# AI Agent Production Shifts and Infrastructure Constraints

**Podcast:** The InfoQ Podcast
**Published:** 2026-10-30

## Transcript

The decisions you're making right now about AI adoption, architecture trade-offs, and how your team works together will shape your systems for years.
Getting those calls right when the landscape is shifting this fast is hard.
QCon San Francisco has spent 20 years connecting senior engineers with practitioners who are a few steps ahead on the same problems.
This November 16th through the 20th, 60 plus speakers across 12 tracks will share what's actually working in production and what isn't.
No hidden product pitches, just senior practitioners helping senior practitioners.
Learn more at QConSF.com.
data engineering community on InfoQ.
I will be facilitating our conversation today.
We have an excellent panel with subject matter experts and practitioners from different specializations in the AI and ML space.
I will go around our virtual room and ask the panelists to introduce themselves.
We will start with our special guest first, Chip Huyen.
Hi Chip, thank you for joining us and participating in this podcast.
Could you introduce yourself and tell our listeners what you've been working on?
Hello, I'm Chip.
I'm very excited to be here today.
It's an honor.
I know that everyone on the panel has amazing background.
I've been following the work for a long time.
Today, I was hoping to share what I know, but I'm still hoping to learn from you all.
I'm working on a lot of AI stuff, AI agent tech, getting AI to work well for workflow.
I'm also very interested in making AI just faster and cheaper.
We saw recently the types of AI that came out.
I think there's a huge demand for cheap and fast intelligence, and that's very exciting.
Thank you, Chipia.
Definitely, if we have time, I would love to learn more about this JAV model from TypeSafe AI, if you have any experience on that.
Daniel, how about you?
Hi, I'm Daniel Dominguez.
I'm the deputy editor at InfoQ.
I focus on artificial intelligence, machine learning, and data engineering.
I'm also a software engineer, and lately I've been particularly interested in AI agents, agentic software development, and how AI is moving from a digital system into robotics and physical AI.
Thanks, Daniel.
Arthur.
Hi everyone, I'm Arthur.
I'm a researcher on technology strategies, basically.
And these days I split my time between consulting and advising companies about how to make AI useful in real systems and apply research basically on multi-agent systems and belief revision.
Thank you.
And Hien?
Hi everyone, my name is Hien Lu.
I mainly focus on ML infrastructures as well as, I guess nowadays, agentic engineering.
Currently, I'm helping with a startup on consulting on AI inference and a long-term fan of QCon, also an editor of InfoQ, also a QCon program chair for QCon AI New York.
Yeah, thanks, definitely.
I'm very excited to talk with you all about this topic because, I mean, you have been part of InfoQ and the QCon family for so long.
It's always great to chat on these topics.
So I'm looking forward to speaking with you about what's happening in the AI and ML space, where we currently are, but more importantly, where we are going, right?
It's a fast pace as the AI technology innovations keep happening since we discussed this report last year.
Before we start with the actual podcast topics, I want to go over a couple of quick housekeeping items for our audience.
There are two major components to these 10 reports.
One is obviously this podcast.
which is an opportunity for you to listen to the panel of expert practitioners and see how the AI innovations are disrupting the industry.
The second part of the trend report is a written article that Arthur also mentioned.
We can cover a lot more topics in that article.
So that will be available on InfoQ website along with the podcast recording.
That's where we actually assess the current technology adoption and we have an adoption graph that we update every year.
to highlight which technologies are further ahead in the adoption cycle than the other technologies.
It's always a great reference for anybody who is following the technology innovations in the AI space.
So definitely look for those two parts, the podcast as well as the article.
And also I definitely recommend to you to check out all the InfoQ publications.
In this podcast, we will definitely discuss the important developments in AI technologies that took place.
since our last report back in August last year.
But due to a limited time in this forum, we won't be able to cover all the topics or even get into in-depth discussions on some of the topics.
But we will try to follow up on those important topics as separate podcasts or separate articles because they are important for us to learn about.
But again, I encourage you to check out the InfoQ website and AI and ML and data engineering community, which includes news items, articles, podcasts like this, and presentations.
So it's also worth mentioning, you know, with all the AI stuff happening, that this podcast is not AI generated or even AI assisted.
It is being recorded with real people who have real world experience.
I just want to get that disclaimer out of the way, right?
