Machine-generated source transcript; recognition errors remain. Timestamps and subtitles use the selected source recording’s clock, not clip-relative time. Section links use cited source moments; sections without a verified moment return to the talk start.
7:55:24Hey y'all, I'm Quinn. Great to see you. You might want to come up in front a little more because I
7:55:28do, it's like an NBA game, have some swag and hats to give out. So if you're in the back and if you
7:55:33like hats, come on up. All right, as whatever, you know, that said, my name is Quinn Slack. I am
7:55:39one of the founders of AMP and had been at Sourcecraft before. I don't want to talk about AMP,
7:55:43though. I want to talk about really a personal story of how agents blew my mind and changed my
7:55:51life. There was a day when I walked outside and the birds were chirping more and the sky was bluer
7:55:56and how they changed my company and how I think they're going to change every company. And it's
7:56:01kind of frustrating sometimes to see friends and people I care about, companies I care
7:56:05about, really struggle with this. So we're going to go on a little bit of a roller coaster,
7:56:08but why don't we get started? I know it's almost dinner time. Anyone want some hats?
7:56:14Some AMP hats? All right. Here we go. I see people getting closer. There we go. All right.
7:56:22I hope I don't hit anyone.
7:56:25There we go.
7:56:28And over there.
7:56:30All right.
7:56:33Yeah, everybody gets a hat.
7:56:34Everybody gets a hat.
7:56:35Come up and find me after.
7:56:37All right.
7:56:38There was a moment in January of 2025 when I was just coding,
7:56:44and I wanted to see what if you just make an agent.
7:56:47This was before Claude Code, and it was like the dumbest stuff ever.
7:56:51It was like change its background color, but it just blew my mind when I saw an agent actually work for the first time
7:56:58it
7:56:58Ran commands it looked at itself. It gave itself some feedback loops that blew my mind and I
7:57:05Had this sense I think that everyone has when they first see an agent that
7:57:09Everything is going to change after that everything is going to change about how you build software
7:57:13I think we've all felt that everything is going to change about how you
7:57:16form companies to work on software and how you sell software, how you distribute software.
7:57:21All other work is going to change. And it was just an absolutely crazy, crazy moment.
7:57:27I was actually at the Frankfurt airport with Thorsten Ball on our team. He wrote that compiler
7:57:32book. Anyone who knows Thorsten on our team? Yeah, he's a cool guy. We were just like minds
7:57:40blowing at the Frankfurt airport. And we knew that everything was going to change. We didn't
7:57:43know where this was all going to go, but we just wanted to go. We just wanted to go run
7:57:47at it. So inside of Sourcegraph, which was a 200-person enterprise software company,
7:57:53I was the CEO, and we made code search, we decided to just start building this, to tune
7:57:58out everything else and just start building this thing, not go to sales calls, not go
7:58:04to like exec meeting, none of that stuff, just tune out all the distractions. And I
7:58:10I was in this position that I think a lot of people say,
7:58:13wow, I'm a software developer at a company,
7:58:15and I want to change.
7:58:16I want my company to be more AI forward, and I can't.
7:58:18I was a CEO, and in theory, I could change anything.
7:58:21I could change teams.
7:58:22I could say, you don't need to get pull request code review.
7:58:26You don't need to do any of the other processes we used to do.
7:58:29And yet, it was very hard.
7:58:32What we did is we ended up taking 20 people from the company,
7:58:35and inside of Sourcegraph still, we
7:58:38made a little team, the AMP team.
7:58:40And that's what we tried.
7:58:42So for those of you who are trying to push back against your company
7:58:44to try to make it more
7:58:46agent-friendly, I have a lot of sympathy
7:58:48for you, and I hope that you can take away some things from here.
7:58:51I know how hard
7:58:52it is. And also, have some
7:58:54empathy for your CEO, too. It's hard for them, too.
7:58:56This is a really hard kind of
7:58:58change and transformation. But we
7:59:00knew that there were some things that were going to change.
7:59:02And this all seems kind of obvious now, but back in
7:59:042025, early 2025,
7:59:06with this idea of enterprise ready,
7:59:08you do this thing and then you can build a sales team
7:59:10and go sell it.
7:59:11We knew that that was just going to totally die off.
7:59:14You had to be on the frontier.
7:59:16If you had a taste of success
7:59:19and then you started doing all the enterprise checklist bullshit
7:59:21to make the median enterprise buyer happy,
7:59:24your product was going to go obsolete in like three months.
