Luyện nói tiếng Anh bằng Shadowing qua video: The Only SaaS Moats Left in 2026
Chọn chế độ luyện tập
Mỗi chế độ mở một phòng luyện tập toàn màn hình.
623 câu
Nếu các câu quá ngắn hoặc quá dài, hãy bấm Edit để chỉnh sửa.
1
How the f* do you survive, if you're a SaaS company, how do you survive in this age of AI?
2
AI is more of a problem probably for every SaaS founder in this room than you think.
3
You don't, I could start one today and probably have pretty close feature parity in a few months.
4
They don't need your 50 features, they only need the three that they use.
5
If you don't agentify your software, build some type of agent on top of it, I think you're going to zero in the next couple years.
6
Because someone's going to come along and they're going to, QuickBooks is terrible software, I hate using it.
7
but I don't care because my accountant uses it and he likes it.
8
You're probably lucky.
9
You're probably doing pretty well if you have one or two of these.
10
I am going to try to scare you a little bit with this talk.
11
This is the, I think I have 91 slides, 92 slides.
12
It is by far the densest talk I've ever given.
13
This one has a lot of gravity to it.
14
You'll notice it's also the, I think the only talk I've ever written, co-written with someone.
15
So it says by Rob Walling and Anner Volset,
16
because Anner Volset spent about a month wrestling with 89 different papers written by think tanks,
17
private equity groups, Gartner, mashing them with Claude, Chachypt, going back and forth, trying to figure out how the f* do you survive,
18
if you're a SaaS company, how do you survive in this age of AI?
19
Because what SaaSpocalypse tells us, tells us on social media, is all SaaS is going to zero.
20
So we can say, well, that's not true.
21
It's just not true.
22
It's objectively not true.
23
But is it going to go back to normal?
24
Are we okay?
25
Should we ignore it?
26
That's not true either.
27
So there's something in the middle, and that's what we're trying to find with this.
28
So where we're going today, normally I have three acts, but today I have four.
29
I want to try to get you a little scared.
30
AI is more of a problem probably for every SaaS founder in this room than you think.
31
I don't want to scare you too much, but just enough to pay attention.
32
I'm going to talk about classic SaaS motes, what I mentioned in SaaS playbook, if you've read that.
33
I have a few motes, and how some of those are dying or have died.
34
then talk about what we think are the modes that still work
35
and then I have like three or four slides
36
that is really could be a whole four-hour workshop on how
37
do you evaluate your SaaS company what questions do you ask I tack it on at the end
38
because it's important but can't possibly cover that in a 45-minute talk
39
so this is a sentiment that we're hearing on the internet microconf tiny I'm a founder, SaaS founder.
40
AI is amazing.
41
I'm shipping so much faster.
42
And obviously, my customers are hairstylists or lawn care specialists.
43
They're never going to vibe code my product.
44
Okay?
45
So that is an attack.
46
That's one attack that AI can make against you.
47
Your customer specifically vibe codes your product.
48
So I'm going to now show you 10 attacks, 9 additional ones, several of which you probably haven't thought of.
49
They're grouped into two buckets, the supply side and the demand side.
50
Okay, so basic economics, if we have a widget and suddenly the supply of widgets goes way up, the price of widgets go down,
51
commodity, they get commoditized, you lose as a company, you lose as a founder, right?
52
Well, SaaS can be the same thing.
53
Imagine you have a single SaaS application that does scheduling links, and suddenly there are 99 others.
54
There's a huge supply.
55
It becomes very messy.
56
So when the supply gets shocked, that's a problem.
57
On the other side, demand is, well, these are my customers, right?
58
People buying.
59
Well, what if I have a million potential customers in the U.S and their businesses, and let's say 90% of them go out of business tomorrow.
60
Suddenly, my demand plummets.
61
This is bad.
62
Both of these are bad.
63
So, I'm going to cover 10 across supply and demand.
64
I'm going to break them into those buckets.
65
The supply shock has six different vectors that AI can come after you with.
66
The thing to remember is everyone gets AI, not just you and not just your customers.
67
Everyone gets AI.
68
So, the first attack is the one we all think about.
69
Your customer builds it, right?
70
Some people say, my app's pretty complicated.
71
It has 50 different features.
72
You're going to get the slides at the end, by the way.
73
50 different features, they're never going to vibe code it.
74
One thing to think about is they don't need your 50 features.
75
They only need the three that they use.
76
All right.
77
Second attack is competitors reaching parity very quickly.
78
This slide shows eight of the 46 different AI note takers that I could find on the internet.
79
and it's, what is it, Otter, Fireflies, Crisp, Fathom, TLDV, Read AI, Circleback, Granola, and I wanted to show them in this order because,
80
you know, you think about your Otter in 2016, and you're like, oh, I got this all to myself.
81
This is amazing.
82
And now we have Granola, and everyone in between who almost, they almost all have feature parity.
83
They're kind of the same product, and I know that would piss off the founders if they were here,
84
but you don't, I could start one today and probably have pretty close feature parity in a few months.
85
That's great for an upstart and it's really bad for your supply.
86
It's a supply shock that everyone can get to feature parity with most apps these days.
87
It buys you, features used to buy you a lot of time, they buy you a lot less time.
88
So attack number one was customers, attack number two, competitors.
89
Attack number three is open source.
90
I have already seen this.
91
There are more open source projects, there are more open source projects working, moving faster than they used to because, if I'm coding, everybody gets AI.
92
They're moving faster and non-technical buyers who used to stay away from open source,
93
they can deploy it because they talk to cloud or cloud code and they can run open source.
94
None of these causes saspocalypse.
