Run Your SDLC on Agents Confidently with Zohar Einy of Port.io

**Cleaned Transcript**

(Zohar Einy at 00:00:00) Our agents are actually operating all the SDLC by themselves. They are writing code, they are reviewing, and the developers are managing a team of agents. Building the factory is easy, but to scale it is, you know, it's not that simple. How do we move from a couple of tickets to 1,000 tickets a week that can be automated by AI? And it's a different challenge than just creating the workflow.

(Zohar Einy at 00:00:23) This is what separates, you know, mature thinking about AI versus cool demos that look great in a video but do not scale for the organization.

(Joel Beasley at 00:00:39) I'm just a curious person, and I've got questions. And so we'll just start at the beginning with, what is Port.io?

(Zohar Einy at 00:00:49) So Port is a platform to build the AI software factory in a governed way and in a way that can scale for the organization.

(Joel Beasley at 00:00:58) So what's an AI software factory?

(Zohar Einy at 00:01:00) Yeah. So now we are moving from the AI-assisted era where, you know, developers would write code and AI would assist them, into the AI-led era where agents are actually operating all the SDLC by themselves. They're writing code. They are reviewing. They are managing incidents.

(Zohar Einy at 00:01:18) They are fixing vulnerabilities, and the developers are managing a team of agents. So an AI software factory is like an actual factory where agents are producing features and doing everything that the engineering used to do.

(Joel Beasley at 00:01:35) Is this something that's happening right now in production?

(Zohar Einy at 00:01:38) Yeah. Many companies already started to scale their AI software factories, and they are able to delegate 80 to 90% of their work to agents. But it's still very, very early days, and very few companies manage to do it at scale. And this is something that, you can... You know, X is all about it, and everyone is talking about it.

(Zohar Einy at 00:02:00) And Gartner recently released a report about AI software factories and how do you get it right and what you should look for. So it's an emerging market, and we believe it's going to be one of the biggest markets for AI in engineering.

(Joel Beasley at 00:02:15) Was that the reason why you founded the company?

(Zohar Einy at 00:02:17) So we started the company four and a half years ago. We started off by creating a developer portal to serve humans, giving developers everything that they need to provision infrastructure and query data, like, you know, who owns this service, where is it deployed, putting guardrails for developers so they can be autonomous without deviating from the company's standards, like production readiness of services. And this is how we started. And we found out that what we did for developers, the unified interface, they can do everything in a governed and a scalable way, is exactly what agents need as they operate in the software factory.

(Zohar Einy at 00:02:58) Because you can just let agents run and connect them to everything via MCP and expect, you know, you will sleep well at night. You need a way to give them, you know, the right context and the right guardrails and the right actions and visibility to management to understand, you know, what's in it for me? What's the ROI for the software factory? So our product evolved into serving agents because they need exactly what humans needed just a second ago.

(Joel Beasley at 00:03:25) So it's kind of like an agent orchestration system.

(Zohar Einy at 00:03:29) So orchestration is one piece of the software factory. So you need a couple of things for running a software factory at scale. First of all, you need the workers. Right? You need agents.

(Zohar Einy at 00:03:40) You already have those if you are paying the Frontier Labs. So you have the agents, but now to make it scalable, you need a couple of other pillars. You need a context lake. So you need to give the workers a way to query high-quality data because if you connect them to all the different tools, usually the data is scattered. It's outdated, and it takes a lot of reasoning power to get answers.

(Zohar Einy at 00:04:03) So context lake is a unified pillar that they can reason about effectively and get high precision. So this is the first element of the factory. Then you have the registry of agents, skills, and plugins, and MCPs. So, you know, to really scale a software factory, we cannot operate in a skills pro and agents pro because every developer is, you know, they're creating their own skills on their own machines. They're duplicating stuff.

(Zohar Einy at 00:04:31) It's not meeting the company's standards. They're building their own agents, and they're running them wherever. So the second pillar that we do is providing an agent registry, MCP registry, skills registry. So you can certify everything that you build in the factory.

