Coding agent reliability isn't a model problem, it's a gates problem. Ran Aroussi, creator of yfinance and founder of Automaze, argues that an "agentic workflow" is a contradiction in terms, and that the only thing standing between you and full autonomy is deciding where a human still signs off. He also thinks you stopped needing junior developers for their coding skills about a year ago, and that hiring them anyway is the only way the industry gets its next generation of architects. What we cover: - Why "agentic workflow" is a contradiction, and when agentic coding is the wrong tool for the job - Where a human still has to sign off, and what coding agent reliability actually depends on - What a software factory changed about delivery times, pricing and the client backlog - Why you no longer need juniors for their coding skills, and why you should hire them anyway - Teachable knowledge vs earned knowledge, and the experience agents cannot compress - The law firm model: why dev agencies may end up looking more like Kirkland & Ellis than a SaaS startup Chapters: 00:00:00 - Introduction 00:03:05 - yfinance, open source and 30 million downloads a month 00:06:03 - Automaze and the MUXI agent application server 00:10:42 - Where to draw the line on agent autonomy 00:15:12 - Software factories and fully autonomous merges 00:22:25 - The architect archetype and the four-phase ladder 00:28:11 - Teachable knowledge vs earned knowledge 00:30:14 - What agents did to delivery times and pricing 00:37:03 - Why "agentic workflow" is an oxymoron 00:41:08 - Anthropic, OpenClaw and the end of flat subscriptions Build your software factory, one workflow at a time, with Tessl: https://tessl.co/vut 🔔 Subscribe for weekly episodes on AI-native development Where's your gate? Tell us in the comments where you still refuse to let a coding agent merge without a human looking.
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
SPEAKER_02For at least a year now, you do not need to hire junior developers for their coding skills. If we stop hiring junior developers, we will never have this archetype of an architect that we are looking for. Both a product manager, a project manager, a team lead where the team is agents in that case. The real skill that we're trying to build is the ability to not multitask but multi-monitor agents. So you'll be able to command a team of agents. And I'm not in under any sort of uh fantasy that you can manage thousands of agents, but you can manage six to ten, maybe.
SPEAKER_01The AI Native Dev is a podcast for developers and engineering leads at the cutting edge of AI and agentic coding. Join your host, Guy Fajani, and meet Simon Maple every week as we chat with the most exciting voices in AI and tackle the biggest questions facing developers today. This is the AI Native Dev. Yesterday, you were writing code. Today, you're prompting agents. Tomorrow, you'll be running a software factory. Agentic coding is moving fast enough to give anyone whiplash.
SPEAKER_01This June, AI DevCon London sold out. Hundreds of developers, engineering leads, and VPs packed in to hear speakers from Anthropic, OpenAI, Netflix, and GitHub cut through the hype and get to what actually works. This November, we're bringing it back to New York. Bigger, bolder, and full of people actually building the agentic future. Three days of real talks, real tools, real answers, whether you're directing the agents, building the loops themselves, leading the teams, or setting the strategy.
SPEAKER_01Our early bird pricing ends on September 30th, so make sure you lock in your place before prices increase. As a podcast listener, you can get 15% off your ticket using promo code POD15. We have so much planned, and we can't wait to have you join us in New York. Hello and welcome to another episode of the AI Native Dev. My name is Simon Mapler. I'm your host for this episode, and joining me is Ran Arusi, the founder of Automaze and MUXI.
SPEAKER_01On today's episode, we're going to be covering major topics such as the difference between teachable knowledge and earned knowledge. And where junior developers today lack that earned knowledge, how can they gain that information? Also, we're going to be talking about the agent workflow. And is the term agent workflow an oxymoron in itself? How do we get that determinism to remain in a workflow? And finally, uh, we're going to talk about the law firm model and how that applies to software development agencies. Do we need to think more about that law firm model of distribution of wealth, or should we stick to that SaaS model that we're familiar with?
SPEAKER_01Welcome to the episode, Ran. It's a pleasure to have you. Thank you very much for having me. And Ran is the founder of Automase and Muxie and uh the creator of Wi Finance. Tell us a little bit about some of those things.
SPEAKER_02Yeah, so um my developer journey started 35 years ago, uh, I'm sad to say. And uh yeah, throughout that time I've been uh working pretty much in any tech adjacent industry as a team lead, CTO, CEO. And um my real love uh journey with uh open source started uh in the late late 90s, where I realized that one of my startups uh that we had to raise uh millions of dollars for just to uh have a database uh can now happen for free.
SPEAKER_02So every time that since then, every time that I developed something that is not like my business secret sauce, I happily uh turned it into an open source library. One of those uh indeed was uh Y Finance, and uh it just took off. I mean, it's uh it's quite amazing. Uh it all started when um I I'm still doing that, but I used to be a lot more invested into algorithmic trading, and um, I've done this uh for a living for a while, and actually with great results, I was very happy about that.
SPEAKER_02And I needed a data source that was free, available, and very Pythonic, uh, so I can use it for downloading market data, and I just uh uh decided to leech on Wi-Fi um on Yao Finance. And I created this library and uh it really took off. And these days I'm getting, I think the latest number is about 25 million to 30 million downloads every month, which is insane. And um I remember just a few years ago I boasted about having 600,000 downloads every month. So um it's it's quite insane, and that this number keeps accumulating because it turns out that every time you ask AI to write anything related to um market data, it uses that.
