2026 State of AI Engineering — Barr Yaron, Amplify Partners

Now joining us on stage is the partner Now joining us on stage is the partner at Amplify Barren. at Amplify Barren. at Amplify Barren. [music] Fantastic. Fantastic. You did a great job practicing. I feel You did a great job practicing. I feel You did a great job practicing. I feel very very loved. Um, let's get started. very very loved. Um, let's get started. very very loved. Um, let's get started. So, like you just heard, my name is Bar. So, like you just heard, my name is Bar. So, like you just heard, my name is Bar. I run a survey every year on the state I run a survey every year on the state I run a survey every year on the state of AI engineering. And the funny thing of AI engineering. And the funny thing of AI engineering. And the funny thing about running a survey on the state of about running a survey on the state of about running a survey on the state of AI engineering is that the field changes AI engineering is that the field changes AI engineering is that the field changes as you make the slides. as you make the slides. as you make the slides. Just in the past week, we've had Just in the past week, we've had Just in the past week, we've had Frontier releases treated like national Frontier releases treated like national Frontier releases treated like national security events. Meta reportedly security events. Meta reportedly security events. Meta reportedly exploring selling AI compute. By the exploring selling AI compute. By the exploring selling AI compute. By the time I get off stage, maybe something time I get off stage, maybe something time I get off stage, maybe something else will happen. So, if I miss a major else will happen. So, if I miss a major else will happen. So, if I miss a major announcement while I'm up here, please announcement while I'm up here, please announcement while I'm up here, please come find me after.

come find me after. come find me after. But that's exactly why we run the survey But that's exactly why we run the survey But that's exactly why we run the survey every year to cut through the noise, every year to cut through the noise, every year to cut through the noise, take a moment, step back and understand take a moment, step back and understand take a moment, step back and understand what AI engineers are actually doing. Uh what AI engineers are actually doing. Uh what AI engineers are actually doing. Uh for the first time this year, we were for the first time this year, we were for the first time this year, we were thrilled to partner with Notion and thrilled to partner with Notion and thrilled to partner with Notion and Verscell to run this survey. Verscell to run this survey. Verscell to run this survey. Very quickly on me, uh this is the least Very quickly on me, uh this is the least Very quickly on me, uh this is the least interesting slide. I'm an investment interesting slide. I'm an investment interesting slide. I'm an investment partner at Amplify. Very lucky to invest partner at Amplify. Very lucky to invest partner at Amplify. Very lucky to invest in companies built by and for AI in companies built by and for AI in companies built by and for AI engineers. And I'll make the same engineers. And I'll make the same engineers. And I'll make the same promise that I make every single year, promise that I make every single year, promise that I make every single year, which is short time on bar, long time on which is short time on bar, long time on which is short time on bar, long time on bar charts. So, let's get right into it bar charts. So, let's get right into it bar charts. So, let's get right into it with lots of bar charts. First, let's talk about well, maybe First, let's talk about well, maybe raise your hand. Did you fill out the raise your hand. Did you fill out the raise your hand. Did you fill out the survey? This is a very large group. survey? This is a very large group. survey? This is a very large group. Okay. Yes, I see you in the front. Um, Okay. Yes, I see you in the front. Um, Okay. Yes, I see you in the front. Um, if the answer is you, thank you so much. if the answer is you, thank you so much. if the answer is you, thank you so much. If the answer is not you, I will find If the answer is not you, I will find If the answer is not you, I will find you in 2027. But genuinely, this only you in 2027. But genuinely, this only you in 2027. But genuinely, this only exists because a thousand of you gave exists because a thousand of you gave exists because a thousand of you gave your time. So, thank you. We had 1,048 your time. So, thank you. We had 1,048 your time. So, thank you. We had 1,048 respondents this year, which is a lot of respondents this year, which is a lot of respondents this year, which is a lot of AI engineers.