(Zohar Einy at 00:04:49) You can assign an owner, so you can create accountability for the organization. And you have audit log for what those agent skills and MCPs are used for. And there is a compounding effect between agent skills and MCP registry to the context lake because now you've managed the identities of the agents and the data in one place. So managing permissions between agents and data becomes very, very scalable, right, to do it...

(Zohar Einy at 00:05:18) To keep up with security standards in the organization. There are a few more pillars to that. There is the pillar of the human in the loop. So we cannot put just, you know, human in the loop on everything. When you go from a POC to a full scale, you can...

(Zohar Einy at 00:05:34) You need to put human in the loop based on risk, based on business priority, and you need the most scalable way to review changes when agents are running across the SDLC. And the element of the measurement of the factory itself, so measuring the ROI, measuring areas that are not efficient that maybe I should optimize for, maybe areas that are not high quality enough and the change failure rate is too high for those workflows. So how do leadership... What does leadership need to measure and to understand what's going on in the factories? Another big one for when you scale.

(Zohar Einy at 00:06:11) And lastly is the governance piece, which is critical to put the guardrails for the agents to make sure that they are not deploying on a Friday. You have full audit log for what they can do or can't do to make sure that you have the right permission model and the policies in place. So all of those things are critical to scale your AI software factory. So agents are not enough. There is a lot more around them that needs to be built.

(Joel Beasley at 00:06:36) That is wild. I'm learning so much. So since... Just for some background for you, twenty... I'm twenty years as a software engineer.

(Joel Beasley at 00:06:44) And for the past several years, I've been doing the podcast. I still maintain a handful of production applications. But at the current point, I am the sole engineer on those, so I'm not collaborating across a large organization. But I am seeing the need for this. Like, for example, I set up a lot of my projects in Cursor, and I found out really quickly that I need to create markdown files for very specific things.

(Joel Beasley at 00:07:11) Because as you mentioned, it takes a lot of resources for it to scan all my sales materials every time I want to do it and come to a conclusion. So what I did was I sat down. I had it scan my sales resources, put all the information into the markdown, then respond to clients and have it help me do this. And then I trained it along the way. So, now I've got this master markdown file for how we respond to sales questions, which I can take anywhere.

(Joel Beasley at 00:07:41) But that's just me in one file in a GitHub. You scale that to an organization where you got multiple... It just means complete chaos without some sort of system like this to organize everything. Are people building their own, or are they just finding you?

(Zohar Einy at 00:07:54) Yeah. Many people are building by... So what we see in the organizations is that developers are not waiting. They are, you know, they have AI. They have knowledge, and they start building and automating all kinds of workflows and skills across the SDLC.

(Zohar Einy at 00:08:10) So companies are... Developers are already starting, but usually when they try to run it at scale, this is where they start to face problems. So everyone understands that this is the North Star, and everyone is trying to build towards that. But there is a question to the industry, like, why are we not there yet? Like, what what is...

(Zohar Einy at 00:08:28) What are we missing? Why are we not moving, like, in an unprecedented velocity? The reason is to do it at scale, just like any engineering project, it doesn't matter if it's AI or whatever we did in the past. Security, governance, scalability, change management, compliance still apply.

(Zohar Einy at 00:08:48) And if you don't take care of them at the beginning, if you do it after the fact, usually this is the reason why we see those projects are failing and taking a step back and saying, okay. How do we move from a couple of tickets to 1,000 tickets a week that can be automated by AI? And it's a different challenge than just creating the workflow.

(Joel Beasley at 00:09:08) Has OpenAI published anything about the architecture for these AI factories?

(Zohar Einy at 00:09:14) Yeah. So recently, the... Actually, the Pragmatic Engineer also released an interesting piece about it, and they released their reference architecture for building a software factory. And, also, Anthropic released a playbook on how to build the AI-native SDLC, which is an equivalent term for AI software factory. And they describe their...

(Zohar Einy at 00:09:38) The North Star. Right? What would ideal engineering organization would look like in the AI era. But to get there... And I think that there is no doubt, like, this is how engineering would operate.