SPEAKER_02Uh so I assume that is part of the the real growth over the past year. And wasn't there a time when actually Yahoo dropped their API for that? They never dropped their API. So the the way the library works is not by scraping or it's using the official um Yahoo Finance API. Uh it's just that they're trying really hard not to document it anywhere, but it is publicly available. Um I actually got um um a CNDC that uh after I pointed them to hey, this is a public API, they said okay, sorry, we're and and they they backed off.
SPEAKER_02Uh so I assume they weren't even hired by Y uh by Yahoo. Uh they were just trying to land them as a client or something. Uh but yeah, this is definitely not the only library that does this. Uh I I just think that it it's doing it really well and that the interface is um API-wise, is very uh Pythonic and very simple, which is something that I really uh believe in. Uh let's cut the number of uh code lines that you need to have. And um yeah, um over the past I don't know, four or five years, I'm uh running a software development company called Automase, as you mentioned, and um we tend to kind of uh do now it's called uh uh FDEs, but we tend to embed ourselves into startups and different organizations with our development team.
SPEAKER_02Uh I act as uh kind of uh the uh the the the spirit of the company. I help uh with the specing of the products because my experience uh uh over the past few decades is uh I have been both in product and in tech. So I'm kind of walk and untalk the the two uh languages. And uh I have a great team that's scattered all around Europe that does the actual development. And uh one of uh my most recent pet projects that turned into a behemoth of code base, although it's not nearly as used as what as Wi-Finance is as MUX is you mentioned, which is um a different take on how I think AI should be deployed, which is having agents as primitive give them their own um application server, and um make sure that you don't have to code your way into having an agentic platform, but just deploy.
SPEAKER_02So it's kind of like a Docker uh in terms of how you deploy, it's like um Maxi pull, push, you pull a formation which is in the entire setup.
SPEAKER_01So yeah, a formation a formation is a unit of is like a collection of agents, is that correct? Yes, right.
SPEAKER_02Yeah, so at the top of the a formation is is it's not just a collection of agents, but it also includes what I call the overlord, kind of the orchestrator that decides of the formation's personality, and uh the agents themselves are sort of minions because I didn't want the experience of having to remember which agent I need to talk to. This one does email, this one's a good writer, and I I just wanted to talk to Jarvis and have uh and have it manage whatever agents it it needs.
SPEAKER_02So um uh affirmation uh basically tells me okay, which models am I going to use, what are the the backup whether or not I have role-based uh access management, what type of uh memory allowance do I give the database? Um is this an organizational deployment where I'm gonna use Postgres with uh all uh different types of extension, or just for me and I'm just gonna use uh MySQL? And it's all configuration files. Um so there's no code involved.
SPEAKER_02Um everything is in YAML files, uh in the format that I call the agent formation schema. Uh it's again fully open source, uh available in agentformations.org. And um yeah, it allows you to define both the obviously the soul of the document, what it can do, the tools, uh the domain knowledge for every agent, uh, access to uh it has a very elaborate memory um uh structure where it has uh it differentiates between uh preferences, memories, decisions, claims, working memory, buffer memory, uh episodic memory, recent memory.
SPEAKER_02So it's a very layered type of memory because what I'm trying to do is mimic how the uh I'm re I'm really uh uh philosophic philosophical here because what I'm trying to do is mimic how the human brain works when it comes to memories. Uh there's the misconception where uh observation and logging is memory. I don't think so. I think the distillation of um of knowledge and uh recall is the memory because we might not remember every single word that we that was said here, but we're gonna have the takeaways, we're gonna remember conclusions and things that we've agreed on.
SPEAKER_02And I think that's how an agentic memory needs to work. So it's all bundled in, you don't have to worry about it, you don't have to um, you just need to essentially configure what your uh agentic AI can do. Which tools do you want to remove from the registry? You know, you're using uh Stripe MCP with uh million tools over there, and you only need three, so there's no reason to hog the memory. Just tell okay, these are the only two or three tools that I actually use in this application. It's not gonna load any of the other ones into memory.
SPEAKER_02Uh so yeah, it's uh it's it's very well thought out based on the needs that I've encountered both in my own own personal kind of Jarvis uh journey and what I noticed with uh with organizations that uh we've deployed we actually created a separate company in the US for deployment uh deploying agentic um um AI at the organizational level. So yeah, that's in a gist.
SPEAKER_01And what one of the interesting things there is the the kind of like operating instructions, as it were, for for for a uh for a formation. Um I believe that that can that the the they can kind of like retrospectively look at those instructions and say actually I want to change I want to change my own instructions based on historic kind of like how things have been done. Um one of the key things though there is I believe that requires a human interaction to say actually yeah I do approve this or I don't approve that.
SPEAKER_01Um wh where do you kind of like st feel today we should draw the line between pure autonomy where things can essentially uh self-heal, fix, uh tune themselves, versus requiring uh us to still provide an overview. And is that line more a technical line today? Or is that more of a a trust or culture line that we need to draw for ourselves?
SPEAKER_02Uh I actually think uh that it's more of a trust line. I think that technically you can create a fully autonomous uh system, and I I was actually talking about this yesterday. Um I think that you know, with AI it can hallucinate. So the first uh line of defense is make sure that it has up-to-date information, it that it knows how to not only store it but retrieve the right uh information, which is not on always just contextual, it it needs to have the the graph of relationships there.
