140: Eating the Burst ![rw-book-cover](https://share.snipd.com/share-image/resize?url=https%3A%2F%2Fimg.transistorcdn.com%2F7-tuF7RMSNRIRfD3M2fNWLQLF-ShDgRiBeceKLOFVZc%2Frs%3Afill%3A0%3A0%3A1%2Fw%3A1400%2Fh%3A1400%2Fq%3A60%2Fmb%3A500000%2FaHR0cHM6Ly9pbWct%2FdXBsb2FkLXByb2R1%2FY3Rpb24udHJhbnNp%2Fc3Rvci5mbS9zaG93%2FLzQzNjc3LzE2OTE1%2FODI4NTItYXJ0d29y%2Fay5qcGc.jpg&width=512&height=512) 🔗 **Source:** [140: Eating the Burst](https://share.snipd.com/episode/1f782f61-e53f-4629-905e-f57313c28a94) ## Highlights - **Dallas Ticket Over Family Chaos** - Aaron Francis told a story about getting pulled over in Dallas for speeding and an expired registration, receiving a warning and a ticket despite pleading family chaos. - The anecdote highlights the everyday friction founders face juggling family logistics and business life, like missed renewals and unexpected fines. Transcript: Ian Landsman No i didn't think they did that in texas i didn't think they did that. I know. Aaron Francis All these rules. Here's the thing. When I'm doing something wrong, I'm just like a good guy trying to make it in the world. Everyone else is like, they're bad guys, right? And I'm like, oh, you guys are driving too fast. But for me, it was like, oh, little old me. I just, oh, I forgot. So I'm coming home. I'm coming home yesterday and i just i just taking the big kids to like an indoor playground we're coming home for dinner the freaking lights go on behind me and i'm like oh no what is this About because i have no idea how fast i'm going it's a six lane road and i'm like oh it's on the highway? No, but it's Dallas, baby. And there's three on either side. And I'm just cruising. And so I pull over and my daughter says, are you going to jail? And I'm like, no, I don't think I'm going to jail. I was doing 48 and a 35, which. Oh, my God. That's like not even worth pulling you over for. Come on. Not just the accepted percentage over. That's what I'm saying. So he's like, why are you going so fast? I said, I don't know. I wasn't paying attention. Ian Landsman Like, have you looked back here? Aaron Francis He said, why is your registration out of date? And I said, well, I got two more of these at home. And he was like, what do you mean? And I said i got two sets of twins so i don't really do stuff on time and he was like all right and he walked off and i got a warning for the registration and a god dang ticket for speeding man Ian Landsman I haven't gotten a ticket in 15 years really okay i got a ticket i got a speeding ticket four or five years ago, I think. Something like that. Aaron Francis I used to get tickets all the time. Oh, I used to get pulled over all the time. Oh, dude. As a teenager, it was like weekly I got pulled over. Yeah. So I'm just going to give it to Sam and have Sam figure out what I'm supposed to do. Ian Landsman You're supposed to pay it. ([Time 0:00:25](https://share.snipd.com/snip/bde25ec0-cd35-4014-9f91-6527dc248666)) - **Jay‑Z Field Seats Through Producer Connections** - Ian Landsman described attending Jay‑Z at Yankee Stadium with Producer Dave in the Legends center field area and enjoying the buffet and special entrance. - The story illustrates industry networking perks and how small team relationships (Producer Dave) unlock premium experiences. Transcript: Ian Landsman I do. On Friday night, me and producer Dave, who produces this very podcast went to jay-z at yankee stadium oh that's pretty iconic yeah we saw jay-z live i never thought i'd get to see jay-z Live at yankee stadium that's cool for the largest crowd ever at yankee stadium for a concert we were there um until the next night and i think they broke the record. But one night it lasted. It was awesome. We were on the field. We were in center field of Yankee Stadium looking up. All the lights, everything's around us, all the people. Yeah, it was great. The concert was, I'd say, very good. It was a couple production things. It wasn't as crisp as Taylor Swift, which was the last concert I went to yeah yeah i mean she's like puts a year into building it now i don't think jay-z is going to put a year into it but good I think the next night so first night was reasonable doubt night which is his original album and this is like to celebrate 30 years of that next night is the blueprint which is a little More of a stadium type album so i think that that night is probably very good uh live and then last night uh apparently was uh kind of uh all his hits which i did not attend the other two nights Just the first night we had a good time went back to hotel had a few drinks a nice evening was had by all it was cool that's super fun i didn't know producer dave went with you i saw the pictures I didn't realize he was there. That's fun. Producer Dave, he made the tickets happen. He got us the fancy tickets. We're on the field. We're in the Legends area. Whoa. But you know that's going to be good. So we go in special entrance. We get right in. You know I love a special entrance. There's nothing better than a special entrance. There's nothing better than being special. Aaron Francis That's all I want in life. Ian Landsman I just want to be special. I just love to go in the special entrance. So we go in the special entrance. And it's like a whole