(Zohar Einy at 00:09:51) But I think that there is still a big question in the industry is how do you do it at scale? And from what we are seeing in the market, for example, from companies like Uber recently released how they do it at scale. So the context lake, the guardrails, the skill registry, all those things are the ingredients to run it at scale. Because building the factory is easy. Building what OpenAI released, the architecture is easy.

(Zohar Einy at 00:10:18) You can do it in two minutes. But to scale it, it's not that simple.

(Joel Beasley at 00:10:22) So what separates the companies that scale it successfully versus failing?

(Zohar Einy at 00:10:26) First of all, I think it's the mindset of... The fact that you have AI and you can build things fast doesn't mean that you should start by building right away. You should wait a second and do some planning before that. And I think that companies that manage to do it at scale spend time thinking about what does it take to do it at scale. What are the checklists that we need to put in front of us to make sure that we can run agents across the SDLC and automate thousands of tickets?

(Zohar Einy at 00:10:54) So questions like, how do I manage access to the data? Like, how do I control what agents can or can't see? Or how do I make sure that they're using certified skills? How do I make sure that developers are not duplicating and creating more and more skills on their own machines? Right?

(Zohar Einy at 00:11:08) How do I give... You know, how do I even teach the developers how to do their work differently from tomorrow? It's a change management question in the organization. So spending time asking those questions that you will face after an initial rollout is what differentiates companies from successful software factories and software factories that you have to shut down and restart. So spending time thinking about those questions.

(Zohar Einy at 00:11:38) I know it's not the most interesting questions. It's not the fun questions because you want to go and implement. But this is what separates, you know, mature thinking about AI versus cool demos that look great, you know, in a video but do not scale for the organization.

(Joel Beasley at 00:11:56) And so this platform, when you're bringing people on, where are they at? Are they at, I've rolled my own, it's complete chaos, or have they not even tried? Like, where are they at when they find you?

(Zohar Einy at 00:12:08) So there are a couple of phases. So the first phase is companies that are running AI coding assistance only, using Cursor, Claude, GitHub Copilot to write code faster. And those companies are, you know, many companies are in this phase, but now we are seeing the next phase is companies that are starting to build, let's call it, local agents on their machines, the RCA agent, the security agents. So teams are starting to build their own agents, their own skills, but they're still running it on their own machine. It's still not running as a centralized factory.

(Zohar Einy at 00:12:48) It's like a decentralized factory running on their... On the developer's machines. And then we see the customers and the prospects that are saying, like, okay. We cannot operate with local factories on every machine of the developer.

(Zohar Einy at 00:13:01) We need everyone to work on one unified place because we need to build one factory. Otherwise, we are risking the organization. So we see AI assist... Fully AI-assisted, then some kind of a hybrid mode with local agents running on developers' machines taking more end-to-end tasks and companies that are running a centralized cloud-based factory that is unified for everyone to collaborate on.

(Joel Beasley at 00:13:26) Help me understand this. What's an example of a skill that would be on someone's local machine that they would then certify and make it available to the factory as a whole? What's an example of one skill?

(Zohar Einy at 00:13:41) Yeah. So maybe a skill on how to write a Jira ticket in the organization. So...

(Joel Beasley at 00:13:46) Perfect.

(Zohar Einy at 00:13:46) Yeah. So there are a couple of ways to write a Jira ticket, and you want it to be according to your standards. So when developers are just writing their own skills or not even, like, you know what? Even not writing a skill for it is a problem because then agents just, you know, do it the way that they think. Or there is a skill, but it's not certified by the organization, so the format is not following the company's standards. So this kind of a skill is mandatory for you as an engineering leader to make sure that you sign off this skill because everyone is using this skill every day, and you want to dictate and control the standards of how Jira tickets are being written in your organization.

(Joel Beasley at 00:14:32) Interesting. Is there an argument for me wanting a customized variation of that skill? Are we not getting there right now? We're just... You're making the skills.

(Joel Beasley at 00:14:43) Someone in the org is certifying them, and then other people can then look at that index and find that skill and then use it.