AI engineers. AI engineers. And to be precise, this is not just AI And to be precise, this is not just AI And to be precise, this is not just AI engineers, as I'm sure you see at the engineers, as I'm sure you see at the engineers, as I'm sure you see at the conference. Every year, we see that AI conference. Every year, we see that AI conference. Every year, we see that AI engineering is more of a discipline than engineering is more of a discipline than engineering is more of a discipline than a job title. It touches founders, CTO's, a job title. It touches founders, CTO's, a job title. It touches founders, CTO's, engineers, product people, folks across engineers, product people, folks across engineers, product people, folks across company sizes and experience levels. company sizes and experience levels. company sizes and experience levels. And that range shows up in experience And that range shows up in experience And that range shows up in experience too. Um for the third year running we too. Um for the third year running we too. Um for the third year running we see the same pattern which is skew see the same pattern which is skew see the same pattern which is skew towards senior engineers but newer to towards senior engineers but newer to towards senior engineers but newer to AI. Of those with over 10 years of AI. Of those with over 10 years of AI. Of those with over 10 years of software experience over half have three software experience over half have three software experience over half have three years or less of AI experience which years or less of AI experience which years or less of AI experience which tracks uh these are very experienced tracks uh these are very experienced tracks uh these are very experienced engineers learning a new paradigm in engineers learning a new paradigm in engineers learning a new paradigm in real time. And the newest cohort, the real time. And the newest cohort, the real time. And the newest cohort, the ones who just started uh engineering, ones who just started uh engineering, ones who just started uh engineering, the median new engineer has nearly as the median new engineer has nearly as the median new engineer has nearly as much AI experience as the median 10-year much AI experience as the median 10-year much AI experience as the median 10-year software veteran. Uh so the newest software veteran. Uh so the newest software veteran. Uh so the newest engineers have never known software engineers have never known software engineers have never known software without this. without this. without this. But doing AI doesn't mean one thing. We But doing AI doesn't mean one thing. We But doing AI doesn't mean one thing. We talked about all these different titles, talked about all these different titles, talked about all these different titles, all these different roles. Before we get all these different roles. Before we get all these different roles. Before we get into models and agents, I have a more into models and agents, I have a more into models and agents, I have a more basic question, which is when people say basic question, which is when people say basic question, which is when people say they're doing AI at work, what are they they're doing AI at work, what are they they're doing AI at work, what are they actually doing?

actually doing? actually doing? So, first up, like to start with the So, first up, like to start with the So, first up, like to start with the modalities. We asked, which modalities modalities. We asked, which modalities modalities. We asked, which modalities are you actively building with at work? are you actively building with at work? are you actively building with at work? Can anyone take a guess? Text dominates. Can anyone take a guess? Text dominates. Can anyone take a guess? Text dominates. I know. Hold your applause. Um, but one I know. Hold your applause. Um, but one I know. Hold your applause. Um, but one piece of this chart that I always find piece of this chart that I always find piece of this chart that I always find very interesting and I always look at is very interesting and I always look at is very interesting and I always look at is the ratio of nope, I'm not using this the ratio of nope, I'm not using this the ratio of nope, I'm not using this modality to I'm not using it, but I do modality to I'm not using it, but I do modality to I'm not using it, but I do plan to. plan to. plan to. I call this the intent to adopt ratio. I call this the intent to adopt ratio. I call this the intent to adopt ratio. Of the people who are not building with Of the people who are not building with Of the people who are not building with a modality today, how many say they plan a modality today, how many say they plan a modality today, how many say they plan to use it? And audio has the strongest to use it? And audio has the strongest to use it? And audio has the strongest intent to adopt this year. Among AI intent to adopt this year. Among AI intent to adopt this year. Among AI engineers who are not building with engineers who are not building with engineers who are not building with audio today, a whopping 56% say they audio today, a whopping 56% say they audio today, a whopping 56% say they plan to adopt it in the AI applications plan to adopt it in the AI applications plan to adopt it in the AI applications they build. And this is not a brand new they build. And this is not a brand new they build. And this is not a brand new signal. Last year audio also had the signal. Last year audio also had the signal. Last year audio also had the highest intent to adopt across highest intent to adopt across highest intent to adopt across modalities but 37%. So audio continues modalities but 37%. So audio continues modalities but 37%. So audio continues to take the lead and have high interest to take the lead and have high interest to take the lead and have high interest but that interest is accelerating.

but that interest is accelerating. but that interest is accelerating. Now there has been an audio swing but if Now there has been an audio swing but if Now there has been an audio swing but if we look at what changed most from the we look at what changed most from the we look at what changed most from the last year in the survey the biggest jump last year in the survey the biggest jump last year in the survey the biggest jump is actually in people using image is actually in people using image is actually in people using image generation. generation. generation. The share of respondents using The share of respondents using The share of respondents using generative AI for images and feeling generative AI for images and feeling generative AI for images and feeling really good about it doubled from 18% really good about it doubled from 18% really good about it doubled from 18% last year to 36% this year. Makes sense last year to 36% this year. Makes sense last year to 36% this year. Makes sense if you look at what we launched in the if you look at what we launched in the if you look at what we launched in the same window. Over the past year same window. Over the past year same window. Over the past year plus survey time, uh we've had models plus survey time, uh we've had models plus survey time, uh we've had models Nano Banana, Nano Banana 2, Chat GPT Nano Banana, Nano Banana 2, Chat GPT Nano Banana, Nano Banana 2, Chat GPT images 2.0. The products have gotten images 2.0. The products have gotten images 2.0. The products have gotten much better. What used to feel like an much better. What used to feel like an much better. What used to feel like an efficient way to generate cursed hands efficient way to generate cursed hands efficient way to generate cursed hands is just increasingly becoming a part of is just increasingly becoming a part of is just increasingly becoming a part of real work. Audio may have the strongest real work. Audio may have the strongest real work. Audio may have the strongest intent to adopt, but image generation intent to adopt, but image generation intent to adopt, but image generation shows us what happens when a modality shows us what happens when a modality shows us what happens when a modality crosses that threshold. So, I'm excited crosses that threshold. So, I'm excited crosses that threshold. So, I'm excited to continue watching these adoption to continue watching these adoption to continue watching these adoption curves every single year. I think we're curves every single year. I think we're curves every single year. I think we're going to see a lot this year. going to see a lot this year. going to see a lot this year. Uh, now models. If you've Who here Uh, now models. If you've Who here Uh, now models. If you've Who here spends time on Twitter?