like fancy restaurant. And it is, there is some stuff on a menu I think you can order. But primarily it's like a buffet. But there's like carving stations, barbecue. There's like burgers and dogs and pretzels if you want stadium food. So we had a couple rounds of buffet. Very nice. Sit down like gentlemen at ([Time 0:06:48](https://share.snipd.com/snip/bcacf80c-c59c-419d-8807-da5757e20cbf)) - **Pay To Play Limits U.S. Soccer Talent** - Ian explained the U.S. youth soccer problem is economic: pay‑to‑play club systems exclude talent from low‑income families, unlike national club models in Europe. - That pay‑to‑play model funnels resources toward parents as customers, reducing incentives to recruit or develop the best players regardless of means. Transcript: Ian Landsman Let's do a little business dad take on this. We'll tie this in because I thought this was very fascinating. Read this article about why the u.s sucks at soccer and all these other countries i mean whatever norway's kicking our heads in or whoever you know all the belgium whatever all these Little tiny countries one one hundredth the size of america all have better soccer teams right and it's really because the business of it and it's. So, and my son was kind of into this, and you've been into this, so you know about this. Aaron Francis Dude, my poor middle-class parents had to pay so much money for me to do proper soccer as a kid. Yep, and so every other country doesn't work like this. Ian Landsman So in America, if you don't know, you pay to go into the better club teams, and then you go up the ranks, you pay more for better club teams and whatever. You go up the chain like that. So all those club teams, the customers are the parents. And the parents who have money then are able to put their kid who maybe isn't the very best in the world at soccer into those club teams. Hence pushing out either other people who should be, have more promise, and or just people who straight are awesome and can't afford to go into the club system and just never get to participate. Whereas in other countries, the club teams are paid by the national system. And so the club teams are rewarded for getting the best player no matter if the parents have money or not they're pulling mbappe off the streets yeah they're going they're going to they're Aaron Francis Going to the tiny villages and finding messy like exactly that's how they got them guys would have never made it inside of the u.s industrial complex right never got to be middle class Ian Landsman And above and even then, you know, that's even then, like, there's just probably better players out there, but there's no reward for the team to go find them or develop them because there's No money that those people can pay. And that's it. And that's where it all dies. I do think in America, you also have a bunch of other sports that some of the best athletes probably get pulled off too, which is a little different than the other countries too. But I never realized the economics of it. ([Time 0:13:08](https://share.snipd.com/snip/1839e927-d5c2-4a97-9e47-6ea35bb22faf)) - **Write Deep Technical Content For Credibility** - Aaron and Ian recommend deep, technical blogging to build credibility and attract engineers rather than short SEO‑slop posts. - Cite examples: Cursor and PlanetScale publishing 3,000‑word engineering deep dives that get shared on Twitter, Hacker News, and lead to hires/customers. Transcript: Aaron Francis Yeah, I think the death of SEO and blogging has been greatly exaggerated. Yeah, I think you still need those things. I think it's still incredibly valuable. Yeah. Especially, and here, let's do this. Especially in the AI age, I think a lot of people think, like, I just, you know, why would I need a blog, write a blog, have a blog when everybody's just going to go for answers? But still, you look at some of the big technical big dogs, like you look at Cursor, and they're writing 3,000 word articles about their new models and properties of like the application. Look at uh you look at planet scale and they're doing deep dives into postgres performance tuning and all of that and i think of course part of that is seo but i think part of that is just Technical credibility like yeah say nothing about driving traffic from google and you look at it as a pure way by which you can gain credibility with a highly technical audience, I think It's still probably one of the best ways to do that. So that gets shared on Twitter and Hacker News and Reddit and passed around in Slacks. And of course, you get SEO benefits, but it gets to the people that make the decisions and it's like oh these people are highly technical and they're very credible so i think there's still A lot of value in doing the hard work i do think the top like let's do 500 words slop that used to be the seo technique of just like hire a ghostwriter slop slap some stuff out. And I think that's all dead, which thank God. Right. Ian Landsman Yeah, I think that's mostly dead. I mean, at the end of the day, the AI still does web searches a lot and it's doing that off an index, right? And so it's still, there's still some aspect. I mean, pure slop is pure slop, but like there's still some aspect of just having some content