95
Ten of them together makes a difference.
96
My estimate, and this is just me doing this for 20 years and taking a guess, said this multiple times.
97
I think somewhere, SaaSpocalypse says everything is going to zero and everyone is plateauing.
98
That's not happening.
99
That's not true.
100
I think somewhere between 10
101
and 20% of all SaaS companies in the world are either going to decline or go to zero.
102
That's a lot.
103
It's a lot and it's not.
104
Does that include little indie hacker projects?
105
Probably.
106
That's still a lot, right?
107
10%, 20%.
108
That gives you an idea.
109
Fact number four is a platform, like a system of record that you're tied into,
110
ships your feature for free because they can build it in a week, a month.
111
They can build it a lot faster.
112
They already have the customers.
113
You've basically given them a spec through your UI.
114
They could feasibly, if you're an add-on to something, they could go out.
115
I'm imagining if you're, well, anything that ties into a CRM isn't Salesforce and HubSpot
116
and Close and any other CRM is just going to start building all the ancillary things
117
because they're just easier to build now.
118
So your platform.
119
This is like an integration partner that's going to come for you, I think.
120
Track number five, we're seeing this.
121
Custom software at SaaS prices, approximately.
122
If you're an agency, you used to charge 30 grand, 50 grand for a build because it took three months.
123
Now you can vibe code it.
124
You can code it with AI assistance in a third, fourth, fifth of the time.
125
And you always want it.
126
Agencies want recurring revenue, so if I'm charging $99 a month for mine, my one size fits all,
127
agency charges, whether it's $299, $499, $999, whatever it is, completely custom, made for you, supported for you.
128
This model is becoming a thing.
129
In fact, if I were in my early 20s, I would think about something like this.
130
This is a really interesting business model.
131
It's still an attack on SaaS.
132
And the sixth supply side is if the model absorbs your product.
133
This may or may not happen,
134
but you remember when we used to pay apps to generate like ad copy and headlines and images and all that stuff?
135
Those apps went away, and there's a bunch of examples of others that have as well.
136
All right, so that's the supply.
137
Demand shock, there are four of these.
138
this demand shock I find and
139
when Anders started bringing these to me I was like oh I hadn't thought of that
140
and most SaaS founders
141
that I've talked to as I've been writing this talk are like f* I didn't even think about that
142
what if your customers just go out of business 50%,
143
90% whatever the number is huge swaths that AI wipes them out so imagine that your SaaS serves translation agencies
144
or transcription agencies that used to have a human sitting there listening to something.
145
The number of words translated and transcribed since AI have gone through the roof, right?
146
There's a lot more volume of this happening.
147
The revenue of these agencies is effectively going to zero because we are basically using AI.
148
Now, one can argue there's going to stay some boutique agencies if you're doing medical transcription, I'm sure if you're, or translation,
149
you know, there's certain areas where this is critical.
150
But this ecosystem is going to take a big hit.
151
I mean, I say 10 to 20% of SaaS is going to experience a problem.
152
80%, 90% of these agencies are going to zero.
153
There's big disruptions in copywriting as well.
154
There's a lot of things.
155
So while the activity grows, your customers' ability to pay you plummets.
156
If your customers start going out of business, there's no recovery from that, right?
157
In terms of that's just involuntary churn.
158
Act number eight, are your seats going away?
159
This one we've heard, remember when Saspocalypse came up and they're like, seats are going away because agents are gonna do everything.
160
Yes and no. In some businesses,
161
seats are gonna have a problem because if you are Zendesk and you are customer service or customer support software,
162
I think when Derek and I left
163
Drip, we had six or eight customer support and we had six customer success.
164
Now, so that was about 12 people.
165
Customer success these days still can't, is not agentic yet.
166
Customer support is 80, 90% automated in a lot of things.
167
So suddenly you have go from eight agents to maybe one or two that is supervising a bunch of,
168
yes, eight human agents down to one or two AI agents, and they're just supervising them, right?
169
So you have massive contraction of seats.
170
And you'll see, you see Zendesk and Help Scout and these folks moving towards outcome-based.
171
There's a bunch of things they're doing with pricing to try to get around this, it's still a big shock.
172
Attack number nine of ten, does the thing you manage still matter?
173
Are people using it at all?
174
Or does it just go away because of AI?
175
Whether it's agents?
176
Imagine that you were software to manage take-home coding tests.
177
We don't do those anymore, right?
178
The whole category of software, managing take-home code, because if you give someone a take-home coding test, what do they do?
179
Well, they're going to use AI for it, but you want to use AI,
180
but how much did they do versus how much did the AI do?
181
So Derek and I were just talking, you mentioned like six times in my talk, I guess.
182
You and I were just talking on the podcast a couple weeks ago, talking about hiring, and it's like, well, we do pair programming now.
183
That's what I would do, right?
184
You're not gonna, I'm not gonna do a take-home coding test.
185
So some people in this room impacted by it, others not, it's still an attack.
186
The tenth and final one is, it's related to seats, but it's different.
187
Imagine if you were software for receptionists, okay, or like, almost like call centers,
188
where they are going to be replaced by AI agents.
189
So you have a bunch of seats, there's a bunch of humans, and now if you, and you have software you're selling for $50 a month.
190
If you don't agentify your software, build some type of agent on top of it,
191
I think you're going to zero in the next couple of years.
192
Because someone's going to come along and they're going to say, yeah, you want that software plus like a fully autonomous,
193
let's say 90% accurate agent for $300, $500 a month when you're paying $50 a month, but you can let the receptionist go,
194
or the call center people, they're going to pay the $300 to $500 or whatever that is.
