YT YouTube2Text

YouTube2Text → Transcripts → Your Software Won’t Matter in 5 Years (Here’s What to Do Instead)

Your Software Won’t Matter in 5 Years (Here’s What to Do Instead) — Transcript

by Frank Nillard · 2,429 words · 399 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Someone recently just used Claude to
0:02rebuild the entire Adobe app suite and
0:04gave the apps away for free, which is
0:06insane.
0:07And the thing is that this is not just
0:10happening with Adobe. There's this wave
0:13going on where people are using AI to
0:15reverse engineer all kinds of apps and
0:17devices. With this one guy even cracking
0:19open his Samsung TV and finding in its
0:22source code that it was detecting and
0:24counting [music] things like the amount
0:25of times babies were crying or dogs were
0:28barking. So, if you're the internal team
0:31that has built this part of the code,
0:33man, you got to calm the down. And
0:36so, after seeing what has been going on
0:38these last weeks, it really got me
0:40thinking if software can be copied that
0:43fast,
0:44then [music] if you're building
0:46something with code, well, the code
0:48itself is not
0:49the thing that is protecting you
0:50anymore.
0:51And so, if you're building something
0:53right now or planning to, I made this
0:56video to just share my insights on where
0:59I believe the remote are to build a
1:01startup or a digital business that is
1:03defensible in this weird [music] phase
1:06that we're right now.
1:07>> [snorts]
1:07>> Let's get into it. So, I've been
1:09building software with AI since before
1:12AI agents and rag were even a thing. And
1:15[music] yeah, you'll you'll see that
1:17first people sold prompts. Then custom
1:20UBTs, then automations with N8N, then
1:23vibe coding came in 2025, and now skills
1:26in 2026, right? And every time startups
1:29followed the same path, right? You you
1:31see first those wrappers on top of GPT,
1:34then the tools that took over one
1:36workflow, one industry, and then trying
1:39to own every workflow in that industry.
1:41And now everyone is building an company
1:43brain or AI agents to automate [music]
1:45all of your work. And when everyone
1:47rushes the same way, you know, you need
1:49to stop for a second and perhaps think
1:51that usually a sign that this is the
1:53next thing to get eaten. And a lot of
1:55these startups only existed because
1:57[music] the models had limits if you
1:59think about it. People built whole
2:01products so AI could read a PDF until
2:03ChatGPT added it as a feature. And so,
2:06every time the model improves, those
2:08gaps close and the startups filling them
2:11disappear. And so, if you build in
2:15>> [music]
2:15>> the AI space by chasing trends, okay,
2:18and specifically features,
2:20it's honestly a never-ending race, okay?
2:22And the answer to that isn't to run
2:25faster. It's [music]
2:27honestly to look 5 years or 10 years
2:29ahead, taking the leaps,
2:30and just build straight towards that and
2:32take that bet.
2:33But,
2:35you need to understand what what's
2:36actually dying right now as well.
2:39The tools that only do one thing, okay?
2:42A lot of them are now just a skill
2:43inside Claude, as you you probably have
2:45seen.
2:46Or a set of hardcoded business rules
2:48that, you know, break [music] the moment
2:50a better model can reason through the
2:52job.
2:53And funnily enough, the the the Adobe
2:56guy that I was mentioning before proves
2:58it itself, okay? A few days after it
2:59went viral, he said, "It's not anywhere
3:01close to ready."
3:02>> [music]
3:02>> His own road map puts it at around,
3:04well, 45% ready for real work and says
3:07the Photoshop files it saves can't be
3:09reliably opened in Photoshop yet. And
3:12I've seen this up close. I'm building
3:15this logistics tool for a few companies
3:17around the area, and recently one of my
3:19guys flew over to Portugal for a pilot
3:23meeting with one of these guys that at
3:25some companies where we're going to
3:26pilot them on.
3:27And he told me that one of the people
3:30there,
3:31who was a bit tech savvy, pulled out
3:32Claude and made an artifact of what we
3:35were building
3:36>> [music]
3:36>> right there in the meeting. And so, now,
3:38if you've ever made a Claude artifact,
3:41you know it's very different from real
3:43software. It just looks nice, it looks
3:45pretty on the surface, but real software
3:47is honestly a different beast.
3:49It has to handle real data, work every
3:51single day, and keep improving with
3:53well, the people that are testing it,
3:54right? Fixing bugs,
3:56acting on feedback, talking to
3:58customers. So, the realistic side of
4:00things is that you right now you can't
4:01one shot an entire software product with
4:03a prompt. Not today.
4:05Well, it's a good way for wake-up call
4:07to realize that this is the worst AI
4:10will ever be. It just only gets better
4:12from here.
4:13So, you end up realizing that the
4:15software alone isn't a moat anymore. And
4:17then, you know, we come back to the
4:19question. So, when anyone can open cloud
4:21code and build software, what actually
4:24protects you? I see six things. And