out there, even if it's not amazing 4,000 world word, you know, deep dive but um the other Thing is like what you're talking about there is like publicity and publicity is still important so it's like okay maybe the seo juice of this is slightly reduced but when people share It on twitter and they find out about planet scale because somebody shared a technical article about planet scale that PlanetScale wrote, then that's publicity and that's still good. And that still gets you customers. And then yes, they go to the blog and they see you wrote a bunch of stuff and whatever you are doing. It seems like you have good engineers and blah, blah, blah, all that stuff. So yeah, I think that stuff's still there. I mean, publicity is a harder game than SEO was. So I think that's, you know, people are sad, myself included, that that's the case because it's not as easy a game to play. But still, I think there's definitely big value in having a blog and writing articles. ([Time 0:24:49](https://share.snipd.com/snip/a16e29f3-d508-4004-ae4c-a7128b44ffd8)) - **Bill AI By Ticket Tiers Not Per Answer** - Ian favored billing AI by ticket volume tiers instead of per‑answer microbilling to give customers predictable, budgetable pricing. - He argued tiered monthly pricing covers varied AI tasks (classifiers, auto‑drafts) and avoids adversarial usage monitoring and refund churn. Transcript: Ian Landsman What do you think about this? So there's two models we're looking at. Okay. And I think I made my decision, but let's get your take and the audience's take. So the most standard way and what most of the other tools are doing is per resolution. So I'm going to do a bunch of AI stuff. An email comes in, whatever. I'm going to it i'm gonna do ai things to it i'm gonna auto draft stuff whatever and then if the customer doesn't respond which i feel like a little skeezy but there's kind of no way around It or they give like you know there's like a thumbs up thing or whatever they say they're happy or they don't say anything boom 99 cents you pay 99 cents for that okay and then if the customer Is unhappy uh or whatever there's other criteria you could have where things don't apply if it's spam whatever then you don't pay for those so you're basically just you're trying to Dial it into like the very specific amount of answerable questions you just pay per answer effectively now what i don't like about that is that it's a very hard thing your hand okay yeah To estimate i am tipping my hand uh it's very hard to estimate as a customer how many are these are going to be actually ai resolvable and what that's going to cost me um there's also like A lot of weird incentives again around like people just saying nothing i'm going to bill you but maybe the people aren't really happy which you know sometimes people are unhappy and Leave like then people reviewing the answers to see if they're good or bad and like now you're doing refunds and like i don't know there's a lot of like nickel and diming aspect to that Whole thing. And it's not a nice clean nickel and diming like AWS. Like, hey, you just made gigabytes. That's what it costs. A gigabyte to gigabyte. That's it. There's a lot of subtleties in this. So the other option I'm considering, which I'm tipping my hand is the one I think I prefer, is you just pay on volume of tickets you receive in a month so if you normally get 500 tickets a Month that is x dollars in the help spot plan if you get a thousand it's y dollars and they're like big bigish tiers so not like 50 difference like you know 100 500, 1,000, 2,000, 5,000, Whatever. So because people know that, generally speaking, they're like, oh, yeah, we do like 300 a month. Okay. So they know they need up to 500 tier. They pay up to 500 tier. And the other thing that I like about this is that at least in HelpSpots AI, some of the other tools don't do this, is that agents can do other things. So they can do stuff like, like we have answering agents and we have classifier agents. So the classifier agents can just read inbound stuff and like pull data out of it and put it in a custom field or do other types of tasks that are not just answering okay so there's going To be ai costs for us people doing other sorts of automated actions that are not just answering so this way that kind of covers everything it's like whatever ai stuff you're doing regardless Of if you're answering or not like this is the monthly price you know what the monthly price is. You budgeted it. That's it. Or we go per answer and you could get into like the other way of these classifiers, for example, is like you could get into the whole credit system. Like, okay, auto answering is two credits, classifying is one credit, whatever. Other stuff we come up with is four credits, blah, blah. You have a whole kind of complicated crediting system. So, I don't know. What do you think about all that? Well, you poisoned the well, but I think... You're a smart man, so I know you can suss through it. I'm not worried about you poisoning the well. Yeah, I can hold two thoughts at once. Yeah, exactly. Aaron Francis So, you obviously have done way more B2B sales than I have. My understanding is that, uh, B2B buyers do prefer knowing what things cost. And