195
So is the human even using your software anymore?
196
Related to seats, but it's different.
197
So that's the part that gets you scared, to make you think, you know, I need to pay attention.
198
Because even if your business is doing okay today, and you're bootstrapping, and you just want to run it forever,
199
some of this stuff's coming for you now, and some of it's coming for us over the next few years.
200
All right, so we're going to talk about some, I'm going to zip through this part.
201
This is like classic moats that I've thought about, that we've talked about in SaaS traditionally, why customers used to stick around.
202
I came up with nine reasons.
203
I saw this slide this morning, and I realized, yeah, there's probably eight, but whatever.
204
It doesn't really matter.
205
So, it used to be, I have unique features, they're not going to leave because they need these features.
206
I have a real strong brand they're going to stick with me for or continue to buy for me for.
207
I have these owned traffic channels, I have SEO that I've been doing for years, clever code, complex algorithms, five years of honing,
208
refining, working around buggy APIs, whatever else.
209
The team is familiar with my UX, they don't want to retrain people, so they're not going to switch.
210
I have 30 integrations, man, those are complex to build, aren't they?
211
High switching costs, it's hard to get my data out,
212
I can't really export and then configuring in the new workflow thing on the other place is a pain in the butt.
213
If multiple people are in my app coordinating, similar to retraining, it's like everyone has to move at once and we cut over on this day.
214
And then I'm a system of record where all this data comes in and it's complicated to move, right?
215
reasons I came up with, there are probably more.
216
And we're just going to check a few of these off, shall we?
217
So what does AI do?
218
AI makes building features, doesn't bring it to zero.
219
I don't want to act through this talk like suddenly I can build a year's worth of software in a week.
220
That's not true.
221
But it dramatically reduces the moat or the barrier to just rolling out new features that competitors have.
222
Feature parity, I think products are going to all kind of converge because whoever's really good at product and UX,
223
people are just gonna wind up copying them with AI.
224
Own traffic channels.
225
This one I'm 50-50 on.
226
Yeah, if you have really good SEO, it's fine.
227
Is AI making marketing and just like it's making kind of starting to trivialize code, is it starting to make marketing really muddy?
228
It is.
229
Clever code.
230
I think we can just admit that someone can probably, now that AI is solving 100 math proofs.
231
Anyone hear about this?
232
There's like 100 math proofs that haven't been solved ever in human history, and they're saying inside it was Anthropic or Gemini or somebody.
233
They just solved 100 of them.
234
And Anar and I were talking, and I was like, well, glad I'm not getting my PhD in math.
235
It's like, what do you do anymore, right?
236
AI's doing it.
237
UX familiarity, yes, if it's still humans.
238
If things go agentic, your UX doesn't matter at all.
239
Integrations, isn't the API really just the spec to hand to my bot now to f*ing code the thing.
240
No, I'm getting depressed.
241
All right, and high switching costs used to be a thing.
242
They still are a little bit, but man, have you seen someone point an AI at an app and say,
243
pull out all my data even when there's no export?
244
Yeah, it really can and does happen.
245
So of these nine, I think three are still, like system of record is a real thing.
246
You're going to see it in this list here.
247
Coordination between people is a thing and a brand plus some stuff is a thing.
248
So we have these three, I'm gonna add five more and we're gonna dig into those now.
249
This is the meaty, meaty part of the talk.
250
So for each of these, hey, real quick, if you're getting value out of this talk, you should subscribe to the MicroConf YouTube channel.
251
We're posting all the talks from MicroConf Europe.
252
So even if you couldn't make it to Iceland, you can still get a lot of the learnings from the event.
253
We also have a back catalog of some amazing talks from our previous events.
254
All right, let's dive back in.
255
I'm gonna talk about them.
256
I'm going to say what the mode is.
257
I'm going to give at least two examples for each one.
258
And that's probably the important part of the talk.
259
I tried to give a big name example we all know.
260
And tried to do a smaller, like, microconf tiny seed company.
261
Just because I don't want to just give, hey, you should be Slack, Facebook and Google.
262
Like, this is not helpful, right?
263
So I tried to do a big one, a small one.
264
And then there's a question at the end to help you almost evaluate your own product in the context of this.
265
Something I didn't say at the beginning is that the M&A market, one of the reasons Aynar,
266
so Aynar came at this problem from private equity and the M&A market.
267
Because a big part of what makes the whole SaaS ecosystem work and so valuable is that private equity,
268
which does 70% of the acquisitions of our types of SaaS companies, they have been paying incredible multiples for the past six,
269
eight years, maybe a little longer.
270
and they follow public markets.
271
So as HubSpot or Atlassian move, so does private equity and these multiples change.
272
Even if you never plan to sell, this ecosystem works because a lot of money has been flowing into it.
273
If you don't have any of these moats, what A&R has been seeing at Discretion Capital, right?
274
He runs a sell-side M&A advisory for two to $20 million SaaS companies.
275
What they've seen is that private equity and strategic buyers as well are really scared,
276
not about the present, but they don't know if you're going to be around in a year, three, five years.
277
So he came at it as what can you have in your business
278
that gives them confidence that you're going to be around in three to five years such
279
that they will pay our old multiples that we've talked about all this time.
280
I come at it from more of the founder side.
281
He comes at it from kind of the finance, private equity side.
282
And that's why these modes are important.
283
Even if you never plan to sell, even if you're bootstrapped, these are defenses against AI.
284
And the other interesting thing, I had a couple slides I pulled out,
285
and it's about Atlassian and HubSpot within a day of each other in August, it's about five weeks ago, they reported their earnings.