SPEAKER_02Uh but even then uh let's take a coding um agent for example. So you're giving it a goal, and if you're not going to intervene at any point and say, okay, after five tries, after two hours, whatever it is that you set as a threshold, come back to me and say, I can't do it, and let me let me help you uh get unstuck. If you don't do that, agents will either um create slop or run uh very high token bill because it's just gonna keep trying, or they're gonna fake the tests uh just to to to have the test passed.
SPEAKER_02Get the test screen quick. By any means. And we've all seen uh code with the comments by the AI, this needs to be changed later to dot dot dot. And um in order to have that trust, I think that the human still needs to be in the planning and in the confirmation uh stage. So uh you need to what is the definition of done? What am I trying to achieve? And uh you need to be there in these these two steps, in the specing.
SPEAKER_02And you need to be there in the just making sure that it actually happened. Everything else you can put on different constraints on the system where you don't really have to look at the code anymore, because if you put enough constraints that says, okay, the we need to achieve this level of uh uh code profiling, and this need we need to achieve uh the these amount of tests and regression tests. And you put all sorts of constraints that by definition the code quality is good if everything is green.
SPEAKER_02Um the fact that uh whether or not it's pretty or not, you're not gonna look at it anyway in the future, so it doesn't matter. Uh what matters is that it's efficient and that that it's um maintainable by the AI. Uh and you can solve this with all sorts of um of rules and constraints. So I don't think there's a lot of uh um technical um constraints for us to let AI do everything, but I think it's more of a quality assurance and trust. Yeah, yeah. And um, you know, going back to Matt to Moxie, you mentioned that yeah, the you have several ways of improving a formation.
SPEAKER_02One is the uh formation keeps learning uh and creates a file that is ready for the human to review. Uh so it will never kind of auto-update itself, but it will create a new version of itself and it will wait for you in a git format, and you can look at the diff. And if you approve it, then you deploy uh version whatever it is, point one. And um uh that that's how I feel that you do it. I haven't even allowed a flag of having it fully autonomous.
SPEAKER_02Yeah, I just don't believe that in a system that you need to trust, you can have flakiness. You know, some systems it's if it's health, it's gonna be life or a lot of money. And so, yeah, just put put in the time to review it.
SPEAKER_01Yeah, it's very interesting actually, because I mean for us at TESL, we have our own software factory uh that we look at every single pull request that is generated, even though it's pretty much end-to-end from issue to pull request, automated in terms of the the testing, the the code generation, and things like that. But we make sure we validate and verify everything that goes through to production. But for our own, for some of our internal tools, for example, our own dark factory, we allow fully autonomous going from issue all the way through to you know a pull request that gets merged without merging.
SPEAKER_01Okay. And I think for the internal, it a lot of it is trust, a lot of it is criticality as well, right? Uh and it's a case of if who's gonna find this first? Is it us? Or is it gonna be a user that actually then affects your brand or your or data, those types of things? And in some cases, we actually have maintenance agents that that will run, they will observe how the dark factories run, how agents are running, create issues, resolve, merge those issues, and we don't even realize those issues are being created or or merged at all.
SPEAKER_01And it's it's a s it's a scary world to live in, but it's almost like progression in small steps in non-critical systems first, till we actually then go to production and say, right, how close are we to actually doing this in lesser critical parts of production? And it's gonna be a gradual cultural change, I think.
SPEAKER_02It's it it is, and and I agree with you that's if it's mission critical and it's part of your brand, you you want to have humans uh in the loop. Uh and yeah, the the role of humans would shift in the even uh in the few months that we have till the end of this year, it will change. Uh and the role will move, maybe maybe diminish, but I don't think it's gonna be removed. No, yeah, and uh it's funny because we we also have obviously our own uh software factory, and we have this uh sentry-like feature we call ladybug that catches uh bugs in while they're still in production, even automatically a diagnosis created, and just there's a PR waiting for us on the other side.
SPEAKER_02Yeah and uh it it's been really fun to have this tool and do just you just wake up to PRs.
SPEAKER_01And that's yeah, it's particularly important in the 3 a.m. you know you don't get some groggy human who's like you know just waking up with pre-coffee trying to diagnose something. By the time they've woken up to look at a pull request, it's already been fully diagnosed, RCA, uh they've got everything waiting for them that they have to review. And and you know, for these types of outages that will cost money, that's so important. It is.
SPEAKER_02And look at the open AI hugging face breach that happened a few weeks ago. Uh if I recall correctly, it was like 17,000 um um hacking attempts or something like that. There's no way a human you know sitting in front of the logs with a cup of coffee can work through this uh because by the time they uh go through the 17,000 uh uh log items, it's gonna be 170 times log items. So you have to have those systems uh in place uh to actually do this for you.
SPEAKER_00Hey everyone, my name is Sahil and I'm one of the engineers here at TESL. I'm gonna be in New York and in Boston on the 16th and 17th of September talking about how we've built our software factory here at TESL and what engineering looks like in an AI native world. I'm gonna be diving into the components that we've built out, why they're there, why they're useful, the mistakes that we've made along the way, the sort of metrics that we focused on early that actually didn't end up helping us, and the ones that actually did, and all the different quirks and features that ended up building out the factory that we have today.