so it's been my experience. Yes. So they're like, Hey, why don't you sign up for this variable thing that could go way up or which is fine could go down but you're as as a b2b buyer you're on you're on a budget you have a budget You're a man under authority you have to report to someone um so and even from like a even from just a little old me perspective i like knowing i don't like having an adversarial relationship With billing. I like knowing if I get my 500 tickets, that's what I'm going to pay for. So I think from I think from all perspectives, it's better to have like a knowable price. And then that lets that lets the agent like the human agents and the buyer just be like, yeah, just do your job and not have to have this background process running of, well, make sure that You guys don't do this because clicking that button actually costs us a little bit of money. And if you all click it, it costs us a lot of money. No, don't get that out of here. So I'm on board. Yeah. Make it simple. Make it non-adversarial and make it just bucketed by tickets i think is reasonable all ([Time 0:31:52](https://share.snipd.com/snip/0f8e692e-aa75-4830-959c-c6a60b213a27)) - **Burst Allowance Solves Seasonal AI Spikes** - Aaron suggested using a trailing average or burst allowance to handle seasonal spikes so customers aren't penalized for short-term surges. - He referenced AWS burst capacity mental models and proposed month‑to‑month burst limits or a trailing three‑month average to auto‑adjust tiers fairly. Transcript: Ian Landsman Right i like it so there's not many other tools that do it that way but that is the way i'm leaning um there is also the world of like what do you think about this so okay we go with bucketed Tickets you cross over 500 yep you're you're about your 500 monthly you cross over 500 do we do the your auto i mean there's like kind of three different things there's the like we auto Move you to the next tier permanently which is actually like what convertkit does which for email lists i think kind of makes sense potentially but for us like you might just have a christmas Time spike and then you go down so i like moving permanently to the next tier, I don't know. I don't think I love that. There is like for this month, you just get charged more. You moved up this month and you charge you more. Next month, you go down to your base tier. Or there's just, no, you pay your base. Within some percentage, we allow overages. Like if you only do it once, if it only happens once in a while, we don't care. Aaron Francis And then obviously if every month you're going over we say hey you gotta move up so yep the last one i think it's i think it's a i think it's like a trailing three month average or something So if you do like 400 400 700 400 400 like yeah that's probably fine um that's what i think yeah if you do 400 550 525 540 560 you're like well you're in you're in the above 500 tier right so Yeah i don't know if it's like 10 percent overage doesn't count or if it's trailing three months it doesn't matter something like that but yeah again trying not to be trying not to be nitpicky And adversarial feels like you get goodwill without having to actually eat a lot of cost um so i i that's how i would prefer it as a user yeah that's i'm thinking something along those lines Ian Landsman It's a little tricky because like since we do have costs with this which is terrible um traditional software, like the weird ones are going to be like you have a baseline of 100 tickets A month or whatever. It doesn't matter what the baseline is. But Christmas season is your busy season. And for two months in November and December or December, January, you know, you do a thousand tickets a month. Yeah. And so like now it's eating that is like definitely a hundred percent losing us a bunch of money. So there's something there. I don't know what that is, but maybe there is like just like some percentage overage is allowed, you know, as long as it's only a month or two. But like if once you're over 25% over, like we will be billed one month at the next tier or something like that. Aaron Francis You know, one place I think you could look for inspiration is, like, going back actually to AWS. They have the concept of, like, burst capacity. I think on RDS, maybe they've got IOPS burst capacity. On Lambda, they've got burst capacity. And it gets, it gets real complicated in the like actual accounting of the stuff. So this may be a reason not to do it, but basically like you're on this tier, but for a month you can burst up to this many and it's still, it's still like it's allowed. You can burst up to this much for one month out of the year or one month out of the six months or something as long as you come back down. And, you know, then you got to track, you know, the burst allocation. And you could do it in a much simpler way than, like, the IOPS does because the IOPS on RDS, like, refill your burst allocation. You don't have to do that. You can just do it month by month. But you might look into how other non you know help spot or help desk companies are doing burst capacity because that might be a good mental model for that yeah now though so i like i like Ian Landsman The name like i like the term burst i