So contrary to all the predictions being made out there, that AI is going to replace everything.
So that being said, let's get started.
It's been almost four years since ChatGPT came out, right?
Back in November, 2022.
It's been two or three years since RAG Technologies kind of started seeing enterprise adoption and autonomous LM agents came out.
So definitely a lot of things have been happening.
Agenting AI has been the kind of main dominant topic for a long time now.
So we'll definitely look at that a little bit today.
We are seeing more developments in the quality assurance space with the AI evaluations and hardness frameworks and all the other good topics.
We'll get into more of those details later.
But the first topic is I would like to get a kind of quick...
take maybe 30 seconds take from each of the panelists on how do you see the current state of AI in 2026?
Chip, could you start us off?
So it's going absolutely crazy.
So I'm tracking like all the money in and out of the AI company.
So money in is in like private investment and then money out acquisitions and IPO, right?
So far for many, many years, we see the charts are very similar, like one of the three numbers because the money in has to come out at some point.
But in recent years, it's like go like this.
The money in is like go exponentially and the money out is no money out at all.
It's very exciting.
It has an amount of money in.
It's like, I think, proportional to like, I guess I use as a proxy for people excitement.
There's a lot of companies.
The number of company sizes going down is the average, like median company size going down significantly.
So that means that like people can do a lot of things that anyone can build nowadays.
Anyone can have a product.
So the issue is not really about how to build things anymore.
Anyone can build things, right?
But what to build and then how do we as a receiving end filter through all of that is quite insane.
And I think nowadays going through these kind of things like winner catch all situations because a few products somehow surface and everyone just talk about them.
And the rest is lost in the shuffle.
It's very, very tricky.
It's just like everyone's like force-fights the same kind of ideas.
And at the same time, we have a lot of new things coming out.
But somehow everyone is talking about the same thing.
We don't get it.
Yeah, it's very, very strange.
How would you hear?
What are you seeing as an AI in 2026?
Yeah, obviously, like Chip said, there's a lot of excitements, innovations, product announcements.
But if you want to talk about like...
At the beginning of like 2026 or the end of 2025, we're talking about the year of agents.
I think we are in a phase where the AI agent stacks.
It's being built in a way, the different layers.
There's a lot of progress there, a lot of exciting, a lot of innovations for all of us, how to learn how to build AI agents.
And now everybody's trying to figure out how to operate them, how to bring them to production.
So that's very exciting space from an enterprise perspective.
I think the other one that I'm sure everybody is aware of is writing code.
It's easy now.
We're coding agents.
So I think there's many exciting ideas going to be built.
And we all need to adjust in terms of how we build software moving forward.
So I think that's another area that's happening in 2026.
And we'll definitely see more excitements in 2027.
I think we'll get into those topics a little bit more here later anyway, so we'll have time.
So how about you, Arthur?
What do you see at a high level?
Yeah, I think follow up on here, I mean, I think what we're seeing now is we're changing or shifting our perspective from the parts of the car to the car itself.
I mean, you have impressive isolated capabilities in each model being developed year after year.
And now you're starting to realize as companies in production that we need them to work and to be dependable.
So, I mean, you can have a pilot, a great agent that can work beautifully, you can have curated data, and that's great.
very good prediction that you have your data is messy, you have different permissions, you have regulations, you have the system has to be affordable, it has to be recoverable.
So that feels like the defining test for AI that we're seeing emerging right now is not whether it can do something once, but whether a team can trust it in a real workflow to solve real problems.
So that's when you see it actually being applied.
in companies that have been successful.
Thanks, Alter.
How about you, Daniel?
What are you seeing?
I think the biggest change in 2026 is that AI is moving from being something that we interact to something that we delegate, like Hain was saying.
I think most AI applications over the last few years were based on a simple interaction.
You just ask a question and the mobile gives you an answer.
Right now, we're starting this year, we are starting to see a system that...
can plan, can use tools, can execute actions, can do their own work and continue working for much longer period of time.
So I think this year is less about another jump in the model intelligence and more about turning that intelligence into actual systems.
Thanks, Daniel.
Thank you for that.
Yeah, we can, that's a good kind of segue from there, right?
With AI agents and all the other innovations happening, the foundation is still on those models, on language models.