SPEAKER_00This is gonna be a much more practical look into the factory that we have here at TESL and some of the specifics on how it works and why we made some of the decisions that we did. Go to Tessel.io slash events and I hope to see you all there.
SPEAKER_01Now, Ran, you've also created a number of books. Um I'm not an author uh other than blog posts. I don't think that's a lot. Not that I'm really an author from that. But one of the one of the um quotes that I really loved is uh in and around uh how agentic uh systems have zero of the properties that would actually make them agentic. Right. Tell us a little bit about that and what what is it that most teams get wrong when thinking about something that is being agentic or building something that's agentic?
SPEAKER_01Right.
SPEAKER_02So I separate this uh into two terms. There's automations, there's fully agentic, which means autonomous uh and proactive, and you it's just another person on your team that you have no control over other than the gates you've set in. And most of what I see is either automations or automation with the AI sprinkled in. And um I think there's a misconception that um agentic workflow, the term itself, is kind of an oxymoron.
SPEAKER_02And uh the reason is that if it's a workflow, then you've already given the road um or the path that the AI needs to work in, so it doesn't get to make its own decisions. Right. Uh and I and I think that that's where a lot of um of the focus is uh on right now. Most people who are doing AI and agentic stuff, this is actually what they're focusing on. They're not focusing on actually creating this entity that can actually think and and and and uh come up with ideas for itself.
SPEAKER_02And the few people who actually went that extra mile are amazed by how capable AI can be. And it's fine because uh I'm I'm not dissing anyone because for most use cases you do need workflow, you don't need the agentic automation and and all of that stuff. Uh, if it's deterministic, you don't need AI. If it's deterministic with some fuzzy bits, like okay, I need AI in this step to write the email. Yeah, then it's uh it's an it's a workflow with uh AI sparkled in.
SPEAKER_02Uh you know, tools like N8N did that a while ago. It's not like a full living-breathing uh entity that um just like OpenClaw or Hermes or anything like that, that just decides on its own. And um yeah, so that that that would be my take on it. I think that uh if you're thinking whether or not you should go agentic uh fully agentic and autonomous, first think of whether or not you either you even need that um in the organization. A lot of business operations are fully deterministic, so you Do m that there's a lot of AI that's being used in places that it shouldn't be used and it just costing you money.
SPEAKER_02Where a simple uh script uh the the number of times I've seen AI try being used to parse strings where a simple regist can just do the work. Faster, cheaper. Faster, cheaper. Literally free. So yeah, it's a shame.
SPEAKER_01Yeah. Now um Automase has this uh four phase ladder approach. Uh from I think it's no computer, pure understanding, to more of a full architect mode. First of all, tell us a little bit about that, and then I'd love to dig in a little bit.
SPEAKER_02Yeah, so we came to realization that for at least a year now, you do not need to hire junior developers for their coding skills. That's just just you don't need them for that. Uh now if we stop hiring junior developers, we will never have this archetype of an architect that we are looking for. And an architect is um um, I don't know if this is a term that's gonna catch on, but it is someone that walks both understands the the product.
SPEAKER_02Uh so it's both a product manager, a project manager, and um a team lead where the team is uh agents in that case. So we work in an operation where we have pods, and that that pod has a um the senior developer, it's kind of the architect that manages uh these agents, and um a client-facing uh individual, which uh is the delivery manager, and they kind of uh work with the client. One is working externally, one is working internally. And the junior developers are there as apprentices for uh they're they're kind of joining along, like you know, go in the Renaissance days, you had the apprentice, and they would join the um the master painter and kind of learn the rope and uh dabble around, and sometimes would get their own project and here just finish this up.
SPEAKER_02And so this is the real skill that we're trying to build is the ability to um not multitask but multi-monitor monitor agents, so you'll be able to command a team of agents, and I'm not in and under any sort of uh fantasy that you can manage thousands of agents, uh, but you can manage six to ten maybe. Um but you need to that goes back to the to the workflow itself, you need to know how to spec products, you need to know what agents are capable of and we're not capable of, and uh be their mentor uh when they when it's time to develop uh software and um you know be at the gate where okay, is it ready for production?
SPEAKER_02Uh did I put in the right um uh workflows uh where when the PR is ready, I want to immediately create an environment where the full system is up and uh end-to-end test run. And so you need to learn these types of things. And um these are things that I don't think you can learn from um reading programming books.
SPEAKER_01And uh that's what that's what's kind of interesting where there is so much new from the point of a workflow, from the point of you know, you we as experienced developers are needing to reprogram ourselves in terms of how we operate. From a junior developer point of view, it's gonna be easier for them to snap into a different workflow because they don't have the baggage of many, many years gone by of all right, I need to sit down, do these deep designs, then code, then do and blah blah blah.
SPEAKER_01But from the junior developer side, uh of course they don't truly understand, you know, the more complex architectures, complex designs when they wake up at three in the morning to solve something. This could be, you know, for experienced developers, they understand what happens when AWS goes out, what the implications are, and how to go around fixing that. What would you say are kind of like how are you seeing both junior developers and senior developers, more experienced developers, snap into this new model of development workflow?
SPEAKER_01Where are the disadvantages for either?