286
And HubSpot beat earnings, beat almost every expectation, did a billion dollar stock buyback, and their stock dropped 21%.
287
and Atlassian had similar results but their MCP server had like doubled in use, it had a bunch of like AI was making it better,
288
they went up 35%.
289
It was a huge swing, like imagine you know tens of billions of dollars moving around
290
and when we look at it it's like well Atlassian
291
and Jira specifically has several of these moats and HubSpot's inbound marketing engine is probably going to, I wouldn't say wiped out,
292
is massively negatively impacted by AI.
293
So we're already seeing it in public markets, we're seeing it in private equity and acquisition markets,
294
and obviously we are seeing or going to see it in the bootstrapper markets as well.
295
Let's dive into the moats.
296
The first one is that buyers know and trust you.
297
I've tried to summarize each of these with this phrasing, but then there's bullets.
298
It's like, well, what does that mean, right?
299
It's a strong brand.
300
And also, you know your buyers and you know how they buy.
301
You have partners, potentially, that have built their businesses around recommending you.
302
And you have a champion inside a company that will fight for you.
303
So the true strongest form of this moat is when a competitor who came in with better software, who can vibe code it,
304
still has to build all of this.
305
All right?
306
A lot of these moats, actually all of them, they're not on or off, it's not a bit.
307
Maybe it's a zero to 10, right?
308
A zero to 100.
309
And you could say, well, Salesforce or Shopify are at 90 out of this, you know?
310
And you can say, well, maybe I'm at 10 now.
311
How do I get to 20 or 30?
312
Examples of this.
313
Strong brand, you know the buyers, partners recommend you.
314
QuickBooks is a perfect example of this huge ecosystem.
315
We could replicate QuickBooks tomorrow and we would be fighting against the accountants.
316
Like every accountant, QuickBooks is terrible software.
317
I hate using it, but I don't care because my accountant uses it
318
and he likes it and a better product doesn't get you in this conversation.
319
I mean, zero is doing okay, but what is zero?
320
10, 12 years into this?
321
Shopify is another example.
322
Their partner network is extensive.
323
There's a bajillion agencies, I think literally thousands of agencies that implement Shopify stores.
324
So yeah, hey, there it is on the slide.
325
It must be true.
326
So when a merchant, if you go to a merchant goes to hire an agency and says, what cart should I use?
327
Well, they're like, well, obviously we're going to do Shopify, right?
328
The platform's already decided.
329
The moat, two questions, I'm sorry, the moat, two examples, and now we have the question,
330
which is who's built their business or their career around recommending or defending you.
331
That's one way to evaluate this.
332
Vote number two.
333
You are a system of record.
334
What does that mean?
335
It's a source of truth within a business for data.
336
This is where daily operations are built around you.
337
It means people are logging in every day to do stuff.
338
You might be the operating system of that business.
339
That if you go down, the business is like, shit, out of luck.
340
They can barely do stuff.
341
There are systems of record that aren't just the operating system.
342
If they have a source of data that can't be easily exported and that everyone's working in all the time.
343
But I think the biggest mode here is daily operations are built around you.
344
You have balances, you have approvals, you have statuses, you have all types of stuff, workflows, living.
345
Roles and rules are configured
346
and a lot of other systems are wired in either reading from it to act or pumping data in to you,
347
and that data doesn't come out.
348
JIRA is a good example of this, right?
349
We think of it as just an issue tracker, but there is so much.
350
I mean, devs and managers and product people, you have multiple teams in JIRA, and you could export your tickets in an afternoon,
351
but migrating all the history, the audit trail, the integrations, and getting all those teams to move to a different system would be very, very painful.
352
So there is, the fact that it is a system of record just has gravity, it has weight, and it pulls, keeps you there.
353
Gymdesk is a tiny seed company.
354
He had an amazing, amazing life-changing exit.
355
And it's basically, imagine your gym, martial arts dojo, yoga studio, your entire business runs on him.
356
And it started as, I don't even remember, it was like a part of the gym.
357
And then suddenly he's like integrating with hardware so people can use like a fob to get in the door, and every aspect of the gym,
358
it just, boop, boop, boop, invoicing and billing and just kept spreading.
359
And so now if you're a gym, you get gym desk, done.
360
No one, very few are going to switch off this for any type of AI.
361
So the question is, if your customer wants to leave, what can't they export?
362
What breaks downstream and what would they have to rebuild?
363
Moat number three, this is where it starts to get esoteric,
364
where Anar showed me these and I was like, what do these mean? and I had to spend hours like thinking through.
365
You're allowed to act.
366
What does that mean?
367
So if you're a tool that recommends that a clerk makes a payment,
368
you're one type of tool versus a tool that can make the payment, that can actually move money.
369
So do you give them a spreadsheet with a list of PayPal emails and dollars?
370
Okay.
371
So you're like a report, you're an analysis, you're reporting to, or do you actually move the money?
372
If you can move the money, then you act.
373
Examples of this, Stripe Connect.
374
No one's switching off
375
Stripe Connect, doesn't just recommend a payment, it actually moves money.
376
Tiny C Company, Reambi, it reimburses, if I wanted to pull a job candidate in, fly them out, and they have expenses,
377
meals, hotel, other stuff, I don't want to set them up as a vendor in my system,
378
I want to pay them via Venmo, PayPal, Cash App.
379
Reamby does the whole thing soup to nuts.
380
The fact that Reamby can act means that it has a pretty nice moat here.
381
You don't have to be Stripe to have the authority to pay.