4:27notice that none of them are are the
4:29code. So, number one
4:31is obsession, right? Remember, on top of
4:34all of this you're building a company,
4:36okay? You're potentially bringing in
4:38some teammates, you're accepting
4:40payments, you're
4:41>> [music]
4:41>> accepting users, right? You know,
4:43turning it into a real business. So,
4:45this sounds soft, [music]
4:46the obsession thing, but it's the
4:48foundation. Everything else on the list
4:50takes years to build.
4:52And
4:53you know, as I said, you're eventually
4:55going to be building a business if you
4:57get some traction.
4:58And so, in something like this, you will
5:00quit before, you know, you get to some
5:02minimum level of success if you don't
5:05generally care about the problem. So,
5:07you really need to think about picking
5:08the thing that you still [music] find
5:10interesting in the third year. Then, we
5:13go on to number two, which is sell the
5:15result and not [music] the tool, right?
5:17And this is probably one of the ones
5:18that
5:19are the bigger highlight out of the six.
5:21>> [music]
5:21>> Nobody wants software anymore. It's
5:23getting increasingly harder. They want
5:25the problem to go away. People are a bit
5:28fed up of just presenting them with
5:30tools.
5:31And some AI startups are taking this all
5:34the way. So, instead of selling a tool
5:36to law firms, they're becoming the law
5:37firm, right? Doing the whole job with AI
5:39underneath and keeping the whole
5:40revenue, not just the small [music]
5:42software budget, and delivering the
5:44outcome, right? That law firms want. And
5:47you don't need to become a law firm to
5:49do this, right? The simplest version
5:51looks like this. So, on the front, you
5:53look like [music] an agency. You sell
5:54the outcome, you know, more leads, more
5:57invoices handled, you know, ads
5:59delivered, whatever the result is.
6:01And then you charge like a service
6:03because the client is paying for the
6:04result, not for a login. But at the
6:06back, you're running on your own
6:08software and AI, right? Every time you
6:10do the job, you automate a bit more of
6:12it, bit more of it, bit more of it,
6:13until you essentially create this
6:16well-oiled machine where most of it runs
6:18on its own. And if you're planning on
6:20exit this something like this, well, you
6:22know, a self-serving
6:24agency can go for
6:26>> [music]
6:26>> big multiples.
6:27And so, you get the best of both worlds.
6:30You charge agency prices and you scale
6:32like a software company. And you might
6:35think I'm crazy, but this isn't just my
6:37theory. If you check Y Combinator, Y
6:39Combinator calls these full-stack AI
6:42companies.
6:42>> [music]
6:43>> And there's a venture firm called
6:44Foundation Capital, you know, you can
6:45search it up, that calls it service as a
6:47software. So, we're flipping the whole
6:48paradigm around. They put the
6:50opportunity on the services market,
6:53which is valued at $4.6 trillion because
6:56it goes after what companies spend right
6:58now of people [music] and outsourced
7:01services, right? Not just the software
7:03itself. And yeah, then we have number
7:06three, which is the data you earn by
7:08actually doing the work manually, okay?
7:11Because everybody has the same models,
7:13you know, Claude, GPT, Gemini. And they
7:16work on a general understanding of the
7:18world, but you know,
7:20you need to understand what happens on a
7:23specifics in the workflows, the edge
7:25cases, with which customer, what worked,
7:28what didn't. That ends up being the
7:30moat. And here's the thing, someone
7:32who's working in the industry for 10
7:33years probably knows it better than you,
7:35right? With Claude Bot, they could build
7:38a first version for their own company,
7:40but they only see their own company's
7:42data. So and so if your tool runs across
7:4510 companies, he starts seeing patterns
7:47across all 10. So nobody inside one
7:50company can get that. So, you know, you
7:52need to do the work yourself.
7:54Number four,
7:55connections in the industry. Not
7:57necessary, but very important. This
7:59decides how fast you get to market.
8:02Being in the room on on sales calls, at
8:03conferences,
8:05with people who trust you. Right? A
8:07competitor could copy your software in a
8:09weekend, but they can't copy two years
8:11of relationships.
8:13Number five,
8:14community. People who learn with you,
8:17give you feedback, [music] stick around,
8:18right? They make your product better and
8:20they're hard to steal because they're
8:21not there for the software. You'll want
8:23them to eventually be there
8:25for each other and for you.
8:28And then you have number six,
8:29>> [music]
8:29>> which is distribution. I'll put this one
8:32at the top if not the first one. Okay,
8:33it matters a lot.
8:35Because as everybody can build now,
8:38building anymore isn't a moat because
8:40one viral video
8:42can give it to [music] your competitor
8:43overnight, right? If your competitor
8:45posts a viral video and you don't, well
8:47all the people are going to go to their
8:49product. And yeah, distribution is the