SPEAKER_02I think that they have to work together because uh the examples that you've uh given, um, this is earned experience. This is uh you cannot accelerate that. You have to be there and you have to understand the consequences of uh uh you know, just how much load can um uh two VCPU's machine can handle. Okay, this is not something you can math your way around because your application differs from my application or that other application that you've worked on in the past. So you you need to have this um it is earned experience.
SPEAKER_02There's there's no way of of condensing that time. But if we can take you know, a junior developer, it used to take like junior developer like three to four years to become like a good level of mid-level developer, and seven to eight to become a senior developer. I think we can compress this, but not to zero and not to a year, probably to a two-year where a junior developer and I'm sure you have this experience. Um I'm sure you're seeing a lot more value from junior developers uh when they're here a year versus junior developers that were year uh after a year, like five years ago, right?
SPEAKER_02Because they can produce a lot more, they can um for lack of a better term, nag the senior developer less because they can actually use agents to okay walk me through this problem, yeah, walk me through this function, why is this doing that? And then come more um baked with the questions ready to to the senior developers for help. And I think that that's exactly the role of an uh an apprentice and a mentor kind of relationship that they have to be together for a while before the junior is ready to um kind of to start their own pod.
SPEAKER_02Yeah, um, and I'm not in any rush to um to to give them their own pod. I mean, I think that if if it's gonna take them a year or two, that's fine. It used to take five, so we're fine.
SPEAKER_01And I I love that what you I think you you call it teachable knowledge and earned knowledge, right? And I guess that is something that is, as you mentioned, teachable knowledge being heavily compressed going forward. Uh when we look at what was earned knowledge five years ago, that's gonna be so much harder to earn now because so many things are being taken away. Well, not taken away, but so many things are being automated and we can hand things over to an agent now. Do you feel like going forward there's gonna be a gap here in earned knowledge where it's important to have that knowledge?
SPEAKER_01It's not really teachable, and we can't earn it anymore because so much of that is being handed off to an agent. Is it is there a gap there that we're gonna struggle with?
SPEAKER_02I think there is. I don't have a good solution for that, but but uh I know what about you, but we moved to the UK six years ago. So my wife and I both got new phone numbers different than the one we have for the past 20 years. And I don't know her number by heart. I barely know mine. Uh because we just uh when we got her new number, it just went straight to my phone. So my brain never had the opportunity to memorize it. And with AI, I think we're sort of in a race between a race between how much it can do for us before we uh forget how to do it ourselves completely.
SPEAKER_02So whether or not we it can cure dementia before it becomes a wider phenomenon than it already is. So yeah, I don't have a good answer for that, but I definitely there are will be things that you won't even need to know because there are new solutions for them now, and uh some things that uh you'll probably have to um um to drill yourself into, just create a situation for you. And uh that that might be a partial solution, just create those uh those false alarms, you know.
SPEAKER_02Every uh every office building have uh fire drills. So you'll you'll have to you'll have to do these apparently.
SPEAKER_01Yeah. With these new workflows um and our teams being adjusted into the kind of you know more agentic workflow leading into a software factory, how much would you say your teams have become more efficient, more effective uh from you know even a in a time of delivery point of view?
SPEAKER_02Right. It's actually very interesting because we were kind of wary about a year ago what would happen to our sales. You know, we we were working on a retainer basis, now we're just gonna finish things faster because of the AI promise. And uh that would mean that we're gonna have uh less um lifetime value of our clients. Uh that didn't happen uh because uh while we are able to shorten the the development time, probably uh for for proof of concept you can you can reduce it to up to like 80% of uh of the time you can you can cut off.
SPEAKER_02But for a true production grade system, you can probably go down by a good one-third of the time because there's still a lot of tests and so many edge cases that you have to test for. So I would say maybe 40 to 30 percent of the time you you're able to cut. Uh but what happened uh for us uh at least that um we just ship more now. Yeah. So the project is not ending, but all that uh bucket list that the client had just goes uh straight on and uh we're just uh adding stuff.
SPEAKER_01The backlog the backlog of oh we might get to it one day is becoming a reality now. Yeah. So do customers do customers expect to pay less for functionality and features now?
SPEAKER_02Or is it you know I assume they they do, but uh since we charge on a monthly retainer anyway, so yeah, they they pay less per peep per feature, uh, but our pricing model ha hadn't changed. So uh we have clients staying with us uh for three years uh uh on average. So um yeah, it's uh it's it's mine. Your customers are actually getting more. They're getting a lot more. Same as the same thing. They're getting a lot more. And what I actually like is uh uh a lot of our clients, uh especially with new clients, that we get uh these are vibecoder uh coding experiments, and they're coming up with us with uh to us with an app that they've developed.
SPEAKER_02Nice. And obviously it's not anywhere near ready for going into production. But I actually like this. I um you know some people resent that as a from as a developer, as an agency that uh provides software development, but I actually like it because I know exactly what the clo what the client is after. Yeah, sure, we're gonna have to replace 99% of the code and redo everything, maybe change stacks or but it's fine as long as you can develop your vision in cloud, even if it looks ugly or whatever it is, but we get a real sense of what you're after from a product perspective.
SPEAKER_02Uh even if you just do mocks, that that's fine. Just a bunch of HTML pages uh that sort of look like in the app you're you're imagining. I actually like it because there's no misunderstanding there. Yeah, misunderstanding there. So I actually like it very much.