So maybe that's where we can start our kind of main discussion today because There's always a lot of new developments happening there.
So in 2026, my observation is the language models have shifted from these massive size models, billions of parameters into the smaller size models, small language models, right?
Where they can be faster in terms of performance, may not be as accurate, but accurate enough for most of the use cases.
And also we are seeing the rise of reasoning models that can think.
through math and logic problems before actually responding, right?
So a lot of good stuff happening.
So we can start with this as our main topic.
Arthur, I know you've been looking at some language model topics and also frontier models.
So what do you see is happening here?
What our listeners should be aware of or should be learning about?
Yeah, I think in general, when you focus on specific models, we have better reasoning, we have better coding, we have better multimodal abilities.
They're emerging.
It's evolving, right?
But then we're going to...
go back to what I said first, that if I were to pick a headline for frontier models in 2026, I'd say that the model itself is becoming less of the whole story.
So you do have the capability, you do have to worry about it, of course, but we're changing the question from which model is best to which model or combination of models gives us the right capability, latency, data boundary costs for this workload, specifically this workload.
To me, at least, the trend seems to be a move from choosing one model to deliberately matching models to workloads, which makes it a system design question, not a model choosing question, right?
And I see that a sign that the field is maturing because the model stops being the whole problem and becomes the one component of the architecture.
So we do have to worry about all the specific capabilities of each model, but they have to work in unison together.
Any other thoughts from anyone else on the models?
I think model size is a little bit interesting because the definition has changed over time.
Because the model size number of parameters, it also depends on how many of them are actually active at the same time.
It's a trillion parameters model, but it's only a few percentage.
And also another is the quantizations.
I do think there's very, very exciting work on quantization recently.
I feel like about two, three years ago, people thought it was crazy to have a four-bit model.
But now we have consistently see people running like one attitude bit model, which is like a significant reduction in size.
And like we think of like as a compute requirement, it's not a compute efficiency of these models.
So I think it's interesting to see.
Another trend they see from models is that of like a release cycle, because like let's say that you're from Chile, you have a very, very strong model, right?
You release this out at the top of every benchmark.
And then a few weeks later, you will see that there's another model, much cheaper, smaller, like almost like.
equivalent, but maybe one or two percent drop in performance.
So at HEMAS, this is usually a trend of like, yes, we have a big model and then you just steal that into like a lot of other techniques, like you think I better make sure of expert, making it more sparse, quantization aggressively quantize into like smaller and cheaper model.
Yeah, there was a lot of interest in like doing fine tuning toward the beginning of the year.
And there's a lot of innovation in that space with lore and all that stuff like that.
all these kind of NeoCloud inference providers started providing the ability to do fine-tuning for at least the open-source models.
I would love to hear from the panel if you have heard successful use cases in the enterprises that actually do fine-tuning for their specific use cases.
I think there's some happening, but I don't hear much about it.
Obviously, it takes a lot of effort to do that with the right skill sets.
maintenance, operations, doing inference, the whole shebang is doable, but whether that is making it reliable and in operations and all that stuff like that is still a lot of work.
So maybe things will be easier in 2027 and people do it more, but that's still a question in my mind of how pervasive that is being actually done in enterprises.
That's a good point here.
Like how do you make the model management lifecycle kind of similar to software development lifecycle, right?
How do we make it repeatable and streamlined?
Exactly.
Yeah.
Unlike, you know, a traditional ML, you start with data and then you train and you evaluation and then you bring it online and here you do fine tuning.
So it's still, it's a lot of work, but I'm sure there are specific use cases, particularly perhaps these AI native company that in the domain of legal, right?
Harvey and such.
I'm sure they do fine tuning there.
But beyond that, I want to, if enterprises.
They actually do it.
That's kind of my question.
I've seen a couple of companies doing that, but they're very specialized in what they do and very, very careful about IP.
So they won't disclose that for a few years, at least, I think.
But usually I'm seeing this basically in healthcare and finance.
It's concentrated there for some reason.
You know the reason, but yeah, it's happening.
Just need the budget, but it's already happening.
It's great.
Makes sense.
Yeah, and was this about you asked, like, has any of your company, like, 5200 model?
After I mentioned, it's like waiting for budget.
Is it supposed to be expensive?
I think it depends on what you're doing, how you're doing.