think that makes sense um but then that kind of brings us back here because should it just always be but then it's like tricky because if we let you burst And what we say I think the thing would have to be like when you burst like you just pay the tier you end up in like you know we allow like 10% overages right then you burst up that month okay You do 200% your normal amount great like we just bill you that month at that. But like, I guess in a sense, it's like, that's sort of the same. If we just let you do that every month, then you're just always up there at the higher rate. Like it kind of ends up not mattering that there is a baseline. Like then it's sort of maybe only hurting the customer on the low end. If you have a low month. Right. That we're just charging you the full amount and we're not discounting that's why i think that's why should it just be month by month by the ticket volume instead of by answers though Aaron Francis That's why i think the burst is nice because you don't get penalized for it right so you're like i'm rocking 300 400 and then i burst to seven and it's like well yeah in fairness i was also Under the 500 for the last two months and so this should be okay right and you're like yeah that's your first capacity eating the burst yeah yeah yeah yeah that's how rds and lambda do it Is like well at least rgs i know is like you you pay for a tier but you can go over as long as you come back under within a certain bound yeah yeah i mean that generally would work that's sort Ian Landsman Of interesting you're getting burst like we could even have credits like it's complicated but maybe it's just a certain percentage of credit there's a certain percentage where it Just makes no sense that it can't be free even with because you started this is your first month and you burst to like 500 times but you know five x like totally like you have to pay for that So there's something there but i do okay a burst with like some ranges i kind of like that we eat a little bit of it we've pad the numbers enough so we're not actually eating much, really. But, like, keep it simple for the budgets. Aaron Francis Mm-hmm. Yeah. Ian Landsman There's probably still some edge cases in here, but I do like that term burst. And I feel like the burst is, like, a concept people will understand. So I like that. I think so. ([Time 0:37:15](https://share.snipd.com/snip/3c76bdc0-7a50-47cf-8ab6-5207c94aeee0)) - **Tiny Task Models Cut AI Cost And Latency** - Aaron highlighted small, task‑specific models (19th.com) as a path to much cheaper, faster inference for narrow HelpSpot classifiers. - He pitched training tiny models (~2–3MB) on ~1,000 examples to detect terminal waiting states, potentially cutting inference latency to milliseconds and cost dramatically. Transcript: Aaron Francis Tell us your use case. And I told them and they were like, oh, this is perfect. And then I haven't heard a single word back from them. It is 19th. So 19th.com. 19th.com. And the H1 on 19th is your task is specific your model should be two okay and so their promise is we will we will train a very tiny very cheap to use i don't know what it costs to train model Um that can do your specific domain task very, very quickly. Like we're talking like 12 milliseconds instead of, you know, 2.5 seconds or something. And the models come in at like two or three megabytes or something. So they're very, very specific. They're not general purpose models. And so what I reached out to them for was, I want to have a model that I can deploy with Solo that is super small and super fast, and it only answers one question. And that question is, given this representation of a terminal, is this terminal waiting on user input? Is it waiting for a permissions decision? Is it waiting for the next user turn? That's all I want to know. Just give me a confidence of this is waiting for a user. And they were like, yep, we could do that. We just need like a thousand examples. And I was like, I could give you a thousand examples today. And so this, you know, as you're thinking about costs and costs and revenue you're going to be able to turn some knobs eventually right on we don't have to take a classifier even to like A haiku five four five which is cheap and fast but if you had something that was tuned specifically for you know help spot classification scoring or ranking or whatever, you could get Real cheap, which would make your margins obviously go way up. So that is provided these people ever actually launch or email me back. It might be interesting. Ian Landsman So I thought about these custom models are pretty interesting. I'm a little dubious that the training cost is ever going to get low enough. It feels like the training cost is like where so much of the money is on these things. ([Time 0:43:58](https://share.snipd.com/snip/ba9e95d8-f079-4c3d-9736-b493a95bc939)) - **On‑Premise Offers A Security Angle For AI Helpdesks** - Ian emphasized HelpSpot's unique advantage: on‑premise deployment for customers who need local control and stronger security than cloud-only AI options. - He noted specific customers—healthcare, government, finance—prefer VPN‑isolated setups to limit external exposure of sensitive data. Transcript: Ian Landsman The