SPEAKER_01More clarity than just a speckle or something flat on the paper. Yeah. Um now, Ryan, um I I understand you I don't know if you still do, you used to certainly act as a an FDE uh for every new automated client. Is that is that still the case today? Uh some. Some. Yeah. Okay. It's true to some extent. First of all, tell us what it what is the role of an FDE? And then secondly, I'd love to understand how how that scales. Is that something you know you eventually think, I need to give this up, I need to I need to grow my FDE team.
SPEAKER_01Tell us a little bit about that.
SPEAKER_02So uh right now we're about um closing in on 30 uh developers uh on the team. And when we start with uh with any new project that we're doing for the first time, that kind of project. So whether or not it's AI transformation project or we're doing uh a new thing with uh healthcare or whatever it is, um, I'd like to be there, and the clients would like me to be there along with one or one of the two development leads that uh that are um running the show there.
SPEAKER_02The reason is that I cannot optimize or turn into a process something that I haven't experienced and that I don't uh fully familiar with. And um and the role of an FDA is basically to be deployed at the company remotely or in person, uh, which for us is remotely, and uh really be in contact with anyone that is related to the project to fully understand what's the best way to implement it and really understand not just what the spec requires, but how it's really going to be used, especially if it's an internal tool.
SPEAKER_02And um even if you're creating um I don't know, just an e-commerce website, you need to understand how the fulfillment on the company side is is, yeah, the we know what the shopping site looks like, but how the back end uh looks like for that company might be completely different than what it looks like for a different company. And uh, you know, I'm all for let's mold the technology around the company and not the other way around. So um I feel that I need to be there, um uh at least in in spirit, at least for the first, like that's a 10% of the of the job, so I can spec it better, so I can make sure that it fits um in a way that we can move it forward.
SPEAKER_02And um, yeah, so that that's how I usually do it. And whether or not it scales, it does not scale. It's it's a human time, it it definitely does doesn't scale. So that's why I'm doing it less on a per client, but more on a per project type. So everything that we're doing for the first time um in that sense, then uh I'd like to be I like to be involved and then hand it off to the team. And yeah, it it works. Yeah, it works. We're not doing the FD in a uh traditional sense of like we're we're implemented in the and embedded into the company forever.
SPEAKER_04Yeah.
SPEAKER_01Hey everyone, hope you're enjoying the episode so far. Our team is working really hard behind the scenes to bring you the best guests so we can have the most informative conversations about agentic development. Whether that's talking about the latest tools, the most efficient workflows, or defining best practices. But for whatever reason, many of you have yet to subscribe to the channel. If you're enjoying the podcast and want us to continue to bring you the very best content, please do us a favor and hit that subscribe button. It really does make a difference and lets us continue to improve the quality of our guests and build an even better product for you.
SPEAKER_01Alright, back to the episode. I'd love to go big picture now and talk about some of the some of the fun things, uh, some of the some of the fun quotes and some of the fun uh uh hot takes you've had. Um The first one I'd love to talk about is uh when you called an agent workflow an oxymoron. Um what did you first of all what did you mean by that?
SPEAKER_02So uh I've talked about this before, and uh a workflow is not agentic by nature. A workflow is a deterministic set of uh of uh steps. Uh you know the inputs and you know the outputs. So just calling it agentic uh is uh either misleading or if it's really agentic, then you're just wasting tokens because it could be fully deterministic or mostly deterministic. So for for something to be called agentic, it has to be fuzzy.
SPEAKER_02It has to um you it has to be non-deterministic by nature.
SPEAKER_01And uh that that's that's just my take on the the reason why I think that um it's just the wrong word to say it's not the workflow that is agentic, the workflow is is I don't want to say static, but a very deterministic workflow that has agentic components in it, but the workflow itself isn't agentic. Yeah.
SPEAKER_02So uh what I uh so kind of my my interpretation for this, if you really want to have a mostly deterministic or somewhat deterministic, so we're just using um a lightweight version of an SOP, uh which is this is how we usually do things. Uh this is all the different forks that you can take um and make up your own mind. But I've given you the guidelines of what how we usually do the if you can find a better way to do it, sure way to do it, go right ahead.
SPEAKER_02But I'm I've given you some context on how this is usually getting done.
SPEAKER_01Going forward, do you see scope or a place for an agentic workflow where the workflow isn't necessarily deterministic, but the agents determine what needs to be done based on certain criteria?
SPEAKER_02Yeah, but then it's it's not really a workflow. Uh right. So it's um um again, uh maybe someone uh else uh have a different uh definition for that. But for me, a workflow is deterministic. It's something that you can draw on a whiteboard and and see how it works. It can have fuzzy parts, like I mentioned, okay, I want to know exactly what to write in the email based on the customer context, but everything aside from the draft is deterministic, and I'm sprinkling AI on on different places.
SPEAKER_02That that's fine. Um but uh if we need to get to a point where you the agent really becomes uh uh autonomous, yeah. So it knows the company and it acts as uh as uh an employee with uh really that's really open-minded and kind of uh try to uh to tackle things that nobody else wants to do because it's too much work, it takes too much time, and it's a machine, it doesn't care about any of those. Um so yeah, I think that we need to go more to that type of direction.
SPEAKER_01And ultimately we do need that determinism in the workflows.