For example, when it comes to open models, you know, there's a, I've seen a small rush of open models and local models and people thinking that, okay, it's going to be cheaper.
But think about open.
When I have the word open, I have open waves, open licensing, transparent data.
I have a lot of things involved.
And they don't necessarily mean lower costs or commercial license or whatever.
So depending on what you're doing, you need massive data or you want to get traced out of big models to train smaller models.
The thing is, I think it's still more of a craft than the artisan thing, the process, you know?
So people are discovering this.
And although you have the process for training models and for training things, applying that to your real world problems is still a process of discovery.
So they end up spending a lot of money.
I think that comes in there fairly.
It's R&D basically.
Yeah, hopefully some of these learnings from these companies will be shared with the community next year, next couple of years, as Arthur mentioned.
Yeah, I want to kind of maybe talk a little bit about the model distillation topic.
Chip, I know you talk about this in your book titled AI Engineering.
I highly recommend everybody to check out the book by Chip, one of the very valuable books in the AI space.
So you talk about model distillation.
Can you discuss how it works and how it can help the overall model usage?
Yeah.
So I think model distillation is extremely important.
It's very exciting.
And I think a lot of us today benefit from model distillations, either done by some providers or some application developers.
So the idea is that you have...
a very big model performs really, really well.
But of course, these big models are very expensive.
Of course, a lot of them are very slow.
So Cambridge is yet another smaller model to imitate and learn from these big models so that you can achieve comparable or most comparable performance, but a much cheaper price.
So a lot of time you can get, so we care about something called compute efficiency.
So we usually think about, okay, for this much drop in the compute efficiency, how much of the performance we have to sacrifice?
And this has, magic has gone and going really, really, really well in recent years.
I think my people are aggressively developing a lot of new techniques and it's working things that I didn't think would be possible even just like a few years ago.
Also in the models, we have local versus on the cloud models, right?
We can run some of these models locally using OLAMA or the other LLM engines.
And also there's a multimodal models like, you know, for the text and images.
So anybody have any comments on those topics regarding these models on-prem versus cloud or the multi-modal models?
I think regardless of the objective of defining what you're using, you need to separate a few questions about what we're doing.
Because we've seen people defining on-prem versus cloud as a full thing or really separate, really binary thing.
I mean, it depends on where the client runs, where the model runs, where the prompts, two outputs, logs go.
Who keeps them?
Everything is involved, right?
So in the end, most setups tend towards a hybrid approach.
They're neither fully implemented or the cloud.
And the thing is, why is it important to define that for the company?
What are they trying to solve?
Is this a budget problem?
Is this an architecture problem?
So, I mean...
okay, you can categorize them, but what for?
What we're doing with it?
So which questions are you asking about this?
It's not as simple as choosing a category and that's it.
Thanks, Arthur.
Yeah, definitely a lot of other stuff happening.
One thing I just want to mention briefly about the JEV models, right?
These are what they call system one models.
I just started learning about this.
I have no in-depth knowledge on this, but yeah, I think we will definitely share more about this in the future.
But has anybody else done any?
Reading on this, like this JEO model from TypeSafe AI, that they talk about, you know, like very fast inference and very low cost compared to the LLMs.
Yeah, it's supposed to be my weekend experiment for tomorrow.
I've been watching a lot of videos on this.
And recently I watched a podcast from Latent Space where they interviewed the founder of TypeSafe.
Very, very fascinating piece of technology.
One of the key things I took away from that is a different type of model, but it's also extremely fast.
I think that's going to help unlock an interesting new set of use cases.
There are a bunch of interesting demos about using that kind of decision to build UI and other use cases that you might have not considered.
Like, hey, is that a relevant use case?
But it's a lot of us to learn and figure out how to incorporate this.
But I think the key idea is, at least for the founder is, what he wants to do is to build a model that is for builders and the output is for software, not for human, right?
That's one of the key things he pointed out.
It's basically evaluates the state and returns the answers.
It's mainly for the software.
So we can maybe, he and you mentioned about inference.
I heard the same thing from my own experience.
So 2026 is being dubbed as the year of inference, right?
So could you talk about how we are shifting from training models to running AI inference and what does all this mean in the real world usage?
What are you seeing in terms of inference frameworks?
Totally.
The inference topics, I mean, it's exploded since beginning 2026.
