Learning Library

← Back to Library

LLM‑Driven Coding: Startup Opportunities

Full Transcript

# LLM‑Driven Coding: Startup Opportunities **Source:** [https://www.youtube.com/watch?v=zbz_PcNt0mg](https://www.youtube.com/watch?v=zbz_PcNt0mg) **Duration:** 00:06:25 ## Sections - [00:00:00](https://www.youtube.com/watch?v=zbz_PcNt0mg&t=0s) **LLMs Spark a New Coding Frontier** - The speaker explains that, like Amazon and Shopify opened downstream ecosystems, large language models are now lowering the barrier to software development, creating a surge of aspiring coders and opening a wave of startup opportunities for building production‑grade applications. ## Full Transcript
0:00you know whenever we have a major 0:01breakthrough technologically we tend to 0:03unlock a stream of Downstream use cases 0:07so for example when Amazon figured out 0:10how to sell Goods over the Internet we 0:12unlocked a host of startups and 0:15businesses that figured out how to sell 0:17on 0:18Amazon when Shopify did the same thing 0:21e-commerce exploded because tools were 0:23easier for people who were good at 0:24making and selling things but were not 0:27necessarily good at internet sales 0:29because Shopify took that problem away 0:32we now have the same thing with llms en 0:36coding so llms are just at the point now 0:40where it is becoming popular to use llms 0:44to code up an app and I have talked with 0:48experienced developers I've worked on 0:49coded Enterprises I know that coding an 0:52app in 30 minutes in cursor with an llm 0:57or in repet is not the same thing thing 1:00is building production 1:02code and one of the things that I think 1:04is really interesting is that I don't I 1:08don't infer from that that we will not 1:11be getting llms who are capable of 1:14getting to production code eventually 1:17what I infer from that is that we have a 1:20host of startup opportunities that are 1:22on the horizon now we have for the first 1:24time got maybe 100x more people 1:27interested in coding over the next 1:29couple years years than we have had ever 1:31before and we have a lot of developers 1:33coding now but because the coding bar 1:36has dropped to zero because my 1:37eight-year-old can code in English 1:40now we have options for getting people 1:45engaged with the profession of 1:46engineering that we just never 1:49had and that presents a whole host of 1:52Downstream challenges because the things 1:54that people assume developers had to 1:56know to be developers is it's just no 1:58longer true 2:00and so we have to sort of think through 2:02and say what does it mean to be a good 2:05engineer writing production code we kind 2:07of know what that looks like now I'm not 2:10saying it's going to change in the next 2:1190 days because these things do take 2:14time but I think there is a whole host 2:16of billion dooll startup ideas around 2:19taking this idea of simple English 2:22instructions to drive code and figuring 2:24out how to take what is essentially a 2:27prototype you can knock together in a 2:28few minutes and convert that into 2:31something that is structured that is 2:33scalable that is sustainable from an 2:35engineering perspective and that fits 2:36well into your existing code 2:39base massive opportunity there there's 2:42opportunity for easier deployments 2:44there's opportunity for easier uh and 2:46more translatable uh error checking if 2:49you're going through and sort of doing a 2:50poll request review there's opportunity 2:53for easier 2:55um Edge case finding actually I think 2:58that's one of the things that is a 2:59product person person I really it drives 3:01me nuts like we will get through the 3:02poll request we'll get to code complete 3:04and then something that was an edge case 3:06that should have been in the technical 3:07plan from the start pops up because we 3:11just didn't anticipate that the code 3:12would touch this particular corner of 3:14the experience in this way because of 3:17this unplanned interaction and now we 3:19have to go back and fix that bug now we 3:21may have to refactor a bit it will add 3:23two weeks I have been in those shoes 3:25multiple times in my career in fact 3:27almost every quarter I get into 3:29something like that because frankly code 3:31is hard holding a mental model of code 3:33is really hard and this is going to make 3:37it easier like llms longterm are going 3:39to make a lot of that easier they are 3:41not making it easier today and I think 3:43one of the interesting thing is we are 3:45we tend to look at the present State and 3:47say this is what it's going to be like 3:49llms do not write good production code 3:51today I don't blame any developer who 3:53says this is spaghetti code this is a 3:55disaster this is not something I want 3:57touching my production environment 4:00fine it will be it will be and I think 4:04one of the reasons it will be is that 4:06what a developer looks like is going to 4:08change gen Alpha my kids age they're 4:12going to come up and they don't care 4:15what we used to call being a developer 4:17frankly they don't care what we used to 4:18call being a product manager they're 4:20just going to come up and say I can 4:21build this I'm going to build it and 4:23that massive expectation from that 4:25market is going to drive startups and I 4:28don't think we'll have to wait until 4:29until they get to be like 5 10 years 4:31from now until they get to be ready to 4:32enter the job market I think it's just 4:35going to be in the next couple years 4:37we're going to see a bunch of startups 4:39crop up that figure out how to take 4:41these increasingly smart llms and get 4:43them to play nicely with production code 4:45architecture so that they act as 4:47effective agents and they write within 4:50the expectations of a technical 4:53environment that is ready for 4:56scale so we will see 4:59but to me when I look at the current 5:02crop of apps and by the way I don't know 5:04if you've noticed this is a little 5:05Easter egg for those of you sticking 5:07around all those guys that did like I 5:10can build an app in 30 5:12minutes they didn't do I can sustain the 5:15app I can bug fix the app they didn't do 5:19I can show you how to make this app go 5:21to you know a million people served and 5:24I can add V2 features it's always just 5:26the 5:27Prototype because that's what llms are 5:29good at right now and I just noticed 5:30that because like there's this Rush of 5:32people doing this in repet and incursor 5:34in the last two weeks and it's kind of 5:35dried up it's like a trend right but the 5:39building will continue people will 5:41continue to discover coding this way and 5:43I think the opportunities for startups 5:45to actually get from this rough idea of 5:48English to code and move from that to 5:52refined code structures that work well 5:56with your existing production 5:58environment that's a YY idea in fact 6:01that's probably something people are 6:02already building if you know someone 6:03building that already you can just pop 6:05them down under the comments um and I'm 6:08really excited to see it because I think 6:10that this is going to be one of those 6:11things where we're going to have you 6:13know 100x more builders in the space and 6:15we're going to have a much much richer 6:18software environment as a result so we 6:20will see but development is changing