spends time on Twitter? spends time on Twitter? All right. Yes. I imagine this is a very All right. Yes. I imagine this is a very All right. Yes. I imagine this is a very Twitter pilled uh crowd. If you spend Twitter pilled uh crowd. If you spend Twitter pilled uh crowd. If you spend any time on Twitter in this uh in this any time on Twitter in this uh in this any time on Twitter in this uh in this circle, you've seen a lot written about circle, you've seen a lot written about circle, you've seen a lot written about openweight models these past few months openweight models these past few months openweight models these past few months and I think we'll see it even more in and I think we'll see it even more in and I think we'll see it even more in the next year. Um so we asked what the next year. Um so we asked what the next year. Um so we asked what models are you actually using in models are you actually using in models are you actually using in production. 94% use closed models. 45% production. 94% use closed models. 45% production. 94% use closed models. 45% are using openweight models. But here's are using openweight models. But here's are using openweight models. But here's the thing, you know, openweight models the thing, you know, openweight models the thing, you know, openweight models are not replacing closed models for the are not replacing closed models for the are not replacing closed models for the most part. at least not yet. The most part. at least not yet. The most part. at least not yet. The respondents using openweight models, respondents using openweight models, respondents using openweight models, over 90% of them are also using closed over 90% of them are also using closed over 90% of them are also using closed models. So they're looking like an models. So they're looking like an models. So they're looking like an augmentation. Teams are mixing and augmentation. Teams are mixing and augmentation. Teams are mixing and matching. We also asked just to double click on We also asked just to double click on this for the top three considerations this for the top three considerations this for the top three considerations when choosing a model. If you're when choosing a model. If you're when choosing a model. If you're choosing a model, what is important to choosing a model, what is important to choosing a model, what is important to you? Um and despite the airtime of the you? Um and despite the airtime of the you? Um and despite the airtime of the open versus closed, it's not what drives open versus closed, it's not what drives open versus closed, it's not what drives model choice. It was a top three model choice. It was a top three model choice. It was a top three consideration for only 5% of the consideration for only 5% of the consideration for only 5% of the respondents.

respondents. respondents. What matters is actually more What matters is actually more What matters is actually more straightforward. It's quality. Quality straightforward. It's quality. Quality straightforward. It's quality. Quality dominates. Followed by agentic dominates. Followed by agentic dominates. Followed by agentic capabilities like tool calling and cost capabilities like tool calling and cost capabilities like tool calling and cost tied right with it. We'll money money. tied right with it. We'll money money. tied right with it. We'll money money. We'll get back to that. Um, one thing We'll get back to that. Um, one thing We'll get back to that. Um, one thing that I found very interesting is that that I found very interesting is that that I found very interesting is that reliability is not near the top. Only reliability is not near the top. Only reliability is not near the top. Only one in five named reliability. That one in five named reliability. That one in five named reliability. That doesn't mean teams stopped caring about doesn't mean teams stopped caring about doesn't mean teams stopped caring about reliability. Uh there are different ways reliability. Uh there are different ways reliability. Uh there are different ways to interpret this data. My guess is that to interpret this data. My guess is that to interpret this data. My guess is that it's more likely to become a threshold it's more likely to become a threshold it's more likely to become a threshold requirement and the models they're requirement and the models they're requirement and the models they're choosing are reliable enough so the choosing are reliable enough so the choosing are reliable enough so the decision moves up the stack outside of decision moves up the stack outside of decision moves up the stack outside of certain circumstances to quality, certain circumstances to quality, certain circumstances to quality, capability, cost, capability, cost, capability, cost, but we could talk after. All right, so but we could talk after. All right, so but we could talk after. All right, so here's where the model story all comes here's where the model story all comes here's where the model story all comes together. together. together. Like I said, teams are not choosing one Like I said, teams are not choosing one Like I said, teams are not choosing one model and calling it a day. Earlier I model and calling it a day. Earlier I model and calling it a day. Earlier I showed that 87% of teams are using more showed that 87% of teams are using more showed that 87% of teams are using more than one model. Uh the model that's the than one model. Uh the model that's the than one model. Uh the model that's the opposite of standardization. opposite of standardization. opposite of standardization. Uh and the way that they choose models Uh and the way that they choose models Uh and the way that they choose models for given tasks varies. Most popular is for given tasks varies. Most popular is for given tasks varies. Most popular is routing by task type. Some run multiple routing by task type. Some run multiple routing by task type. Some run multiple models compare outputs. Some route based models compare outputs. Some route based models compare outputs. Some route based on cost. Uh but models are good at on cost. Uh but models are good at on cost. Uh but models are good at different things. What was interesting different things. What was interesting different things. What was interesting was that more than half of respondents was that more than half of respondents was that more than half of respondents said that their organizations starting said that their organizations starting said that their organizations starting to standardize on fewer AI tools.