7:59:28And the whole game had to change.
7:59:30Nothing mattered other than making it
7:59:32so that the people at these other companies
7:59:34loved your product and the CEO, the budget was infinite, and all bets were off.
7:59:39So we knew that.
7:59:40We didn't quite know all the challenges we'd run into.
7:59:45Also, in this world where agents can build a product or build a feature,
7:59:49how do you even think about differentiation?
7:59:51I mean, to this day, there's people who say, well, why is AMP better?
7:59:54Why is Claude Code better than Codex or whatever?
7:59:57And so many people are begging
8:00:00to have the answer in the form of, well, it's these three features.
8:00:02You know, AMP has orbs, which are really good cloud agents, and we have multiplayer.
8:00:07But the crazy thing is, you all know that if we can do that and if we prove it,
8:00:10then any of our competitors could as well.
8:00:13So, you know, why even play that game?
8:00:16Like, what does it even mean to build and sell software?
8:00:18Why would you go and hire a sales team?
8:00:20Why would you go and do anything in the traditional way if you don't even know what it's going to look like in a month?
8:00:25And if you don't even know how to describe differentiation,
8:00:26there's no sustainable differentiation.
8:00:29For AMP, people end up loving it.
8:00:31But you can even see Adam Wathen, who's the Tailwind guy.
8:00:37He loves AMP.
8:00:38But he's like, it's one of those things
8:00:39where you can't understand why it's good until you use it.
8:00:44Can you imagine having a whole product management team,
8:00:47a design team, an engineering team that's
8:00:49so disconnected from the actual customers, a whole company
8:00:52apparatus?
8:00:52if that's the best way that you can market a product.
8:00:55It's really tough.
8:00:57So we knew that we just had to change everything.
8:00:59And we rethought so many things
8:01:03because kind of unintentionally,
8:01:04we were under some weird constraints.
8:01:07We decided to spin out AMP.
8:01:09So for those of you who have been at a company,
8:01:11like a small, you know,
8:01:13you're on a small team in a company,
8:01:14you're trying to do things different,
8:01:15you're trying to be the startup within a startup,
8:01:17you probably have these fantasies of,
8:01:18let's just go off and start a new company.
8:01:20Well, we actually did that.
8:01:21We did the right thing by investors and employees, and everyone has ownership in both.
8:01:25But we actually spun out, and for a year, until last Friday, we couldn't hire anyone.
8:01:31We couldn't grant options.
8:01:32The legal spinoff process was a huge pain.
8:01:35And we were forced to make do with the 21 people that we brought over.
8:01:39We didn't have time in this crazy competitive environment to do...
8:01:43We couldn't move as slowly in order to do mandatory pull requests.
8:01:48We couldn't have meetings.
8:01:50We couldn't have our time be sucked up by enterprises
8:01:53that would need to take many months to decide on something.
8:01:55And so we were forced into this radically different way of working.
8:01:59And we also had this environment where every single person on the team
8:02:02has a huge stake in the company.
8:02:04And that skin in the game that you get,
8:02:06when AI makes it so you can operate with way less people,
8:02:09skin in the game, it's like this spiritual thing for the soul of great people.
8:02:14It makes them better at everything they do.
8:02:16It makes them grow faster.
8:02:17And so this is what we have seen.
8:02:19And this, I think, when you've seen a company that operates like this without any of that bullshit,
8:02:24it's obvious how companies like this are going to crush everyone else
8:02:27and how every company in the future is going to look a lot more like that.
8:02:31So, yeah, we spun AMP out, and we had to figure out a lot of new ways of working with these constraints
8:02:38that we had no idea that we were going to be under.
8:02:40And we couldn't hire, as I said.
8:02:43So as a result, while our, this is a log chart, by the way,
8:02:46while all the other big competitors in the space
8:02:48were growing and building old-fashioned companies
8:02:50in order to go and exploit this amazing technology of AI,
8:02:54we had to stay at 21 people.
8:02:56But that was awesome because it was a forcing function
8:02:57for us to work differently and to figure out
8:02:59how can you do so much more with an agent
8:03:01where we couldn't hire people.
8:03:04There was actually one guy who was like,
8:03:05I don't care that I can't get options for like eight months
8:03:08and I'm going to come join, and Hamish is pretty awesome.