SPEAKER_02For a lot of things you will always need the determination, uh uh, but you can have workflows, but you can have an agentic uh entity that decides on which workflow is uh needs to be called right now. Yeah, so um you know you you come to work every day, you know how to do different things, you have workflows, but you're not working on an assembly line where you have your this is where you stand and this is the the box that you need to move. Your work is a lot more uh um uh dynamic, but for every task that makes up that work that work, you have uh a workflow.
SPEAKER_02So I think if we want uh agents to be more like humans but just faster and more efficient, which I think this is what we're after at the end of the day, I think that that's the direction we need to go. We we need to provide them with a set of workflows, SOPs, whatever you want to call them, but and within those um uh environments, just let them let them do their thing.
SPEAKER_01Another hot take, Ran. Um anthropic risks turning its strongest advocates into its loudest critics because of a specific strategic signal. Tell us about tell us about your take on that.
SPEAKER_02So it uh kind of go goes back to OpenClaw that was banned from using the the Max plan. Yeah. And their whole uh so for for anyone who doesn't know, uh OpenClaw is a great uh successful project uh by Peter Steinberger, and um it used uh Claude and specifically the Claude um uh Pro and Max plans in order to um to act as its brain. And um they just shut it off the access.
SPEAKER_02If you want to use it, use it through the API. Their rationale was it messes up our statistics. We don't know how many people are actually using it for uh coding purposes, and how which is yeah, you could have found a different solution around that, right? The solution is not let's shut it off, the solution is have a toggle mode. You do have a co-work mode now, so you could have done an open claw mod. Uh so it's it's uh it was a very weak argument. And um the the reason was um was financial, obviously.
SPEAKER_02And uh I'm fine with that. That that that's fine. It's you're it's it's you're a company, you need to make money, but don't be us, you know. We we can go without that. And um I've seen this, there's a new project uh called Buzz by Jack Dorsey. It's really cool. And I also tweeted jokingly that it's it looks very promising. I can't wait for Anthropic to shut all access to it. And um yeah, it's fine. If you want to charge API, charge API, but for me, you know, I'm I'm paying$200 a month for a single subscription and I have multiple.
SPEAKER_02Um why can't I use it the way I like it? If if you decided you want to give me some qu some allowance, then let me use the allowance however I see fit. Um yeah, it's it's kind of uh Do you see other providers following suit there or I think that uh we're seeing the exact opposite, actually. Um maybe it's just uh temporary strategic uh move by OpenAI to say, hey, you can use it for whatever you want. Uh Grok is also doing that.
SPEAKER_01Uh the Peter Steinberger uh situation was quite specific because Peter now works for OpenAI.
SPEAKER_02But but I think they actually told you you can do whatever you want with it even before they they started uh uh discussions with him. Uh but yeah, it would definitely a counter move to what Anthropic uh uh has done. But um I think I think this whole subscription model, it's eventually gonna go away, uh, or it's gonna be a lot more specific. So you're gonna get uh so let's say that they're charging five dollars and twenty dollars for input and output. So your subscription will get you a discounted uh um um but you're not gonna get uh sorry, discounted price on X number of tokens, but you're not gonna get 60x uh on your plan compared to what the uh API provides, which is around what you're getting right now with cloud subscription and open AI subscription.
SPEAKER_02This this party has to end unless they come up with an ultra-efficient uh uh GPUs or some some miracle chip um and that everyone's working on and we'll probably see, but I don't think we're months away from it. I don't think we're even uh um a year or two away from it. Unless the uh until the the they'll be able to dramatically cut down their costs, there's uh it's it's it's a race to the bottom. And I think that as soon as uh I think it's uh as soon as Anthropic goes IPO or something like that, they're they have no um the the the motivation for such an aggressive uh user uh acquisition and growth uh slightly reduces once they're IPO'd.
SPEAKER_03Right.
SPEAKER_01Let's take one more hot take. I like this one. Uh it's around your law firm model. Right. So your law firm model argues that dev agencies are heading more towards something like a Kirkland and Ellis than a SaaS startup. Why do you think that one?
SPEAKER_02Okay, so right now we're very close to every single developer being able to create their own software factory. Maybe on a technical level, they're obviously we're already there, but on a really production level, we're almost there. So you can be um a reasonably good developer and you can start your own firm because you're not limited by the capacity. The however meetings you can hold to close sales, that that's your true capacity. It's not about the development. So if you really want to attract and hold on to talent, you have to give them a stake in the company.
SPEAKER_02Um so that's why and law firms uh they they've done this for ages. You just become an associate, then a junior partner, then a partner. And yeah, you buy yourself into by the number the business that you bring in. But um if if it wasn't for this model, then there'll be just more and more law firms instead of uh more and more partners in inside existing law firms. And it it is a matter of motivation. If you want something to stick around with a startup, you can offer stock options.
SPEAKER_02Uh, there are different things that you can do. Uh, but for a business that is um not um VC backed or focused on hypergrowth, the only real uh benefit that you can give them is because you're not heading for an exit in those types of um businesses. Uh so what you can offer them is just a piece of a cake. And you say, okay, you bring in you you handle your own pod, uh, so you get this amount of uh cut from uh the revenue or from uh or from the profit.