382
And I threw a third one in here, churn key, which is if you're a SaaS company you can put churn key in
383
and they have offers when people go to churn
384
and they they help they offer discounts
385
and they offer all kinds of different variations to try to reduce your churn
386
and there's a lot of tools that will tell you who's about to churn
387
but churn key is literally allowed to get in the cancellation flow
388
and make an offer they're allowed to act they're allowed to step in front of your customers
389
and you don't you don't even check on them
390
so question here is can you act can your your software actor can only advise.
391
Are you a reporting tool or something that actually does something?
392
Mode number four is that, this is another one, when he said it, I was like, what are you talking about?
393
They don't check your work.
394
The best example of this is Gusto.
395
I've used Gusto for 10 years at least.
396
No, no, since 2013 probably.
397
13 years.
398
And I've never once checked their work.
399
I could go to jail for IRS something or other
400
if Gusto decided to f* me because I have never once looked at anything but the W-2 they issued to me.
401
You just don't.
402
And we can say, well, it's complicated.
403
Why don't I?
404
Well, I don't know.
405
I don't understand it.
406
Great.
407
So am I going to switch when someone AI vibe codes a Gusto competitor?
408
I'm not because these guys got it dialed.
409
Another example, this is Avalara, also a tax.
410
This is sales tax, though.
411
Once again, we don't check the number.
412
This was when I had, I struggled to find like a small, like a tiny seed microconf example of this.
413
So I have two big ones.
414
And it's not just, hey, we're trustworthy.
415
They've earned this permission, right?
416
They've been right for years and years and years.
417
So if you were to, like I said, Vibecode, Gusto or Avalara, switching is not likely.
418
All right.
419
We're in the home stretch here.
420
The fifth one, oh yeah, this is the one we added the other day because it had gotten lumped in.
421
Man, you do this long enough and they start lumping in.
422
At one point we had 10 and then realized, yeah, you can imagine this, how this goes trying to build something like this from nothing.
423
So this one is that someone official already said yes.
424
And what I mean by that is this is where there are regulations, licenses, accreditation, certifications, right?
425
Anything where there's a third party
426
board or body that is monitoring these things and that you as the vendor have to qualify,
427
get certified, etc. Just because your customers do it, yeah, yeah.
428
Serving regulated customers isn't the same as holding an approval yourself and because the competitor can clone your code, again, every one of these is like,
429
alright, if cloning the code, the code let's just assume it gets faster and faster easier
430
and easier and it goes to zero much like hard drive space did internet speeds bandwidth all
431
that right it just approached zero didn't get to zero
432
but it approached zero let's assume for the sake of argument
433
code approaches zero what else are these modes doing for us
434
and um this is where you have to get this is this can be the iso 27001
435
It can be, I want to say GDPR, that's not what I mean.
436
What is, SOC 2 is what I'm thinking of.
437
But it can all, and these are just little speed bumps in the road.
438
But it can also be things like, well, I'll get to it in a second, I guess.
439
HIPAA is another one.
440
So compliance is so brutal that Vant has literally a billion dollar company and they just spring up over the past.
441
I mean, they haven't been around for more than five or six years.
442
Because these things are tough, and they take a while, they're expensive.
443
An example of a tiny seed company that works with this, and if you were to try to vibe code this exact same thing,
444
it would be very hard.
445
Because no one in here knows what ABA is, except for the founder.
446
I forget what it is.
447
But it's like an element of psychology, and it's related to that.
448
But they have courses that count towards credentials.
449
And if you try to vibe code this, you have to get board approval.
450
And are any of us going to go through that?
451
No. So she has a moat around her business.
452
Another example of this is one that I just thought of in the last couple days.
453
Sending text in the U.S is shockingly difficult.
454
And the laws are way worse than email.
455
We thought we had it bad with Drip.
456
When I deal with the text, the SMS companies, in TinySeed, I'm like, good Lord.
457
I would never start a business in that.
458
Well, that's a moat right there, right?
459
So, you have to, there's carrier registration, there's consent laws, there's all types of ways.
460
It is very cumbersome.
461
The software for this is not that hard to code, but dealing with texting is not fun.
462
So, that is a bit, you know, is it 100% no one's ever going to do it?
463
No, but it is going to be a barrier.
464
So, the question here is what approval, complexity, what legal stuff do you know, have you navigated that a competitor would have to earn?
465
Number six is that everyone's already in the software.
466
This is the idea of network effect.
467
Who has to move if they leave your software?
468
Is it one person?
469
Well, they can switch in an afternoon, right?
470
You have a whole team?
471
All right, everyone agrees, we move next week.
472
It's a whole community or a whole company all at once.
473
It's painful.
474
So the more folks that are in there coordinating, the bigger the moat.
475
Obviously prototypical example of this is Slack.
476
Some of us love it, some of us hate it.
477
We all try to move off of it.
478
No one does.
479
There's competitors to Slack.
480
We can't name them because Slack is the king.
481
But you don't have to be Slack.
482
Commit Swimming is a tiny seed company and swim team management built for coaches.
483
What they do is, and a lot of the swim,
484
there's adult swim teams, but imagine there's a bunch of kids swim teams and they get all the parents in,
485
signed up for it and it's notifications, right?
486
Here's the meet, here's the location, we're all doing this pizza party, blah, blah, blah.
487
Every family's all signed up.
488
These are all non-technical folks generally.
489
So switching doesn't just mean to move the software.
490
It's like everybody has to move all at once and the whole league has to agree.
491
Like it's a pain in the ass, right?
492
So the question here, this is a pretty, This one's very interesting and obviously hard to achieve.
493
Would a competitor have to move everyone at once is the question.
494
Number seven of eight.