8:03:12So I want to talk about the ways in which we built AMP,
8:03:14which is our agent product,
8:03:17around the way that these constraints forced us to work
8:03:20and how we grew up around the agent here.
8:03:25We got rid of all of the other tools.
8:03:27We got rid of all of the other things you got to check.
8:03:29Obviously not JIRA.
8:03:30No task tracking.
8:03:32We don't have a support system.
8:03:33Everything we do in the agent.
8:03:35In AMP, they're called threads.
8:03:37We even communicate in the agent.
8:03:39You can say, you know, I made a,
8:03:41something's bad in the product.
8:03:42Let's do it.
8:03:43The agent can listen to everything,
8:03:44and then go fix it.
8:03:46If you're in like a local agent that runs on your laptop,
8:03:50you can't bring someone else in
8:03:51if you realize you're in over your head.
8:03:53Really valuable.
8:03:54We push way too often.
8:03:56And I want to show, you know,
8:03:58how we actually reduce the risk here.
8:04:00Because you might say, let me start that again.
8:04:02You might say, well, if you're pushing constantly,
8:04:04then how do you not take down the site all the time?
8:04:07And the difference is when you can push
8:04:10and you can do it any time of day just by yourself,
8:04:13you can chunk up your work into way smaller pieces.
8:04:16Before you're going and doing a big change,
8:04:18you can go and say, let me add the logging
8:04:20to make sure that my understanding,
8:04:21the agent's understanding of the system is correct.
8:04:23Let me push a little change, a little change,
8:04:25and then the agent can monitor it for an hour,
8:04:27then push the next change.
8:04:28You can run a whole shadow system side by side
8:04:31with the old system to see,
8:04:32does it compute the exact same results in every single case?
8:04:35And that way of working actually reduces the risk
8:04:39because by the time you finish the old way of doing it,
8:04:42And the pace that things have to move these days
8:04:44and what your customers expect,
8:04:46you've probably forgotten what you're building,
8:04:47and it just totally sucks.
8:04:50Let me give us some more hats.
8:04:51Who wants some more hats?
8:04:53All right.
8:04:54There we go.
8:04:55Sorry.
8:04:56The one guy who's on his laptop.
8:04:57That's a lesson, right?
8:05:00There we go.
8:05:01All right.
8:05:04Another thing is making it so the agent is able
8:05:08to irrefutably prove what the thing that's building is correct.
8:05:13There's a lot of ways I can do this. Turns out they're really good at proving these things.
8:05:17So if you're trying to remove like jitter in the UI, make sure that when you load
8:05:20it's like nice and crisp. You can ask the agent to do something that no human would ever have
8:05:24the patience to do, which is take every single frame and compare them visually. And that's what I did
8:05:29here. You can ask the agent to go make a
8:05:32ton of different versions of something. You can never ask a human to like, you know,
8:05:36if you ask a designer, go make nine iterations or nine styles of this. They're going to get really annoyed.
8:05:41I know because I've worked with designers,
8:05:42and I know they don't like that.
8:05:43But for the agent, yeah, you can have it
8:05:46make a ton of different things.
8:05:47This is the waveform for dictation in AMP.
8:05:50And this is one of the solutions of people who say,
8:05:52I don't like waiting on my agent.
8:05:54They're asking for one thing at a time.
8:05:56And then they say, make this little change.
8:05:58You shouldn't do that.
8:05:58Go have it make a ton of things at once,
8:06:00and you're going to get a lot more done.
8:06:04We have this magical system where if you report a bug in AMP,
8:06:07has anyone here reported a bug in AMP?
8:06:09All right.
8:06:10Cool.
8:06:11I apologize for the bug,
8:06:13but we have a system that is a full loop
8:06:17where when you report a bug,
8:06:19we will fix it, of course.
8:06:22You know, Android app.
8:06:24We'll fix it, obviously, with the agent.
8:06:26I think everyone does that now.
8:06:28Mostly everyone.
8:06:29A lot of times still,
8:06:29the engineers are different from the support team,
8:06:31which is kind of crazy.
8:06:33We'll send an email to the person thanking them.
8:06:35We give them free credits,
8:06:36which is always kind of nice.
8:06:38I think it sucks.
8:06:39I think it sucks when there's a bug in a product
8:06:41and the company doesn't apologize.
8:06:43I just think that's morally wrong.