SPEAKER_02And I prefer revenue because then you don't have to uh let everyone know how your business operates. Uh so that's why we also created an open royalties type of agreement, uh, which it's very clear. We we know what we get from the client, that's all you need to know as a royalty partner. Yeah, so uh and it's our problem to solve uh all the the expenses part. But yeah, if if I want to hold you, I need to make you a junior partner, then a senior partner, and um give you the ability to earn.
SPEAKER_02And I think that companies that especially software uh shops that are not going to use that model in the next few years, and we're not still fully there. I I'm just you know, for for if I'm being honest, we're not there, but we're heading there. We built we built our pod model and then we're slightly uh slowly moving to that direction. But there's you know, once you have a senior developer that manages a pod and they're handling two, three, four clients, um, and you're paying them, let's use kind of round low numbers, you're paying them one dollar a month.
SPEAKER_02But they know that their client these clients are bringing ten dollars a month. Why would they stick around with the company if they can say, okay, I'm gonna earn zero for a couple of months, but I'm gonna lend my own clients, even one client, I'm gonna have less work and more money, you know. Uh I'm just gonna keep uh everything. So you you you need to be able to say, you know what, 10, 20, whatever it is percent of the clients that you manage are yours. Another 10-20% goes in a pool that is being split around the pods.
SPEAKER_02And um I think this is a healthy model. You know, I used to do like 20 years ago, I used to do a lot of uh uh affiliate marketing. And what I liked about this model is that uh it's uh the the um uh the motivation uh is aligned. Yeah. So everyone is wants the the business to to make money.
SPEAKER_01And and is this would you say this is very much favoring the startups because they're gonna be more flexible in in being able to offer more of those types of things to to their employees? Or uh and also does this does this lead to almost an explosion of new startups being created? If not, because people will think, well, actually I could look as you say, I need one client and I'll get more than you know the the the 10 clients or the same as 10 clients worth of revenue.
SPEAKER_02So it really depends on the startup. Uh most startups are not profitable by definition, that's part of their growth strategy. Uh so for those startups, especially if their goal is to become a unicorn and go IPO, then I think the existing model is already good enough of the stock options and equity. And yeah, it's it's uh pretty much doing its job. But if you're a bootstrap business or one that's focused on uh revenue generating, yeah, you know that that's that's a focus. You're not playing a goal IP. It's never gonna be uh even with Automase, we're not gonna be a hundred million dollar business.
SPEAKER_02Yeah, uh, but we are gonna uh uh continue to generate a good amount of money, and that's the cake that we have uh to offer. So um maybe we'll I'll sell it in the future, I don't know. Um it's it's not part of the strategy, yeah. So uh I I don't know. A good offer will come, we'll we'll see. But uh this is definitely not part of the strategy. We're here to stay. So this is the only way for me to uh really uh hold those people is to give them part of this revenue. And I think that for service-based companies, um you'll definitely gonna see that.
SPEAKER_02So you're you're definitely gonna see, or hopefully, I don't know if definitely, but hopefully, you're gonna see this, and not only for software uh for software stuff. Uh you know, every if if you have a network of plumbers and you want to keep the best plumbers around, you have to pay them more than hourly. Otherwise, again, they'll just they just need a lot fewer clients to match their salary than what they serve right now under this umbrella. So um just align those interests.
SPEAKER_01Amazing. Uh Ran, where where can people find, where can our listeners find out more about Automase and Maxi?
SPEAKER_02Uh well, so I'm mostly on X, so that's the best place to uh to go. Uh link to everything that I do. Uh Maxi is on uh Maxi.org, M U Xi.org, and Automaz is automaz.io. Uh but yeah, just come on X, follow me, I'll follow back, we'll have a good discussion. And your handle on uh I can't call it X.
SPEAKER_01I can't call it X. Your handle on Twitter? Twitter, yes. R O U S I. Perfect. So that's it. Amazing. Rand, it's been a real pleasure chatting with you. Thank you very much for having me. And I love the I love the hot takes. Really, really wonderful. I hope you at home uh enjoyed all those hot takes as well. And uh thanks very much for listening. See you next time. The AI Native Dev is brought to you by Tesla, the package manager for skills and context. Your hosts are Guy Pajani and me, Simon Maple. Our producer is Tom Dowler. The AI Native Dev is not just a podcast, it's a community.
SPEAKER_01And we host monthly meetups at the Tesla offices in central London. Visit Tesla.io forward.community to learn more for ice.
Developers are about to write skills instead of code, and Microsoft's Daniel Meppiel argues the software factory that runs them needs versioning, a bill of materials and signed receipts before enterprises will trust it....
Paul Stack helped build Terraform. Now he says he'll never write another line of infrastructure as code, because an agent harness runs his team's production infrastructure instead. Watch all the AI DevCon NYC 2026...
Automated software delivery is moving faster than the people who used to approve every release. LaunchDarkly co-founder and CEO Edith Harbaugh tells Guy Podjarny why rollback should be autonomous, why one-shot coding...
Agentic coding has outgrown the text box. Maggie Appleton, staff research engineer at GitHub Next, argues that walls of CLI output and AGENTS.md files serve coding agents far better than the humans steering them. She...
AI attackers now move at machine speed, while most security teams still move at meeting speed. Former Robinhood CSO Caleb Sima has seen an AI attacker up close, and he argues the only way to keep up is to take humans...