495
Every customer that you have makes your software smarter.
496
You're going to hear me talk and use the phrase learning loop.
497
Anar is going to be talking about learning loops in a couple hours.
498
Example of this is Stripe Radar, which identifies fraud.
499
and every merchant that
500
Stripe has sees fraud patterns and that feeds up into a Borg-like mind that protects all the other merchants.
501
So if you were to try to go start your own
502
Stripe Radar with just your data, it's worth way, way, way, way, way less than Stripe Radar and their system gets smarter.
503
No single merchant could ever build it, right?
504
And if you leave it, you lose it.
505
The diagram here is a situation comes up, human or software takes an action,
506
you see what the result is and then you correct and make it better.
507
If you have a lot of data, it doesn't matter.
508
You have to be able to use that data to predict across all your customers
509
because that then becomes the moat because switching means you lose it.
510
Another example of this is Ramp,
511
which is bookkeeping, and every correction that one of their customers makes to a Verizon line item,
512
oh, I'm going to do that as a cell phone, whatever.
513
Well, everybody else gets that too, right?
514
That's a trivial example, but you get the idea.
515
And then churn key again, every cancellation in theory should make their engine better.
516
I don't actually, I didn't have a chance to contact the churn key guys.
517
I don't know if they're actually doing that, if they're doing a feedback loop,
518
but you can bet that I'll be contacting them after this talk and say, build that in because you should have this mode.
519
So this one only counts if you have clear three gates.
520
Number one, every customer makes it better for every other just be siloed
521
that this customer makes it better for themselves because then they can go build it or a competitor can build it.
522
Number two, customers can take their data but not what you learn from it.
523
Yeah, they can export it but they don't know what you learn.
524
And the last one is you have to keep learning, right?
525
Because if it's stagnant and you don't get better over time, then it doesn't matter, they can just leave.
526
Another Another example I thought about this is MailChimp.
527
MailChimp is the only ESP that ever made freemium work
528
because the biggest thing in email is spam
529
and they built an engine called Omnivore that figured out how to detect spam at a level, at a machine learning level,
530
this was in 09, like way before any of this AI came out.
531
They're another good example.
532
So your question is, does what happens to one customer make your product smarter for the rest?
533
And the last one that I added just a couple days ago based on this
534
Peter Lovell's tweet, it's hard to keep running.
535
So Peter Lovell's tweets, I've tried for weeks now, he's Peter Lovell's indie hacker, half a million followers on Twitter, he's just vibe coding everything.
536
I mean even like 20, 30 dollar stuff he wants to like vibe code and host it himself because that's what Peter does.
537
So he's at the extreme end of that.
538
He's trying to vibe code everything.
539
So he tried to do Scrapingbee, which is a proxy service that helps you not get blocked as you try to scrape things, right?
540
Because everybody blocks your IPs.
541
Tried to code it, tried to code it, finally he's like, it, I'm just going to pay him 99 bucks a month.
542
So if you look at his list, he said, these are the only things I'm paying for now.
543
It's Cloudflare, like raw server, bare metal server boxes and VPS and blah, blah, blah, and Scrapingbee.
544
And it's like, what a trip, these massive companies are scraping people, why is that?
545
It's because web scraping isn't that hard to build
546
but to keep it working and not get blocked is a pain in the ass.
547
It's a bunch of DevOps, it's servers, it's cycling IPs, it's a huge headache to run, it's a burden.
548
Scraping big guys sold, they had a great exit, eight figure cash exit, they've been public about.
549
But they didn't want to run it anymore.
550
It was such a pain in the ass, but that was their moat.
551
Cloudflare, same exact thing, right?
552
Are you going to vibe code to Cloudflare?
553
If someone else vibe codes Cloudflare, am I going to switch?
554
No. It's a crazy moat.
555
It's just hard to run.
556
Drip.
557
God rest our souls, Derek Reimer, that we survived this.
558
Drip was, anyone can send an email.
559
And this is why when I see people, I've seen someone like vibe code kind of an ESP, right?
560
And say, just plug in your Amazon SES and you can send your own email.
561
And yep, Benedict, you as well.
562
We just kind of chortle because it's like, no, no, no, you don't understand.
563
We had multiple full-time people.
564
All they did was try to fight spam.
565
You hook it into your SES account, like, it's a problem, right?
566
So, these are moats.
567
The question here is, is it hard to build or is it hard to keep running?
568
Because that thing that you hate and resent can actually be a moat.
569
These are the eight.
570
And now I'm going to wrap up with just a couple slides.
571
These are more of a thought experiment that we're, I mean, we're going to be fleshing this next part out into hours and hours of stuff.
572
So, it's going to be quick.
573
But I want to answer the question quickly like
574
which of these are possible for you
575
and there's eight questions each one corresponds to a moat
576
there's probably two three four questions we could ask for each of these right
577
so it's like who has built you've heard some of these who's built their career
578
or business around recommending
579
you do you hold the definitive view of the work right
580
are you a record of the work are you a system
581
of record are you just a view of it do you take action
582
or do you just recommend it do customers trust your output enough that they don't check you, what approval, this is the regulation one, right?
583
What approvals, regulations do you hold that a newcomer has to wait for?
584
Who has to agree to replace you?
585
Does one customer's experience make it better for everyone?
586
And what do you maintain that you resent maintaining?
587
That's my favorite.
588
And I think this is the last slide.
589
It is, these are just examples that I was like, for each of the moats,
590
how would I go from where I am today to try to have that moat?
591
So if I was, and these are in the same order as well.
592
So it's like, if I had a vertical tool, I wanna build an ecosystem,
593
I want the vendor, I wanna become the one that the industry recommends, right?