8:06:45And then what we'll do is we'll go through the logs,
8:06:47and this is something that no human would ever do by themselves.
8:06:50We'll see everyone else who was affected by that issue.
8:06:53Because where there's smoke, there's fire.
8:06:55So you find the 17 other people that hit that same obscure error.
8:06:59They remember it, probably, and you proactively tell them.
8:07:01How many times have you gotten an email from some product that says,
8:07:04hey, I know that you saw an error yesterday and that was really annoying.
8:07:07I'm really sorry about that. Here's $10. We fixed it.
8:07:09And by the way, that, it's a self-perpetuating cycle
8:07:12where then people who get that know that you care
8:07:14and they're going to send more feedback in the future.
8:07:17And then you ship those all out and then, you know, boom.
8:07:19And that's the cycle.
8:07:21That, people talk about vibe-coded products being junk.
8:07:25AMP is basically 100% built with AI
8:07:29and people say it's very high quality.
8:07:31And this is one of the things that you can do differently with an agent.
8:07:34But there's no way in hell you could do this
8:07:35if you had to go through a PM or go to the team that owns this bug
8:07:38and the designers each time.
8:07:41Dictation.
8:07:42Who dictates to their agents?
8:07:45Yeah?
8:07:45Wow, that's pretty good.
8:07:47That's a lot more than I thought.
8:07:49It is the fastest way to get bits of information
8:07:52from your brain to the agent.
8:07:54And I think that there's probably something around it
8:07:59ushering in the return of remote work.
8:08:02Because I was just in an office today
8:08:04with Biong, another AMP co-founder, dictating,
8:08:06and it was very annoying.
8:08:09Dictation, if you're not doing it yet, you've got to do it.
8:08:12Just overcome the embarrassment.
8:08:16Channel Nathan Fielder and tell yourself that you can withstand any amount of awkwardness as you talk.
8:08:22We've hooked up our agent to a lot of different tools.
8:08:25I think a lot of companies have gotten here to this point.
8:08:29We've been doing this in production for a long time.
8:08:32And what we've found is the more that we start using these tools,
8:08:36where just the agent is using it directly,
8:08:38we don't even need the software product on the other end.
8:08:40And it's a very small number of products
8:08:42that we think actually have sticking power for us.
8:08:45So, you know, we built some new things in AMP,
8:08:50for example, a CRM.
8:08:51And a year ago, you heard people talking about,
8:08:54oh, no one's going to Vibecode Salesforce from scratch.
8:08:56And yeah, that's totally true.
8:08:57Why would you want to?
8:08:59Have you ever used it?
8:09:00Would you be proud if you Vibecoded that
8:09:02and that came out the other end?
8:09:03And also, you've used like 2% of the features of Salesforce.
8:09:07So actually, you know, people ridiculed this idea,
8:09:11but actually, would you code the 2% subset of Salesforce features
8:09:17and then another 2% that integrate with the rest of your stack
8:09:21in the most perfect way because the agent is so good at that?
8:09:25And that, you know, 4% of Salesforce total,
8:09:27that's like the easy shit,
8:09:29and you know what you want out of that better than Salesforce,
8:09:31and you certainly don't want to spend your time
8:09:33talking to some Salesforce product manager
8:09:34and hearing from them,
8:09:36I'm going to advocate with my PM and my executive team
8:09:38to get that on our Q4 roadmap, all that bullshit.
8:09:42So compared to that,
8:09:43it's actually getting easier and easier
8:09:46to just build the tools that you want from scratch.
8:09:50And this goes back to some of the existential
8:09:53kind of dread but also mind-blowing feeling
8:09:56that we felt when we first tried an agent.
8:09:57This is what
8:10:00we kind of thought would happen, that it would change everything about the software industry.
8:10:05If Salesforce, which is like the king cobra of SaaS, they have the most reliable customers,
8:10:12most reliable renewals, all that.
8:10:13They like invented the category.
8:10:15If we're starting to see, you know, why would we use them?
8:10:19Well, everything is starting to fall.
8:10:20And all of SaaS starts to break down if you can no longer assume that you can pay a salesperson 100K for closing a deal, and that deal's going to last for 10 years, and it's going to get bigger by a little bit.
8:10:32All of that is called into question now.
8:10:34So if you're at a software company, and that is how you ultimately make money, and you're trying to ring the bell, you're trying to raise the alarm here, absolutely raise it.