(Zohar Einy at 00:14:51) I think that there will be versions of... Or branches of this kind of skill because, you know, for large enterprises with thousands or tens of thousands of developers, sometimes they run different business units. So it's important that the way that you strategize about how you build your skills marketplace or your agent marketplace should fit your needs as a company, your DNA, your business requirements, and because every company is a snowflake. Right? So I think that there is a case for having different types of Jira ticket formatter or writer skill, but it really depends on the way that you operate.

(Joel Beasley at 00:15:33) Okay. Cool. So you... You're more of a, here's the infrastructure. Here's the pillars.

(Joel Beasley at 00:15:38) Here's the core tools you're going to need. You also provide some insight because there are those public papers from the large companies on how they do it. But then you let the company essentially build its own with your infrastructure.

(Zohar Einy at 00:15:52) Exactly. We understand that every company is different. They have different architecture. You know, they're running microservices, monolith, monolith and microservices, mono-cloud, multiple clouds, you know, migrating from GitHub to GitLab or vice versa.

(Zohar Einy at 00:16:10) You know, they have different processes. They have different culture. Dictating the way that the software factory would look like would probably break because every company is operating in a different way. So we give the builders the infrastructure to build their AI software factory in a way that is scalable.

(Zohar Einy at 00:16:30) So this is the way that we strategize about our product.

(Joel Beasley at 00:16:33) Nice. Kind of reminds me of Ruby on Rails. Right? Here's the standard. Here's all the components.

(Joel Beasley at 00:16:38) You bet. You can do anything you want with it, but there is a preference or an opinion on how it could be laid out, which is great because it allows new people to learn, and we're all trying to learn this new structure as well.

(Zohar Einy at 00:16:50) Yeah. It's also... You know, we don't know what's coming next, and it needs to be future proof. And you don't know... You want to be able to change in the future.

(Zohar Einy at 00:16:58) If tomorrow we have a new something that is coming up, probably will come tomorrow based on what we see and the pace that we see. You want to be able to be adaptive to those changes.

(Joel Beasley at 00:17:09) How's leadership taking this? Because, you know, the technology already had its budget, already had its SaaS and its cloud expenses. And now they're coming back and saying, okay. Now, are they seeing it as, well, obviously, this is impacting our entire life, and so the budgets are open, or is it difficult?

(Zohar Einy at 00:17:29) In the beginning, we saw companies that are just like... They had a token leaderboard. Well, you know, the more you spend, the better. We are using AI. Like, you know, we spend so much money.

(Zohar Einy at 00:17:40) Then they said, like, okay. Explain to me what is the value, because I got the bill. So I need to understand what does it mean, and I need to give some answers to the board. Right? And then people took a step back and said, like, okay. Let's think how we challenge the teams to demonstrate the ROI and show us how they become more efficient. I think it would be very similar to the way that we operate with the cloud. Right? Like, we have access to AWS, to GCP. You know, if you don't use it the right way, you will pay high bills, and you should think before you build.

(Joel Beasley at 00:18:15) Do you and your tool provide any reporting related to ROI for leadership?

(Zohar Einy at 00:18:22) Yeah. So I think that the ROI is the highest level metric, right, to demonstrate. But I think that to have some proxy KPIs that will help you understand if you're on the right track. So we either recommend to demonstrate efficiency, like, how efficient is my workflow, and what do I need to do in order to optimize the cost for it. For example, maybe the context window, the input tokens are too high for my agents. Do they need all these input tokens?

(Zohar Einy at 00:18:54) You know, do they need all this context to achieve the same goal? Or do I need to maybe use a lower model? I don't need the Opus 5.5. Maybe I can use Sonnet. Right?

(Zohar Einy at 00:19:07) So those kind of things to find optimization options of the agents that are running in their SDLCs, also things that you can measure and you should measure. And then at the end of the day, you want to also measure the change failure rate of agentic workflows. You know, how many deployments made by AI really sustained and really gave the... Were able to operate in production, and we didn't have to roll back those deployments. And then, you know, and then you get to the ROI question of, okay, how many new features did I deliver to the business? What did I pay for it? And, you know, do I see higher and higher gains over time? So I would say it's measuring velocity, efficiency, quality are the three things that I would focus on to demonstrate the ROI question.