594
If I'm a point solution, can I become a system of record?
595
If I'm a reporting tool that gives reports, can I figure out how to act and actually move the money or actually put the churn thing,
596
not just recommend it.
597
If I have something that calculates something, how do I get to be the output that nobody double checks?
598
If I'm in a regulated industry, how do I become an approved or the approved provider?
599
We see, don't we see Anthropic
600
and all the AI providers trying to basically regulate their industry so that they pull the ladder up behind them?
601
If I'm a single seat tool, can I build that network effect such that everyone's coordinating?
602
If I collect data, can I build a learning loop?
603
And if I have infrastructure that I babysit and I despise and hate, will that be a burden that no one else will carry?
604
So in conclusion, try to figure out which ones are yours or which ones you can move into.
605
Maybe you have one of these today.
606
You're probably lucky, probably doing pretty well if you have one or two of these, I think.
607
So if you figure out which ones are yours or you move towards them, you're worth more.
608
Business is more defensible.
609
And as I said, this talk was about moats, about defense.
610
And we're going to hear from Anar in a couple hours about how to go on offense.
611
So that's it for me for today.
612
We have time for questions.
613
of the Q&A from after the talk.
614
We're not posting that one on YouTube, so this is the only place to get it.
615
And if you wanna be in the room next time in 2027, MicroConf is headed to Austin, Texas in April and Dubrovnik, Croatia in October.
616
Watching these talks on YouTube is great, but the real magic happens in person.
617
You're in a room with about 250 founders who get exactly what you're building.
618
And some of the best conversations happen in the hallway between the talks.
619
The last seven, maybe eight MicroConf's has sold out.
620
So grab your tickets now at microconf.com slash events if you wanna go.
621
These events will sell out and I hope to see you there.
622
And if you already have your ticket or you're just ready for another great talk, check out this one right here.
623
I'll see you over there.
Tải Ứng Dụng
Mọi thứ bạn cần để nói trôi chảy
Chấm điểm AIChấm từng câu nói
Luyện IPAThành thạo mọi âm
Từ vựngMở rộng vốn từ
Game Từ vựngHọc qua chơi
Cùng kênh
FOR YOU
Video gợi ý
C141 câu
No, Really — Is A.I. a Bubble? | The Ezra Klein Show
Daily ConversationUniversity StudentsWorking Adults
C11
78 câu
The last real programmers are dying out...
Daily ConversationUniversity StudentsWorking Adults
C15
182 câu
How to Become an AI Product Manager in 2026 | Ex-Google, Microsoft
Daily ConversationUniversity StudentsWorking Adults
C19
20 câu
Alexandr Wang on what matters in an age of abundant intelligence
Daily ConversationUniversity StudentsWorking Adults
![The Great A.I. Debate: Should We Fear the Digital Revolution? [2015] | Intelligence Squared](https://img.youtube.com/vi/QWUeu3TLZN0/mqdefault.jpg)
C11
183 câu
The Great A.I. Debate: Should We Fear the Digital Revolution? [2015] | Intelligence Squared
Daily ConversationUniversity StudentsWorking Adults
B21
435 câu
Don't Lose the Room: How to Make Ideas Stick
Business EnglishDaily ConversationUniversity Students
B24
132 câu
Are we heading to an era of stagnation or innovation?
Daily ConversationUniversity StudentsWorking Adults
C11
315 câu
How Disruptive Will Generative AI Be? A Micro-Level Analysis of Evidence and Expectations
Daily ConversationUniversity StudentsWorking Adults
C1246 câu
Why Software Factories Fail
Daily ConversationUniversity StudentsWorking Adults
C13
553 câu
How AI Is Changing Strategy—and 4 Skills Leaders Need Next
Daily ConversationUniversity StudentsWorking Adults
C111
66 câu
The one OpenAI announcement that can actually make you money...
American AccentDaily ConversationUniversity StudentsWorking Adults
C11
127 câu
The Alignment Problem Explained: Crash Course Futures of AI #4
Business EnglishDaily ConversationUniversity Students
Từ vựng và ghi chú luyện nói cho bài học này
Bài luyện nói này dựa trên video “The Only SaaS Moats Left in 2026”. Video này có 622 câu và 7311 từ để luyện shadowing. Phần lời nói dài 39:50. Người nói nói nhanh, khoảng 184 từ mỗi phút, nên sẽ có nhiều chỗ nối âm và nuốt âm. 84% số từ nằm trong 3.000 từ tiếng Anh thông dụng nhất; phần còn lại nên xem trước khi luyện. 94 trong 622 câu là câu hỏi, nên bài này hợp để luyện ngữ điệu câu hỏi và cách trả lời ngắn.