to standardize on fewer AI tools. to standardize on fewer AI tools. They're trading standardiz flexibility They're trading standardiz flexibility They're trading standardiz flexibility for standardization. for standardization. for standardization. A share of those are mixed. They say A share of those are mixed. They say A share of those are mixed. They say they're standardizing on some layers they're standardizing on some layers they're standardizing on some layers while staying flexible on others. But while staying flexible on others. But while staying flexible on others. But the headline here is that there's we're the headline here is that there's we're the headline here is that there's we're in the early great standardization of in the early great standardization of in the early great standardization of the platform and tools, not the models. the platform and tools, not the models. the platform and tools, not the models. All right, this is the slide where All right, this is the slide where All right, this is the slide where anyone who's opened an AI bill in the anyone who's opened an AI bill in the anyone who's opened an AI bill in the last year starts nodding. So it turns last year starts nodding. So it turns last year starts nodding. So it turns out that infinite intelligence still out that infinite intelligence still out that infinite intelligence still comes with a usagebased bill. comes with a usagebased bill. comes with a usagebased bill. Once teams are managing many models and Once teams are managing many models and Once teams are managing many models and AI workflows, the next question becomes AI workflows, the next question becomes AI workflows, the next question becomes cost. cost. cost. Cost is now a first class engineering Cost is now a first class engineering Cost is now a first class engineering constraint. We see this in the data. 40% constraint. We see this in the data. 40% constraint. We see this in the data. 40% of respondents say that cost regularly of respondents say that cost regularly of respondents say that cost regularly shapes how ambitiously they use AI shapes how ambitiously they use AI shapes how ambitiously they use AI and another 36% say that it sometimes and another 36% say that it sometimes and another 36% say that it sometimes does. does. does. Well, this is pretty straightforward. So Well, this is pretty straightforward. So Well, this is pretty straightforward. So all in about uh three out of four all in about uh three out of four all in about uh three out of four respondents are adjusting their AI usage respondents are adjusting their AI usage respondents are adjusting their AI usage based on cost and maybe the fourth has a based on cost and maybe the fourth has a based on cost and maybe the fourth has a company card.

That might be surprising or maybe it's That might be surprising or maybe it's obvious but 12 months ago it was not. obvious but 12 months ago it was not. obvious but 12 months ago it was not. Token maxing is cool. Being able to find Token maxing is cool. Being able to find Token maxing is cool. Being able to find real use cases is amazing but cost is real use cases is amazing but cost is real use cases is amazing but cost is becoming a real big part of the product becoming a real big part of the product becoming a real big part of the product decision today. decision today. decision today. And it shows up in monitoring too. What And it shows up in monitoring too. What And it shows up in monitoring too. What folks are monitoring in production folks are monitoring in production folks are monitoring in production includes cost and token usage as the includes cost and token usage as the includes cost and token usage as the number two thing they watch for. It's number two thing they watch for. It's number two thing they watch for. It's being monitored like an SLA right under being monitored like an SLA right under being monitored like an SLA right under quality itself. quality itself. quality itself. Which brings us to the biggest line item Which brings us to the biggest line item Which brings us to the biggest line item of them all. Agents. of them all. Agents. of them all. Agents. We've been talking about agents for a We've been talking about agents for a We've been talking about agents for a while. Uh this year, as you've seen, as while. Uh this year, as you've seen, as while. Uh this year, as you've seen, as you'll see today, as you've seen in you'll see today, as you've seen in you'll see today, as you've seen in previous days, you're going to talk a previous days, you're going to talk a previous days, you're going to talk a lot about harness engineering. They're lot about harness engineering. They're lot about harness engineering. They're escaping demo world. escaping demo world. escaping demo world. So, we asked respondents what level of So, we asked respondents what level of So, we asked respondents what level of tool permissions their agents typically tool permissions their agents typically tool permissions their agents typically have. And this is where agents start to have. And this is where agents start to have. And this is where agents start to look more real. There are two things look more real. There are two things look more real. There are two things happening at once. First, and I don't happening at once. First, and I don't happening at once. First, and I don't think this is surprising, relative to think this is surprising, relative to think this is surprising, relative to last year, there are far more teams last year, there are far more teams last year, there are far more teams using agents. This year, 95%, this seems using agents. This year, 95%, this seems using agents. This year, 95%, this seems high to me, 95% say they're using high to me, 95% say they're using high to me, 95% say they're using agents, roughly double last year.