(Joel Beasley at 00:20:00) Help me understand agents handling things like self healing. What is that?

(Zohar Einy at 00:20:05) Yeah. So self healing is a term describing the fact that, you know, you have the PagerDuty calling you overnight, and then you turn it up. Right? You wake up, someone calls you, and then hangs up right away. So this is the future of incident management the way that we see it, but not just for incidents in terms of the production downtime.

(Zohar Einy at 00:20:29) We believe that self healing will come into play with CVEs, with vulnerabilities that will self patch themselves. We believe that CI/CD pipelines will be self remediated. So one of the most annoying things as a developer is when your CI/CD pipelines fail, and then DevOps just throw a bunch of logs at you. You don't understand what they are. Maybe you have AI, but you still... Yeah. You just pass it on to DevOps to fix it. And then, you know, and then they are getting angry because they need to fix so many issues, and they have so many tickets. So self healing CI/CD pipelines is a big one. We are seeing this use case coming up a lot to free developers' time and to be able to run faster because sometimes the CI/CD fails because the developer did the misconfiguration, and sometimes there is a bug in the CI/CD pipeline, and you don't want to deal with it. Like, you want self healing CI/CD, self patched vulnerabilities, self healing incidents, and so on. So self healing is a big, big umbrella of issues that you want to be taken care of for you.

(Joel Beasley at 00:21:33) Last week, someone texted me a video of David Heinemeier Hansson, the creator of Rails, him saying that they're now doing pencils down. So that's what he's calling the trend. I don't know if a lot of people have heard that phrase, but it's referring to this trend of humans no longer write code. So I've interviewed Trevor at Soundstripe, DHH, a couple other people.

(Joel Beasley at 00:22:00) They've all have this trend that they're requiring their engineering teams, which are of significant size. It's not like three people. Their entire engineering orgs to not be writing code anymore. Are you seeing this as well?

(Zohar Einy at 00:22:14) Actually, we are seeing some companies, like, doing a no IDE week. Like, you are not using an IDE for one week as a practice to... You know, people start to say, okay. Okay. So what do I do? I can't open the IDE. I need to work just with Claude and such. So, it's a good practice to get developers rehab from looking at the code.

(Joel Beasley at 00:22:39) It's hard. I'm a fan of Cursor. I use that personally. And they have slowly weaned you out of the IDE into the chat with the agents. They've done a pretty good job of it because I'm reluctant, but they're pulling me along in small increments, and it tends to be working. I've gotten... It was so hard to let go, Zohar. Like, I'm so used to opening the files, reviewing the code, seeing... running the functions with my eyes and everything. To... Have you ever done a blind deploy? It is wild. I built a hobby app just to do it, and I was blown away by how good it is now.

(Zohar Einy at 00:23:19) Yeah. Yeah. Yeah. I think that we need to, like, you know, to start and feel more comfortable with it. But I think that the way to get... to regain the control and the trust is to make sure that when you are needed, you will be in the loop. Right? But when it's, you know, when it's high confidence deploy with low risk, if it's not touching, like, a P0 service. So it's... maybe it's fine.

(Zohar Einy at 00:23:46) Right? Because the cost of mistake is low. And if it does, you should build the right measures to make sure that if it's touching a business critical piece of code or a sensitive asset to the organization, like a customer or a product or... so then you want to drill down into that deeper. So...

(Joel Beasley at 00:24:09) Question. Earlier, you mentioned... We're talking about governance, and you gave an example of agents that can't deploy on Fridays. So I have noticed that when I've given specific commands, sometimes it ignores them. How do you make sure that if the rule is... are you still at the mercy of the model, or does your platform have something else that will further enforce the rule?

(Zohar Einy at 00:24:35) So it's both. Like, we would... We realized that we cannot depend on the model to, you know, to listen to the rules that are nondeterministic. So there are also, like, deterministic rules that no matter what will be applied. So the agent cannot operate beyond those rules. So it's important because, again, like, you want predictability.