Obviously from one end, you have all these inference stack, right?
From VLM, SG Lang, and so on that are helping the rest of us.
to spin up our own inference service in our own environment.
And at the same time, these open-weight models are becoming more and more capable.
And that space is like initially not a whole lot of people how it works and how to operate it.
But now there's a lot more information and resources out there about how to actually use this software, how to optimize them.
And at the same time, some of the use cases we talk about the, you know, financial companies where you have PIIs.
So the ability to host these open waste models internally for those kind of specific use cases are very, very relevant.
And I believe that that's going to continue to grow, continue to be the case where a company will do more of this moving forward.
Now that is a lot more approachable, more easier to do.
There are some companies that because they want to save costs, they actually want to host internal open-way models for coding, for example.
So for many reasons, that inference is going to continue to be a very, very useful, very popular topic.
And more and more companies will start doing this moving forward, in my mind.
Yeah, it's very exciting.
For the work on inference, do you have focus on specific chips or hardware architecture?
Do you inference on?
Not so much at this point because AMD is just coming online, but more and more inference provider, at least the NeoCloud, they want to be able to do inference across chips, you know, NVIDIA and AMDs and potentially near in the future, Intel and so on.
But at the moment, still predominantly NVIDIA from what I gather.
I'm curious because I think it's a trend of as you get like newer and newer chips, right, more powerful chips.
like many, many, many older generation of chips, right?
And somebody, like, they almost come for free because a lot of funcillas don't use them.
I thought, like, what do we do with them?
So I think, like, getting those done on, like, older generation of chips, like, making things faster could be very interesting.
Having, like, smaller, I think it's tied with, like, model distillations.
You've got to have a smaller, cheaper model that requires less compute footprint.
You can feed them onto an older generation of chips.
And also a lot of upcomers for chip because it's very hard to compute NVIDIA on their flagship architecture.
But you can definitely compete in a lower tier architecture.
That's probably the next area of the competition, right?
I heard companies like NVIDIA and Apple are definitely dominating the hardware space and in-depth demand.
Cerebris is an interesting chip player.
And I think we'll see more of that maybe later on this year.
Ian and Chip, you mentioned in a previous conversation about agentic loops and hardness engineering and context engineering.
Can you share, maybe starting with Chip, can you share what you see happening in the context engineering space?
I do think definitely in the last couple of years, agentic coding use has dominated everything.
But the thing with the TypeScript launch, it shows that actually people have been so focused on coding.
That's a huge tap area for like...
non-coding use cases.
For me, coding is definitely very important, but I definitely realize it's not the only thing.
Yeah, Arthur, what are your thoughts?
Context pollution is actually the biggest problem, the problem worth emphasizing now.
Because we've seen in the past year or so movements toward bigger context windows in general, but they don't really solve the issue if the system cannot distinguish what's useful, what's current, what's authoritative information.
from irrelevant, stale, or virtual material.
So I'm bringing something from research here, but in general, more context can make an agent slower, more expensive, less reliable.
So the practical question would be, if you have your context window, regardless of the size of it, what enters the context?
From which source?
With which permission?
For what step of the workflow?
So you don't have this huge pollution.
It tests topics on retrieval qualities, social attribution, compression rules, context modeling in general, context compression.
So, I mean, there is a huge, huge heap of problems underneath the word context.
That's how I see it.
Thanks for seeing that.
So, yeah, in the interest of time, I do want to kind of move to the topics.
I want to focus on the agents and the evals a little bit more in this discussion.
So we can jump into the agents topic.
Daniel, I know you've been...
Looking into this space for a while now, so this continues to be one of the major trends in the AI space.
Agents, they have gone from, like you mentioned earlier, from just coding assistants and now they can act on several different aspects.
AI agents are probably one of the biggest trends of 2026.
What I find interesting is that agents are moving through several levels very quickly.
So we started with personal agents helping one person.
Then we start seeing agents working with teams, and now companies are thinking about agents operating across entire organizations.
So I think the important change is that we're moving from beyond just the chatbot model.
Instead, we are simply generating an answer, and agents can have goals, can use tools, can interact with software, can make decisions, can keep working until the task is complete.
So I think that changes AI from interface into something closer to an execution layer.
We can see this mainly in software development.
Software development is probably where this transition is more visible.