agents, roughly double last year. agents, roughly double last year. Second, amongst the teams that are using Second, amongst the teams that are using Second, amongst the teams that are using agents, those agents are much more agents, those agents are much more agents, those agents are much more likely to have write access. Last year, likely to have write access. Last year, likely to have write access. Last year, 52% of folks building with agents said 52% of folks building with agents said 52% of folks building with agents said their agents could actually write data. their agents could actually write data. their agents could actually write data. This year, that number is 89%. This year, that number is 89%. This year, that number is 89%. So when you combine these two shifts, So when you combine these two shifts, So when you combine these two shifts, more teams using agents and more of more teams using agents and more of more teams using agents and more of those agents having write permissions, those agents having write permissions, those agents having write permissions, the share of all the respondents and the share of all the respondents and the share of all the respondents and again it's a survey using write enabled again it's a survey using write enabled again it's a survey using write enabled agents is up more than three times agents is up more than three times agents is up more than three times relative to last year. So this is really relative to last year. So this is really relative to last year. So this is really the big shift. Agents are no longer the big shift. Agents are no longer the big shift. Agents are no longer reading, summarizing, drafting. They're reading, summarizing, drafting. They're reading, summarizing, drafting. They're taking actions inside of systems. And taking actions inside of systems. And taking actions inside of systems. And that raises the obvious question, how that raises the obvious question, how that raises the obvious question, how are we controlling all of this? are we controlling all of this? are we controlling all of this? um with pretty blunt instruments. Uh um with pretty blunt instruments. Uh um with pretty blunt instruments. Uh very there are many ways that folks are very there are many ways that folks are very there are many ways that folks are controlling agents today. The top two controlling agents today. The top two controlling agents today. The top two are human in the loop approvals and are human in the loop approvals and are human in the loop approvals and gating permissions which are the right gating permissions which are the right gating permissions which are the right instincts but kind of the same toolkit instincts but kind of the same toolkit instincts but kind of the same toolkit you'd use to manage an intern.

you'd use to manage an intern. you'd use to manage an intern. Below that the results scatter. Task Below that the results scatter. Task Below that the results scatter. Task decomposition, retrieval, memory, decomposition, retrieval, memory, decomposition, retrieval, memory, sandboxing. People are trying sandboxing. People are trying sandboxing. People are trying everything. Nobody has settled the everything. Nobody has settled the everything. Nobody has settled the control layer for agents. uh memory and control layer for agents. uh memory and control layer for agents. uh memory and persistent context is one that I'm persistent context is one that I'm persistent context is one that I'm watching very carefully right now. I watching very carefully right now. I watching very carefully right now. I think it's going to evolve a lot in the think it's going to evolve a lot in the think it's going to evolve a lot in the next year. And when agents fail or when next year. And when agents fail or when next year. And when agents fail or when people complain about agents failing to people complain about agents failing to people complain about agents failing to be more precise, it's usually the be more precise, it's usually the be more precise, it's usually the thinking, not the plumbing. So, uh you thinking, not the plumbing. So, uh you thinking, not the plumbing. So, uh you know, like twothirds say that know, like twothirds say that know, like twothirds say that hallucination or losing context mid task hallucination or losing context mid task hallucination or losing context mid task is what frustrates them the most. is what frustrates them the most. is what frustrates them the most. All right, so agents are out in the All right, so agents are out in the All right, so agents are out in the wild, which makes it a good time to look wild, which makes it a good time to look wild, which makes it a good time to look at what everyone's actually running at what everyone's actually running at what everyone's actually running underneath. So let's take a peek at the underneath. So let's take a peek at the underneath. So let's take a peek at the stack. stack. stack. Um, we asked, what is the biggest Um, we asked, what is the biggest Um, we asked, what is the biggest challenge in your stack? Every single challenge in your stack? Every single challenge in your stack? Every single year that I ask this, the answer, the year that I ask this, the answer, the year that I ask this, the answer, the number one answer is eval. Um, so Eval's number one answer is eval. Um, so Eval's number one answer is eval. Um, so Eval's lead here is same as always, but by a lead here is same as always, but by a lead here is same as always, but by a very thin margin, like that margin is very thin margin, like that margin is very thin margin, like that margin is getting smaller. And I'll say the quiet getting smaller. And I'll say the quiet getting smaller. And I'll say the quiet part here, which is that 96% of the part here, which is that 96% of the part here, which is that 96% of the people in the survey in this room have a people in the survey in this room have a people in the survey in this room have a problem with the stack. You just can't problem with the stack. You just can't problem with the stack. You just can't agree on which one. Um, so if you're agree on which one. Um, so if you're agree on which one. Um, so if you're deciding what to build next, if you're deciding what to build next, if you're deciding what to build next, if you're interested in infrastructure, that interested in infrastructure, that interested in infrastructure, that scatter is the map.