Từ vựng quan trọng trong video
15 từ đáng học trong video, kèm phiên âm và nghĩa:
| Từ | Phiên âm | Nghĩa |
|---|---|---|
| vibe danh từ | /vaɪb/ | cảm giác |
| founder danh từ | /ˈfaʊ̯n.dəː/ | người sáng lập |
| slide danh từ | /ˈslaɪ̯d/ | cầu trượt, cầu tuột |
| competitor danh từ | /kəmˈpɛt.ɪ.tɚ/ | đối thủ |
| equity danh từ | /ˈɛk.wɪ.ti/ | sự bình đẳng |
| shock danh từ | /ʃɔk/ | choáng, ngạc nhiên |
| buyer danh từ | /ˈbaɪ.əɹ/ | người mua, khách mua |
| ecosystem danh từ | /ˈikoʊˌsɪstəm/ | hệ sinh thái |
| server danh từ | /ˈsɝvɚ/ | trình chủ, máy chủ |
| swim động từ | /swɪm/ | bơi, tắm |
| merchant danh từ | /ˈmɝt͡ʃənt/ | thương gia |
| complicated tính từ | /ˈkɑm.plɪˌkeɪ.tɪd/ | phức tạp |
| exit danh từ | /ˈɛɡzət/ | lối ra |
| loop danh từ | /luːp/ | đường nhánh, thòng lọng |
| mathematics danh từ | /mæθ(.ə)ˈmæt.ɪks/ | toán học |
Cụm động từ bạn sẽ nghe
| Từ | Nghĩa |
|---|---|
| switch off động từ | tắt |
| check out động từ | tính tiền, trả phòng |
| go back động từ | trở lại |
| go out động từ | ra, chết |
| sold out tính từ | đã bán hết, bán hết |
Những câu đáng luyện lại
Các câu ngắn, trọn ý trong video mà bạn có thể dùng lại khi giao tiếp hằng ngày:
- Act number eight, are your seats going away?
- Who has to move if they leave your software?
- You have a whole team?
- Who has to agree to replace you?
- If I'm a point solution, can I become a system of record?
Ngữ pháp trong video
Những cấu trúc người nói dùng nhiều nhất, kèm đúng cụm từ trong video:
| Cấu trúc | Trong video |
|---|---|
| “Used to” used to + động từ — thói quen hoặc tình trạng trong quá khứ nay không còn | used to buy · used to because · used to stay |
| Thì hiện tại hoàn thành tiếp diễn have/has been + -ing — việc bắt đầu từ trước và vẫn đang tiếp tục | I've been writing · I've been doing · have been paying |
| Thì hiện tại hoàn thành have/has + quá khứ phân từ — việc đã xảy ra nhưng còn liên quan đến hiện tại | I've ever given · I've ever written · you've read |
| Câu bị động be + quá khứ phân từ — nhấn vào việc xảy ra, không phải người làm | be replaced · been solved · are built |
Phát âm cần chú ý
Người nói dùng 217 dạng rút gọn, ví dụ you're, don't, I'm. Hãy nói theo dạng ngắn đúng như bạn nghe.
- Âm “th”: mathematics /mæθ(.ə)ˈmæt.ɪks/, sixth /sɪksθ/
- Âm “sh” và “zh”: shock /ʃɔk/, integration /ˌɪn.tɪˈɡɹeɪ.ʃən/, translation /tɹænzˈleɪʃən/, speculation /ˌspɛk.jəˈleɪ.ʃən/, cancellation /kænsəˈleɪʃən/
- Từ dài — đặt trọng âm cho đúng: competitor /kəmˈpɛt.ɪ.tɚ/, ecosystem /ˈikoʊˌsɪstəm/, integration /ˌɪn.tɪˈɡɹeɪ.ʃən/, complicated /ˈkɑm.plɪˌkeɪ.tɪd/, mathematics /mæθ(.ə)ˈmæt.ɪks/
Những âm người Việt thường gặp khó:
- Cụm phụ âm cuối — đọc đủ từng âm, đừng nuốt âm cuối: export /ˈɛks.pɔɹt/, merchant /ˈmɝt͡ʃənt/, mathematics /mæθ(.ə)ˈmæt.ɪks/, accountant /əˈkaʊn.tənt/, solve /sɒlv/
- Âm /tʃ/, /dʒ/ và /ʒ/ — tiếng Việt không có, đừng đọc thành “ch” hay “gi”: merchant /ˈmɝt͡ʃənt/, infrastructure /ˈɪnfɹəˌstɹʌkt͡ʃə/, strategic /stɹəˈti.dʒɪk/, conclusion /kənˈkluːʒən/, stretch /stɹɛt͡ʃ/
- Âm /æ/ — mở miệng rộng hơn âm “e”: mathematics /mæθ(.ə)ˈmæt.ɪks/, spam /spæm/, translation /tɹænzˈleɪʃən/, gravity /ˈɡɹævɪti/, barrier /ˈbæɹi.ə/
Cách luyện với video này
- Nghe hết video một lần, chưa cần nói, và ghi lại những từ bạn chưa biết.
- Bắt đầu ở tốc độ 0,75×, nói đuổi từng câu, rồi quay lại tốc độ bình thường khi đã quen.
- Ghi âm giọng mình rồi so với bản gốc, chú ý các từ như vibe, founder, slide.
Phương Pháp Shadowing Là Gì?
Shadowing là kỹ thuật học ngôn ngữ có cơ sở khoa học, ban đầu được phát triển cho chương trình đào tạo phiên dịch viên chuyên nghiệp và được phổ biến rộng rãi bởi nhà đa ngôn ngữ học Dr. Alexander Arguelles. Nguyên lý cốt lõi đơn giản nhưng cực kỳ hiệu quả: bạn nghe tiếng Anh của người bản xứ và lặp lại to ngay lập tức — như một "cái bóng" (shadow) đuổi theo người nói với độ trễ chỉ 1–2 giây. Khác với luyện ngữ pháp hay học từ vựng bị động, Shadowing buộc não bộ và cơ miệng phải đồng thời xử lý và tái tạo ngôn ngữ thực tế. Các nghiên cứu khoa học xác nhận phương pháp này cải thiện đáng kể phát âm, ngữ điệu, nhịp điệu, nối âm, kỹ năng nghe và độ lưu loát khi nói — đặc biệt hiệu quả cho người luyện IELTS Speaking và muốn giao tiếp tiếng Anh tự nhiên như người bản ngữ.