other thing that helps it has which again, nobody at all has this, is that it's on-premise. And so if you want to have an AI forward help desk, but you want to run it on-premise, there's literally nobody else that can do that. There's probably a couple of open source things out there. So that's sort of an interesting angle too. Aaron Francis It's just insane to me that people are still putting stuff on-prem. Like that is a great indication of um the unknown unknowns like i just i just didn't know that i didn't know that people were still doing that yeah and it's it's just indicative of the entire Ian Landsman World there's it's so much broader than just x.com the everything app you know yeah and in the ai age i think it's sort of interesting because like people are worried about ai hacking Them so you have that pressure that's not gonna be everybody most people are just gonna be fully cloud right but for certain types of industries certain types of organizations all right You worry about ai hacking but now with ai you want to centralize a lot of your data which is like kind of bad from a hacking perspective in a sense now you're even more worried about security. What tools, how do you open up your tools to allow access to all of your data? Like, okay, where do we want to put the things that can literally access everything in one call? So I do think there is like this pressure for certain types of organizations. Like, okay, well, we're going to control all that stuff. And if you control all that stuff. Is it primarily health care and universities that want it on-prem uh yeah there's like the government um health care is a big one financial institutions are a big one uh yeah you know Aaron Francis All the like we got a lot of money or we got a lot of private data um and do they put it on do they put it on like an intranet that is just not visible from the outside world at all? And so it's like this safe cordoned off area that, that's pretty cool. Ian Landsman You have to be in their VPN in order to like access it. And yeah. Aaron Francis So obviously, obviously HelpSpot is secure, but then even then it's just not accessible at all from the outside world even better additional level of yeah to be like ([Time 0:52:49](https://share.snipd.com/snip/115799bf-0ae0-4d08-b5df-7aa13b0856bb)) - **Require Human Approval For AI Customer Replies** - Ian advised enforcing human review as the default for AI responses, letting AI draft but requiring approval for outbound messages. - He also recommended narrow triggers and category rules so AI only handles safe, preselected request types like password resets. Transcript: Ian Landsman Possibly just disappears before we go live, but right now it's optional and off is to query the history in help spot to use it to answer questions. But the thing is like that is just querying in there and like finding other people's information, right? Like all these old tickets have other people's information on them right so what we have that's a clean the clean way to do it is we have verified answers which these are the generic question And answer pairs and other information that have been stripped of any personal information uh but you know it's going to have access to other data systems you know limited in the beginning More later um but you know it's kind of interesting we built this before we've had a whole other versions of this one which was on the rag style which was terrible didn't work very well And what we've gone to which i'm really excited about is just the way we all work every day which is tool oriented so like the ai just decides if oh i can't find let me search verified answers Oh there's nothing in there let me search this you said that i could use this website documentation on this website oh let me search that and then like whatever we you can just set up all The tools and it can search all different things you let it be smart about it and the smart thing is doing everything not the stupid thing is trying to figure out what to send the smart thing. And it works way better. So that's been cool. Aaron Francis Yeah, I think if I were in charge of a system like this, like I implemented this stuff that you're talking about, I implemented HelpScout. HelpScout! I would have a business rule that nothing goes to the customer without a human first. That would be, I would be comfortable. And you can run it that way for sure. Yeah, I would be comfortable saying the AI can do anything it wants except take free world action, like respond. Right. That's probably Maybe it drafts new verified answers and you have to approve the new verified answer. That's how I would run it. I'm not, I'm still as pilled as I am. I'm still not a YOLO, like let it respond to customers. No freaking way. Ian Landsman Right. Well, I think there's a world too where there's going to be like some middle grounds of like, if you get the answer from a verified answer, then you can just send it. Because you found the solution in the verified answers. Go send it and occasionally it's probably still going to make an error there right but like you're not sending anything crazy unexpected but then the other tiers are like yeah it's like No just draft and then the human agent reviews and