scatter is the map. scatter is the map. And the leading challenge, how to And the leading challenge, how to And the leading challenge, how to evaluate your AI outputs requires many evaluate your AI outputs requires many evaluate your AI outputs requires many different methods, but as always, the different methods, but as always, the different methods, but as always, the vibe review is number one. So there are vibe review is number one. So there are vibe review is number one. So there are some consistent things that we'll see if some consistent things that we'll see if some consistent things that we'll see if they change over the time, but they they they change over the time, but they they they change over the time, but they they have not changed. have not changed. have not changed. Okay, this is interesting. So across Okay, this is interesting. So across Okay, this is interesting. So across eight layers of the stack, we asked what eight layers of the stack, we asked what eight layers of the stack, we asked what do people build versus buy. Um again, do people build versus buy. Um again, do people build versus buy. Um again, maybe the corporate card is is going to maybe the corporate card is is going to maybe the corporate card is is going to play a part in this, but there is a wide play a part in this, but there is a wide play a part in this, but there is a wide range and mix for every layer of the range and mix for every layer of the range and mix for every layer of the stack and a few clear takeaways. stack and a few clear takeaways. stack and a few clear takeaways. So the first is that inference and model So the first is that inference and model So the first is that inference and model serving is the layer that people buy the serving is the layer that people buy the serving is the layer that people buy the most. Many people don't want to build most. Many people don't want to build most. Many people don't want to build inference infrastructure and fair inference infrastructure and fair inference infrastructure and fair enough. Uh prompt management is the enough. Uh prompt management is the enough. Uh prompt management is the opposite. 61% build it themselves. Um opposite. 61% build it themselves. Um opposite. 61% build it themselves. Um apparently everyone's prompts are apparently everyone's prompts are apparently everyone's prompts are special. And this is true of a lot of special. And this is true of a lot of special. And this is true of a lot of the product logic prompts rag eval. They the product logic prompts rag eval. They the product logic prompts rag eval. They tend to stay inhouse on a relative tend to stay inhouse on a relative tend to stay inhouse on a relative basis. fine-tuning is the clearest not basis. fine-tuning is the clearest not basis. fine-tuning is the clearest not yet. Like most people don't have it at yet. Like most people don't have it at yet. Like most people don't have it at all and uh folks are pretty locked in.

all and uh folks are pretty locked in. all and uh folks are pretty locked in. So those who bought aren't looking as So those who bought aren't looking as So those who bought aren't looking as much to build. Those who built aren't much to build. Those who built aren't much to build. Those who built aren't looking as much to buy. Uh but those are looking as much to buy. Uh but those are looking as much to buy. Uh but those are those are the core takeaways from the those are the core takeaways from the those are the core takeaways from the usage in our stack. So many of you work on teams and like we So many of you work on teams and like we said at the start these range from solo said at the start these range from solo said at the start these range from solo founders to large enterprises. founders to large enterprises. founders to large enterprises. What is this doing to teams? What is this doing to teams? What is this doing to teams? And remember this is a builderheavy And remember this is a builderheavy And remember this is a builderheavy sample. But among builders the vibes are sample. But among builders the vibes are sample. But among builders the vibes are good which you know I'm sure if you look good which you know I'm sure if you look good which you know I'm sure if you look to your left and your right you're to your left and your right you're to your left and your right you're feeling that the vibes are pretty good. feeling that the vibes are pretty good. feeling that the vibes are pretty good. 97% report a net positive effect on 97% report a net positive effect on 97% report a net positive effect on their organization. their organization. their organization. The top effect isn't really just speed. The top effect isn't really just speed. The top effect isn't really just speed. It's cheaper failure, more It's cheaper failure, more It's cheaper failure, more experimentation, more prototypes, more experimentation, more prototypes, more experimentation, more prototypes, more bets. It didn't just make engineers bets. It didn't just make engineers bets. It didn't just make engineers faster, but it made trying things nearly faster, but it made trying things nearly faster, but it made trying things nearly free. And so there's some happy campers free. And so there's some happy campers free. And so there's some happy campers as a result of that. as a result of that. as a result of that. But it's not free free. You know, But it's not free free. You know, But it's not free free. You know, there's no free lunch as nothing is. So there's no free lunch as nothing is. So there's no free lunch as nothing is. So the same tool that increases the same tool that increases the same tool that increases experimentation also increases review experimentation also increases review experimentation also increases review burden. Both can be true. And um you burden. Both can be true. And um you burden. Both can be true. And um you know o over nine and 10 respondents are know o over nine and 10 respondents are know o over nine and 10 respondents are feeling negative downstream effects in feeling negative downstream effects in feeling negative downstream effects in some way. The most common ones being wid some way. The most common ones being wid some way. The most common ones being wid you know widely discussed at this you know widely discussed at this you know widely discussed at this conference uh online and anywhere that conference uh online and anywhere that conference uh online and anywhere that you see AI engineers erosion of deep you see AI engineers erosion of deep you see AI engineers erosion of deep technical skills and understanding of technical skills and understanding of technical skills and understanding of the codebase.