8:51thing that gets you the shots,
8:53but the other five are what makes,
8:56>> [music]
8:56>> you know, the thing last, whatever
8:58you're building last.
8:59And now,
9:00if you want to think about this
9:02properly, there's one question that
9:03matters more than anything else, which
9:05is how
9:06will people and businesses be using
9:08digital tools in five to 10 years?
9:10Because that's eventually what you
9:12should be building towards, not today's
9:14habits necessarily. You have to have
9:15also that long-term vision, long-term
9:18understanding on how people will consume
9:21and use your whatever you built. And at
9:24the same time, understand what's
9:25happening at right now.
9:26And you can already see the shift. I'm
9:29just giving you an example. So most of
9:30the emails I send now,
9:32I write with Claude. And a lot of the
9:34replies I get were clearly written with
9:37Claude, 2. So, it's basically AI talking
9:39to AI with me approving. Now, take that
9:42to shopping. Instead of clicking through
9:44five pages, an an AI agent in your
9:47WhatsApp gives you a few options, asks
9:50what [music] you like, picks one, and
9:51buys it.
9:52And you end up never opening a website,
9:55okay?
9:55And that's why [music] also, another
9:58tip, it's important to be at the edge of
10:00things. Because if you see what people
10:01are doing at the edge,
10:02well,
10:03these are like 0.01% of the population.
10:06You are somewhat predicting the future
10:08in some sense [music] by being at the
10:10edge, at the bleeding edge of
10:11technology. And here's what I believe
10:13this means for builders, okay? Right
10:15now, everybody's building apps, but a
10:18lot of my actual work already happens
10:20inside Cloud,
10:22not inside the apps. I barely go into
10:24the apps.
10:25Everything is working right now inside
10:26Cloud, or most of it. So, the businesses
10:29that win, I believe, are not going to be
10:31the ones with the prettiest interface.
10:34They'll be the ones an agent can reach
10:36and use through some API or an MCP,
10:39okay? And so,
10:40also, I want to give you a quick
10:41exercise. Look at your normal day, okay?
10:43Your emails, your shopping, your
10:45bookings, your admin. How much do you
10:46think of that will an agent be doing for
10:48you in [music] 5 years? And which
10:50businesses will would that agent need to
10:53talk to? That's a little glimpse of
10:55where some of the opportunities might
10:57be, okay?
10:58And so, if I was starting from zero
11:00today, here is the order I would follow.
11:02First one,
11:03you pick an industry you're generally
11:05obsessed with that you already
11:07understand because
11:09or that you're willing to spend some
11:11years learning at incredible speed. Two,
11:14go and sit with the people in it. If
11:16you're one of them, that's a good start,
11:17but you also need to
11:19reach to other people similar to you in
11:21that network and find the problem that's
11:23causing them the most. Number three
11:25is solve it,
11:27okay?
11:28Try to avoid not solving it with, you
11:32know, just [music] just building an
11:33automated pipeline or whatnot, right?
11:34Try to solve it first scrappily by hand
11:37with,
11:38you know, okay, perhaps a bit of flow
11:39code, right? Whatever, but don't build a
11:41proper solution first. But just get it
11:43working. Whatever is quicker. Number
11:45four, get paid for it as a service
11:47ideally, right? You want to sell the
11:49outcome and not the software cuz money
11:51right now is the only real proof. And
11:53you can charge really quick if you
11:54charge as a service.
11:56And five, while you do the work, collect
11:58the data and build the relationships.
12:01Only then you'll turn it into something
12:03like software.
12:04And notice what that gives you. By the
12:07time you build the software, you already
12:09have some customers, you have some
12:10connections,
12:12the data, okay, because you're [music]
12:13doing things manually, and hopefully you
12:15have that obsession as well. So, then if
12:18the next model makes your software
12:20[music] 10 times easier to copy, well,
12:22doesn't matter too much anymore because
12:23the software
12:25was never what protected you. So,
12:27yeah,
12:28main take of this is that, you know,
12:30stop building, [music]
12:31you know, basic AI wrappers
12:33or basic workflow wrappers. The
12:35advantage isn't anymore who can code the
12:38fastest.
12:39>> [music]
12:39>> It's who understands the problem the
12:40deepest and who people trust to solve it
12:43and who gets to to own the distribution
12:46quickest.
12:47>> [music]
12:47>> And yeah, if you want to hear more about
12:49this, let me know in the comments. Tell
12:51me your thoughts, honestly. Hope this
12:53helped. Thank you so much for watching
12:55and I'll see you in the next one.

About this transcript

This page contains the full transcript of Your Software Won’t Matter in 5 Years (Here’s What to Do Instead) by Frank Nillard, generated from the public captions YouTube serves with the video. The transcript has 2,429 words across 399 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.