are giving you, you know, telling you to do things that don't actually affect how your users are using the platform, then it's not very valuable to you as a user to be using this platform. So make sure you are giving consistent alerting uh for anything that you're building.

SPEAKER_00 1:00:40

How do I choose between building or buying a solution?

SPEAKER_03 1:00:46

So the most important thing to keep in mind is your organization's priority, right? Your purpose and your mission because that's going to help you make a decision on what is important for you at this current time. So if you are thinking about buying or building a solution, you should make sure that that solution gives your organization a competitive advantage inside of the mission that it has, right? It's not just that it has a business purpose, right? If you need the solution because you're having a business problem, then there's definitely a business case to be made. But when you're deciding between building and buying something, you are making a resource usage or resource consumption problem. You're trying to decide if it's more valuable to dedicate these resources to building something today that may require a spike in resources and a long tail uh support of that application, or if you buy something which requires a low upfront cost, but maybe significant more a more significant long-term uh cost in the future, then you have to like you have to balance that type of problem. And that problem is not something simple. You can't just say, oh, if it costs this much, it costs more today to just buy it, or it costs more today to just build it. You have to make sure that the decision that you make over the long term actually aligns with the purpose and the mission of the business over the long term in that same category that you are, you know, uh considering right now. One thing to keep in mind is that if the solution does not offer a way for your business to differentiate itself, if you don't show up differently to your customers in your market by using this solution, then I would just go with building or buying it. If it's something that your customers are going to look at you differently and either consider you a leader or a loser in that space, if you solve this problem a specific way, then I would consider building it myself. But I would give the caveat that most solutions are going to be better to buy because they're commodities, right? There's no distinction between a database, right? There's no distinction between the infrastructure you use. If you use AWS, you can do the same thing with AWS that you can with GCP, that you can with Azure. You're not going to build your own most companies should not build their own programming languages. They're going to use the industry standard languages to do the things that you need to do, right? So consider rebuilding only only if you see that there, and this this should be directly to the tied to your actual business uh purpose and your actual business advantage if you're gonna be rebuilding something. Because if you're not solving a significant problem with the thing that you are rebuilding, then you're just wasting your time. If you could get 80% of the weight done with a bought solution, get 80% of the weight done, and then figure out how to add on some other value inside of your business to get that last 20% to meet any of your you know extra customer needs. But like your core business solutions should get most of your energy and your time.