the codebase. the codebase. And these are consequences of cheap code And these are consequences of cheap code And these are consequences of cheap code generation. generation. generation. And the org chart is really feeling it. And the org chart is really feeling it. And the org chart is really feeling it. So So So many folks, 81% are saying that AI is many folks, 81% are saying that AI is many folks, 81% are saying that AI is blurring the line between their role as blurring the line between their role as blurring the line between their role as engineers and product design and engineers and product design and engineers and product design and marketing. These stats shocked me. Um, marketing. These stats shocked me. Um, marketing. These stats shocked me. Um, where you feel it the most is shipping where you feel it the most is shipping where you feel it the most is shipping software once exclusively the engineers software once exclusively the engineers software once exclusively the engineers domain. I know folks talk about vibe domain. I know folks talk about vibe domain. I know folks talk about vibe coding and how that's accessible to more coding and how that's accessible to more coding and how that's accessible to more folks than ever before in different folks than ever before in different folks than ever before in different roles, but today over a third of teams roles, but today over a third of teams roles, but today over a third of teams have non-developers shipping features, have non-developers shipping features, have non-developers shipping features, which was pretty wild to me. Mostly which was pretty wild to me. Mostly which was pretty wild to me. Mostly smaller, mostly internal, but 17% say smaller, mostly internal, but 17% say smaller, mostly internal, but 17% say that non-developers are regularly that non-developers are regularly that non-developers are regularly shipping customerf facing features shipping customerf facing features shipping customerf facing features across the stack. across the stack. across the stack. And even when non-developers aren't And even when non-developers aren't And even when non-developers aren't shipping, a third of teams see them shipping, a third of teams see them shipping, a third of teams see them building really useful things, building really useful things, building really useful things, prototypes, front-end mocks, and more.

prototypes, front-end mocks, and more. prototypes, front-end mocks, and more. So, shipping software is not gated on So, shipping software is not gated on So, shipping software is not gated on being an engineer. We knew this, but uh being an engineer. We knew this, but uh being an engineer. We knew this, but uh the extent to which it's being pushed is the extent to which it's being pushed is the extent to which it's being pushed is is higher than I expected. is higher than I expected. is higher than I expected. All right, so where does all of this go? All right, so where does all of this go? All right, so where does all of this go? We always ask people to place bets rapid We always ask people to place bets rapid We always ask people to place bets rapid fire. So, let's talk about those fire. So, let's talk about those fire. So, let's talk about those results. results. results. Um, Um, Um, so present tense first. 76% say AI so present tense first. 76% say AI so present tense first. 76% say AI boosted their job satisfaction. So boosted their job satisfaction. So boosted their job satisfaction. So that's good for most of this crowd. I that's good for most of this crowd. I that's good for most of this crowd. I hope you're uh as uh Alphaba and Glenda hope you're uh as uh Alphaba and Glenda hope you're uh as uh Alphaba and Glenda say, I hope you're happy now. Um, that's say, I hope you're happy now. Um, that's say, I hope you're happy now. Um, that's great. But 59% fear today's AI code great. But 59% fear today's AI code great. But 59% fear today's AI code creates long-term liabilities. creates long-term liabilities. creates long-term liabilities. Only a third call software engineering a Only a third call software engineering a Only a third call software engineering a solved problem. Although uh when I have solved problem. Although uh when I have solved problem. Although uh when I have conversations with folks sometimes the conversations with folks sometimes the conversations with folks sometimes the way in which they define software way in which they define software way in which they define software engineering is different. So you can engineering is different. So you can engineering is different. So you can read into that stat as you will. Um read into that stat as you will. Um read into that stat as you will. Um happier faster but embracing the happier faster but embracing the happier faster but embracing the maintenance build is the TLDDR and maintenance build is the TLDDR and maintenance build is the TLDDR and people are unsure what's going to happen people are unsure what's going to happen people are unsure what's going to happen with hiring.