(Zohar Einy at 00:25:02) You want to have assurance that specific policies are taking place. So, like, for example, policies can be PII data access or deploying on a Friday or using specific skills or blacklist specific access to resources in the cloud. Those kind of things, you want deterministically defined. But, also, it's easier to define them in a deterministic way. I think that writing, you know, text that says don't do that, don't do this, like, it's not as effective as, you know, I have, like, a deterministic rule policy that is being applied and evaluated, and they want to give proper, you know, proper guidelines to... and leave no place for opinion.

(Joel Beasley at 00:25:56) For people who want to learn more about these systems, these AI factories, where... Do you publish information on your side about that, or do you write about that?

(Zohar Einy at 00:26:07) Yeah. So we've recently... We took the Anthropic Playbook, the entire playbook that they published, and we implemented live on our YouTube. So this is actually an interesting one, like, how to take the theory and put it into practice. So we write a lot about it.

(Zohar Einy at 00:26:22) We have a bunch of community sessions that we are inviting guests from, from Cursor, from GitHub, from from a bunch of Frontier Labs. So, yeah, like, we have a lot of content on our YouTube, on our blogs, and we talk a lot about it just because we are excited about the topic.

(Joel Beasley at 00:26:41) Well, we'll make a producer note for David to find that specific YouTube video and put it in the show notes below because that's fascinating. It's great to have the article or the content come out from a big player who's doing it at scale, but it's even better to have someone sit next to you and walk you through it actually being applied in real time. And that's what you guys are doing. Right?

(Zohar Einy at 00:27:08) Exactly.

(Joel Beasley at 00:27:09) That's so cool. You guys... It's almost like you're a modern company.

(Zohar Einy at 00:27:14) Yeah.

(Joel Beasley at 00:27:15) Yeah. Do you have an example of a company that's actually implemented the systems, and they've had success and seen it pay off?

(Zohar Einy at 00:27:26) Yeah. So GitHub is one of our customers. They're using Port as the foundations for their factory. So there is, you know, an entire study that we've put together. And, also, we have another few on our website that they're describing their story, how they were able to delegate almost all their tickets and why they couldn't do it without Port.

(Zohar Einy at 00:27:49) And, yeah, we... I believe that the very few companies that were able to scale their software factory are Port's customers, and we are very excited to be the vehicle for a fully autonomous SDLC.

(Joel Beasley at 00:28:03) That is so... You must be excited. Are you the founder or just the CEO?

(Zohar Einy at 00:28:07) I'm the founder and the co... cofounder and CEO.

(Joel Beasley at 00:28:09) Dude, how excited are you? You're living the dream. You had the idea. You brought it to market at the perfect time. It's got the tidal wave. You're surfing at the top. Do you feel good about life right now?

(Zohar Einy at 00:28:22) Yeah. It's a, you know, it's a dream. Right? Like, you stand in front of, you know, in front of all the company. We grew from, you know, like, just a few people to almost 300 people today, you know, from a small garage where, you know, we had an all hands with just a few people around there. After we finished eating lunch, we do the all hands.

(Zohar Einy at 00:28:44) And now I have a camera here, here, microphone, prompter. Like, it's becoming a production. Right? Like, you... And this experience is happening super fast, and it's fun.

(Joel Beasley at 00:28:54) How did you meet your cofounder?

(Zohar Einy at 00:28:56) We met each other in the intelligence unit, in the 8200 unit. He was part of the DevOps and infrastructure team, and I was part of the develop... the R&D team. And then, actually, we got to know each other because I had so many requests from DevOps, and I needed so many infrastructure related things. So I got to know him because he would handle those requests.

(Zohar Einy at 00:29:19) And, yeah, this is how we got to know each other.

(Joel Beasley at 00:29:22) And what military is that?

(Zohar Einy at 00:29:23) It's the IDF.

(Joel Beasley at 00:29:25) The IDF. Oh, this is great. Okay. So there's a lot of tech startups coming out of Israel, and I've gotten to interview several of them. But I found out... I've been doing this show, Zohar, for about ten years, and I found out... Right. I interviewed my first Israeli startup maybe three or four years into it. So let's say six, seven years ago. And the individual who I got to interview was so direct.