So we moved from autocomplete to copilots, and now we have coding agents.
We have tools like Cloud, Codex, Cursor, Replit, and others.
that can navigate repositories, can modify multiple files, can run commands, can execute tasks, can instead failures, and can iterate on their own.
So the interesting question is no longer can AI write code.
We already know that AI can write code.
I think the more interesting question is how much of the software development loop can we delegate to an agent while we still maintain quality, security, and control?
that is starting to change the developer's job.
So I think developers are spending less time producing every line of code and there is more time defining to intent, giving context, reviewing decisions and validating what agents can produce.
So I think the recent developer surveys also show how quickly agent coding has become part of everyday development workflows.
So I think there are a lot of...
discussion right now on the interface disappearing, the agents and enterprise integration, multi-agent systems, all the challenges that it's bringing.
So definitely this is something that we're going to continue seeing probably next year and a couple of next years.
All this agent and agentic phase working across organizations and on a daily workflow of not only developers, but...
in other areas of the work.
Yeah, thanks, Daniel.
I think as the AI tools generate more and more code, I don't think that will be any different than humans doing it.
Maybe they'll be faster.
But in terms of the defects, I think I'm sure the code will not be perfect right off the bat.
So how do we see these generated code?
How do we make sure that it's safe and secure and in compliance with the corporate standards?
Arthur, how do we safely connect AI agents into the...
software delivery platform and the SDLC process?
That's a big question, isn't it?
First, we need to separate coding assistants from coding agents.
Assistants would be something like an agent helping a developer navigating a code base, you're writing a task, drafting a patch, something like that.
Agents, on the other hand, means the system can run comments, modify a repository, open pull requests, affect deployment in any way.
And for that, we need operation design.
So for operation design, you need all the steps.
You need authority levels, permissions.
You need to audit your workflows, see what's happening there.
And the thing is, it's really interesting to see how the tools that work with that are emerging because the first ones are obviously based on the human flow.
And we just had a discussion about Jeff, right?
We are seeing more and more tools that are...
realize that, wait, an agent is not a promoter here, okay, but it's not really a person.
It's a piece of software.
So we need to write tools for the software to use, not for a person to use.
We're seeing this change right now.
I'm curious to see how it's going to affect this operation design for great permissions for structuring what agents can and cannot do from the coding agent's perspective, not from the coding system perspective.
Yeah, definitely.
That will be a big...
I think that's going to be, you know, to keep watching, right?
Thank you for that.
Yeah, one thing I want to add to discussions of agent is, there's two things.
One is like harness engineering.
I think that an extension of contacts engineering.
But the thing that I see is quite interesting, is relevant to the workshop that I'm going to be giving, is these agent SDKs, right?
Started out with just helping us to write the loop.
But now these SDKs have been expanded.
to provide harness engineering capabilities, including memory management for contacts, including code execution, including some of the out-of-the-box tools for doing interactable local file systems.
So that's another trend.
So you can see that these agent SDK from lane change or anthropic or open AI, they all move in that direction.
of adding more capabilities as part of the harness engineering.
As anyone building agents moving forward, that's something to look out and leverage.
And there's a whole other trends of managing these agents, right?
So these companies are providing a runtime environment for you to host your agent.
Whether you want to do it or not, that's a question, but that's what they're seeing.
Are we able to build?
Are we able to build it well?
And how do we operate and run it at scale?
I think there is an architectural shift happening underneath all of this.
So for agents to become useful inside enterprises, they need to access tools, data, APIs, talk to other agents.
So all the protocols like MCP or A2A are going to become more and more different.
So I think the question would be whether agent protocols become a new enterprise integration layer.
So, for example, today we connect applications to APIs.
Tomorrow we may increasingly have agent discovery capabilities, negotiating tasks, coordinating work on top of those APIs.
So I think all of this architectural layer is also changing on the way that agents interact with each other.
Anybody else have any thoughts on quickly how all these agenting AI systems are changing the team organizations, team hiring and everything?
with AI systems these days is impractical because it's very expensive.
I mean, people are more and more qualified.
So people that are qualified to use AI, they carry that on their price tag.
Let's put it that way.
So you have to use it anyway.
So once you get there, how do you hire people to fit your own AI-assisted workflow?
I think that's the biggest question.