with hiring. with hiring. And for the five-year bets And for the five-year bets And for the five-year bets we have 67% expect a leading lab will we have 67% expect a leading lab will we have 67% expect a leading lab will declare AGI in the next five years. Note declare AGI in the next five years. Note declare AGI in the next five years. Note the wording. We said will, we asked the wording. We said will, we asked the wording. We said will, we asked about the press release, not the about the press release, not the about the press release, not the achievement. So, will they declare it? achievement. So, will they declare it? achievement. So, will they declare it? Yes. What does that mean? Not sure. Uh, Yes. What does that mean? Not sure. Uh, Yes. What does that mean? Not sure. Uh, only 9% bet on Transformers being only 9% bet on Transformers being only 9% bet on Transformers being state-of-the-art in 5 years. Most are state-of-the-art in 5 years. Most are state-of-the-art in 5 years. Most are unsure. Uh, but that was interesting. unsure. Uh, but that was interesting. unsure. Uh, but that was interesting. And then my favorite, will there be more And then my favorite, will there be more And then my favorite, will there be more AI compute in space or on land? 36 yes. AI compute in space or on land? 36 yes. AI compute in space or on land? 36 yes. 38 no. The most divisive question in the 38 no. The most divisive question in the 38 no. The most divisive question in the survey is about outer space. survey is about outer space. survey is about outer space. I promised you a lot of bar charts and I promised you a lot of bar charts and I promised you a lot of bar charts and that was a lot of information. So a that was a lot of information. So a that was a lot of information. So a review or our 2026 wrapped impact is review or our 2026 wrapped impact is review or our 2026 wrapped impact is overwhelmingly positive. Image genen overwhelmingly positive. Image genen overwhelmingly positive. Image genen doubled or happy image gen doubled while doubled or happy image gen doubled while doubled or happy image gen doubled while audio has the highest adoption intent audio has the highest adoption intent audio has the highest adoption intent the same as last year. cost really the same as last year. cost really the same as last year. cost really became a first class constraint and we became a first class constraint and we became a first class constraint and we see that everywhere in monitoring in how see that everywhere in monitoring in how see that everywhere in monitoring in how ambitious folks that are going out and ambitious folks that are going out and ambitious folks that are going out and building AI products are behaving.

building AI products are behaving. building AI products are behaving. Open weights augment but they don't Open weights augment but they don't Open weights augment but they don't replace. So we're seeing a multimodel replace. So we're seeing a multimodel replace. So we're seeing a multimodel future with a consolidation of the future with a consolidation of the future with a consolidation of the stack. stack. stack. Agents got right access more than ever Agents got right access more than ever Agents got right access more than ever before, tripling relative to last year. before, tripling relative to last year. before, tripling relative to last year. While the guardrails stayed pretty While the guardrails stayed pretty While the guardrails stayed pretty primitive, and inference is the buy primitive, and inference is the buy primitive, and inference is the buy market, everything closer to product market, everything closer to product market, everything closer to product logic tends to relatively stay more logic tends to relatively stay more logic tends to relatively stay more in-house. It is a very exciting time to in-house. It is a very exciting time to in-house. It is a very exciting time to be an AI engineer. I cannot wait to see be an AI engineer. I cannot wait to see be an AI engineer. I cannot wait to see how the next year unfolds. how the next year unfolds. how the next year unfolds. So, you can find the full report in the So, you can find the full report in the So, you can find the full report in the link up here. Every chart plus some cuts link up here. Every chart plus some cuts link up here. Every chart plus some cuts that we didn't have time for today. Um, that we didn't have time for today. Um, that we didn't have time for today. Um, I won't ask you to fill out a survey I won't ask you to fill out a survey I won't ask you to fill out a survey about the survey, but if there's about the survey, but if there's about the survey, but if there's something that you want on the books for something that you want on the books for something that you want on the books for 2027, something you're curious about, 2027, something you're curious about, 2027, something you're curious about, you can come find me here on the you can come find me here on the you can come find me here on the internet. I'm easy to spot. Thank you so internet. I'm easy to spot. Thank you so internet. I'm easy to spot. Thank you so much. Uh, we will see you next year or much. Uh, we will see you next year or much. Uh, we will see you next year or per 36% of you, maybe in orbit. Thank per 36% of you, maybe in orbit. Thank per 36% of you, maybe in orbit. Thank you.