(Joel Beasley at 00:29:51) And then I met some of his friends and we did podcast. And I realized that culturally, there's this amazing communication. I was raised like that, very direct communication. But then when I went into the business world and started hiring people and growing teams, I realized in the United States, it is not like that.

(Zohar Einy at 00:30:09) Yeah. Correct.

(Joel Beasley at 00:30:11) They get upset. It's offensive. And so I had to learn. So whenever I meet people that have that same communication style, I'm always like, yes. It's good to know you, Zohar.

(Zohar Einy at 00:30:23) Yes. It saves time. If you are direct, it saves time.

(Joel Beasley at 00:30:26) It does.

(Zohar Einy at 00:30:26) I think that directness is an important quality and something that we look for for people that are joining Port, whether they are from anywhere. Right? Like, I think that good people are direct because they have enough, you know, confidence to... And, you know, self confident and respect to the other side to be direct and tell them what they think. So I find it as a great quality and something that can help us move very fast as a company.

(Joel Beasley at 00:30:57) You're exactly... It's selfless. It's saying this outcome that we're trying to achieve together is so important that we just need to be as efficient as possible to achieve that outcome. I can be wrong. You can be right.

(Joel Beasley at 00:31:10) It doesn't matter. What we're trying to do is get to the truth and achieve the outcome.

(Zohar Einy at 00:31:14) Correct.

(Joel Beasley at 00:31:15) Yeah. I love it.

(Zohar Einy at 00:31:16) It's something that I found out as different in different cultures and... Yeah. It is different.

(Joel Beasley at 00:31:25) Yeah. There is some cult... Oh, yeah. We won't get it. We won't look at every single culture.

(Joel Beasley at 00:31:29) Let's talk more about Port. For people who have been listening to the conversation and they want to improve their agentic SDLC within their organizations, where do they go? What's the first step they should take?

(Zohar Einy at 00:31:42) Yeah. So first of all, I would start with building the context lake. And this is the first place that I would go to model the different assets inside the semantic layer. So think about your microservices, deployments, environments, vulnerabilities, resources. Start with those things, those key elements with ownership of who owns what so the agents can know where they live and how you speak in your engineering organization, because they are very strong, but they don't read minds. And they need you to tell them how you articulate... What do you mean when you say service? So it's important to give them... To infuse the semantic layer into their brains, if you will.

(Zohar Einy at 00:32:26) So context check is the place to start, and then I would start... I would go from there to build the agent registry, skill registry, and go from there all the way to orchestration, human in the loop, and measurement.

(Joel Beasley at 00:32:37) At what point do we install the Neuralink?

(Zohar Einy at 00:32:41) It's a... This is the next episode, we will find out. Things move so fast.

(Joel Beasley at 00:32:46) Yeah. I want to get those context lakes worked out before we start doing the Neuralink stuff.

(Zohar Einy at 00:32:51) Correct. Yeah. We have some time, like, a week. Maybe two. Maybe two.

(Joel Beasley at 00:32:57) I noticed you have... And I don't know if we're allowed to talk about it. I'm going to say it, and if we can't, we can edit. But there's PortCon. This is a conference you guys are putting on in San Francisco, November 17th and 18th of 2026.

(Joel Beasley at 00:33:11) You've got some great speakers. You're having GitHub, all sorts of interesting people. Two days, what are you excited about?

(Zohar Einy at 00:33:20) PortCon is the first AI soft factory conference in the world that is in person that we are really excited about. We are going to have leaders from GitHub, OpenAI, Anthropic, Expedia, 1Password, NVIDIA, really like the best companies that are going to share how they are building their AI software factory and what does it take to make it scalable. And yeah, it's going to be a great event in Shack 15, which is a very nice and cool place, and we are really, really excited about it.

(Joel Beasley at 00:33:58) Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague who you think would get value from it. And if you have topics that you'd like to hear discussed on the podcast, either add me on LinkedIn or send me an email, [email protected]. Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.