8:10:46and get out of the cage that you are in
8:10:49because, in my opinion,
8:10:54once you have seen how the new kind of companies
8:10:57that are two orders of magnitude smaller,
8:10:59where everyone can do everything
8:11:00and where everyone has skin in the game,
8:11:02then you don't want anything else.
8:11:04And so this is what we call this new kind of company
8:11:06where we take that old org chart
8:11:08and now we just got 21 people on AMP.
8:11:12We have no PMs.
8:11:14Everyone talks to customers.
8:11:15Brady is this guy who is an SDR at Sourcegraph,
8:11:18really smart guy.
8:11:19I love the guy.
8:11:20And one day he's like,
8:11:21I'm going to make a wire,
8:11:23ACH wire transfer thing
8:11:24so customers can pay us, you know,
8:11:25if it's too big for a credit card.
8:11:27And he's like two years out of college
8:11:29and he's just doing that.
8:11:30And nobody knows on the other end.
8:11:31It's like that old New Yorker cartoon.
8:11:33On the internet, nobody knows you're a dog.
8:11:34You know, nobody knows that he's an SDR,
8:11:36that we're all kind of, you know,
8:11:37devs and doing everything.
8:11:38But you can.
8:11:40And you also get people
8:11:41that are really real with customers.
8:11:43And I think sometimes customers need real talk.
8:11:45So Alex Kemper on our team,
8:11:47just today, he was like,
8:11:49hey, to a customer who wanted our stuff,
8:11:52that sounds like we're just going to get bogged down
8:11:53in a bunch of BS,
8:11:54and I don't think that's a good use of anyone's time.
8:11:56If you can get to a place where you have people
8:11:59that feel empowered to say that
8:12:00and to truly say the best thing for the customer,
8:12:02that's great.
8:12:03You're not going to get that out of a salesperson
8:12:04who is commissioned.
8:12:06We have, this is some people's favorite part,
8:12:09we have one meeting per week,
8:12:10and then everything else,
8:12:11the guidance is follow your energy.
8:12:14You couldn't run most companies like that,
8:12:16but if everyone at the company has a massive stake in the company,
8:12:19take 100% and divide it by 21,
8:12:22and all that, you can actually trust people.
8:12:25Because again, when people have skin in the game,
8:12:27it makes them better at everything they do.
8:12:29It's like an incentive to make them think whole company,
8:12:32to make them do more, to make them work harder.
8:12:34And that's why I included Saturday and Sunday as well on this.
8:12:39And we are shipping just so much more.
8:12:43I think that AMP, for many people, is a product that they love.
8:12:47We also have a great business,
8:12:48and we're serving as proof that this new kind of company
8:12:51that is profitable, that's more than 100 million ARR,
8:12:55that is doing so on a dollar-in, dollar-out basis
8:12:58that's 40 to 50 times more efficient
8:13:00than old-fashioned companies here, that that can work.
8:13:04And so if this is appealing to you,
8:13:07if you want to push your company in this direction,
8:13:09if you want to go start a company like this,
8:13:11then reach out.
8:13:12I'm happy to help.
8:13:14And my guidance is just get out of the cage.
8:13:17I've heard from so many people who say,
8:13:19I hate that there's a gossip channel
8:13:20with the people at my company.
8:13:22We're all talking about the change.
8:13:23You know what?
8:13:24The longer that you are in the gossip channel,
8:13:26the longer that you are at a company
8:13:27getting paid a lot of money
8:13:28and you're really smart
8:13:28and you're not doing shit about it,
8:13:30you're starting to become part of the problem.
8:13:32And the longer that you let yourself
8:13:34be just a developer
8:13:36when agents have completely changed the game,
8:13:39you're really hurting yourself.
8:13:40So if this is the one thing that you take away,
8:13:45just get out of that cage,
8:13:46let agents really shape how you work,
8:13:49and get to a company that's willing to be shaped by the agents.
8:13:53Thank you.
8:13:54And if you want to try AMP, you can get up there.
8:13:58You can copy that.
8:13:59I've got more hats, so find me afterwards.
8:14:01And I'll probably be getting dinner at Chipotle or something,
8:14:03so come find me if you want to chat about all this stuff.
8:14:07All right, thank you all.
8:14:08Thank
At 08:04:05, Slack reports roughly 130 deployments a day and about ten minutes from push to production. These are Amp's presented operating figures, not a recommendation to remove review from every system.