there's like a quick way to just send it that's going to be the default um and then the other thing that's cool about how Helpshot's doing It is like it makes it harder to initially use, which is unfortunate, but like all this is kind of run through the triggering system. And like, again, you've got multiple agents and multiple triggers. So you can really dial it in. You can be like, this category of request is answerable by AI agent X. But none of the other stuff. Or these two categories have two different agents that answer their questions, but none of the other stuff. So you can really pick and choose, whereas most of the other tools are going for like, just send in all your stuff and we try to answer it all. And like, maybe we send through stuff, we determine we can't answer, but it's not like, you can't super dial it in to a very specific slices that you want the AI to answer, which to me in This early phase of AI, I think you need that. It's like, OK, like password reset questions are easy and we'll just, yes, let it answer it. But like other things are not so easy. And let's just let that go to the human with auto draft or not with auto draft. ([Time 0:55:58](https://share.snipd.com/snip/f1c7bc74-3469-40db-8b3d-cdf40945be5d)) - **Multiple AI Accounts Fueled Solo's Rapid Shipping** - Aaron and Ian described using multiple Codex/Soul accounts and usage resets to experiment rapidly, calling it the best ROI for shipping Solo and desktop apps. - They credited cheap high‑quality API access with letting them prototype, iterate, and ship features that would have been impractical before. Transcript: Ian Landsman Mean I'm still like at Fable. I haven't doing some stuff. I haven't, again, it's last week. I haven't done that much, unfortunately. But, uh, today, this morning I got into a little Fable and over the weekend, I have not even used Soul yet. So starting today, this afternoon, I should be having some time to do some Soul. I know. Aaron Francis I ripped through three accounts and like four usage resets because i have all these usage resets banked and i just like i let her rip and i load balance i load balanced across three codex Pro max whatever what were you doing i can't tell you i can't tell you um but it's very good um of course and so, yeah, I was orchestrating across, you know, three accounts. And then I had to enact four or five usage resets. And it's, I'm still just a codex boy. It's just incredibly good. Yeah. And you're on soul for all this, I assume. Soul for all of this. I will say, check your usage resets I did not know that they expire. Ian Landsman I'm trying to right now. Where do you even get them? Aaron Francis On the easiest way is just to do slash usage inside of Codex CLI. It is available on the web dashboard. All right. Ian Landsman Yeah, I can't do that right now because I have to go to my other account, as we've talked about. No, your other user account. I don't see a way to do it on here in the settings anyway, but maybe it's in Codex or something. Aaron Francis Well, you know, everything's named everything, so it's hard to even speak intelligently about it. Ian Landsman The naming is totally nuts. We shouldn't talk about that. The naming that they're doing is absolutely insane. I didn't realize. Did you realize that in the new work codex app, there's not even chat anymore? If you do the Mac app of chat GPT, there is no chat. There is no chat in the chat GPT app. Aaron Francis Listen, I don't want to sound like I know everything, but I would not have shipped that. That's crazy. What the hell are you doing? You got the smartest brains in the entire history of the universe. So, chatgpt.com. And then it says, do you want to try chatgpt work as a pop-up? And you're like, I don't know what the hell that is. Close that part. Bottom left, go down to settings. That'll do a modal. To usage usage limit resets i've got a full i've got two full on this account in particular i've got two resets that expire on august 11th my first one three resets my first one expired On july 17th which is four days from today and i did not know that and so i'm glad that I ran up against a usage limit so that I could enact that. Ian Landsman Now, I have one on, yes, July 17th, July 26th, and July 31st. I have limits that are resettable. And I guess I didn't get any later ones probably because I didn't like use it at all last week. So I wonder if they just are smart about that now and like they don't give you any credit if you haven't used it at all um but anyway i do have three all right so i gotta i gotta do yeah i gotta Aaron Francis Throw it out i gotta just waste these credits on something so i will burn it down yeah that rainforest is not gonna burn itself so you gotta you gotta let her rip and know that it's crazy. Yeah. But having three Codex accounts for $600 a month is still the greatest business ROI that I've ever seen in my life. Of course. Ever. Ian Landsman It's insane. You shipped an entire desktop app in a language you don't know, in desktop apps you don't know, in all kinds of other aspects of this thing that are crazy and you don't know. Inter-app communication and all this stuff. ([Time 1:00:12](https://share.snipd.com/snip/9d4ef665-ce45-42cd-9e71-1ccb1fa3a21e))