Again, we go back to the point that we don't have real standards yet.
We're experimenting in companies with different companies, different size.
It's very different to work on an AI-focused company and a bank or a repaid supplier that has its own IT department using AI for something.
One common point that I've seen is people are focusing on how well a new hire.
can actually describe or structure the system verbally because we go back, I don't like this, but we go back to prompt engineering, right?
You have to have a full understanding of what you're doing because most people are not really in the code anymore.
I mean, the AI generates all this boilerplate and it's very hard to do a code review these days the way you're supposed to do a few years back.
So with the lack...
of a process or tools to deal with that.
People are sticking to what they know or they perceive which is valuable at that time.
And the one common point is that new hires that can actually specify the system and get something useful out of AI, something that works within their workflow.
And then you have the variants.
Okay, so if I use a workflow that's based on, I don't know, I have one API model or I have my own.
workflow, pipeline, and then have the specifics.
But the common point is that we are kind of skipping the junior level and going to the architectural design level because for a new hire to be able to really specify a system, they need a certain baggage of knowledge, you know, so it kind of generates the problem.
How are you going to train your new hires if you're not really hiring junior people, which is part of the best we're seeing these days.
It's a big discussion.
I do see one thing that's changing as part of the interviewing process, where companies are now asking candidates to demonstrate their competency in terms of using AI tools.
So it's one of the new models.
Not all companies do this, but I wouldn't be surprised if more and more companies would incorporate this in the future.
For example, in my experience, right now we stop.
Hiring junior developers.
I mean, junior developers are no longer needed because of the also the front end.
Front end, we stop hiring front end developers.
Front end is no longer needed.
So what we're doing right now with the developer that we're hiring is that, as he said, we look for people who knows how to use AI tools, who know how to implement all the AI into the workflow.
And the developers are becoming more orchestrators, agents orchestrators, rather than see the code and the daily things that are happening.
So, for example, in our case, we developed a lot of agents for our platform to do a lot of things.
So the developers, what they do is to make sure that those agents work fine, that we don't have any, I don't know, data lake or that we control the expenses of the coding.
We control all these things that we can control.
But the developer, the code, the source code, all that stuff is completely out of the way.
of our equation right now.
And I think that's where the industry is moving.
It's not that developers are going to disappear.
It's that the developer role is going to change.
Mainly, it's going to become an orchestrator of all those agents and make sure that the agents work as they're supposed to work.
Thanks, Daniel.
Also, Hian, I just want to kind of quickly cover the self-learning and self-improving agents that you mentioned in the offline conversation.
So can you share any thoughts on that?
That sounds like a very fascinating topic.
Yeah, obviously this topic is very interesting and new at the same time.
I think it's a progression, right?
We know how to build agents, but how do we fine-tuning?
How do we make it better?
How do we make it improve over time?
There's a lot of research out there in terms of techniques on how to build these agents that they can self-improve over time.
The startups are building products where you can give them agents and they will help fine-tune your agents.
So it's a very natural trend.
I think we're still in very early days of that.
But the idea is pretty fascinating to me because if at the end of the day, if you're writing an agent, most of your logic is going to be in the prompts, in the system prompts.
Now you have tools and whatnot, but the core kind of logic of your agents is in the prompt.
And now we learn doing evaluations and tracing becoming very, very valuable.
as part of the agent and how to iterate.
So I think these bits are coming together of one day, we just got to be expressing what we want and then the system can help fine tuning it and make the agent better over time.
But we still have an early day, but it's a very exciting space to watch.
Yeah, thanks.
And also you shared that link, the article from Amthropic that is very good.
Actually, we will include as part of the references.
So it goes from, it talks about a transformation from like person, computer, and then the AI solution to like, agent working with workers, automated, basically build an AI building itself, right?
So very interesting.
Tang-Hian brought up the point about self-improving for agents.
And I think the term is a little bit ambiguous.
What does self-improving mean?
On the one hand, you can just say, okay, it can tell the agent to never use Rust, for example, right?
And it remembers you never use Rust in the future.
Is that considered self-improving or does it involve more of like...
like a digital change?
Yeah, what does self-improvement usually mean in client-employment agents?
Yeah, good question.
One example, if you think about Hermes, as a personal assistant that has that capability to improve over time, wants to know you better, knows how you work and stuff like that.
It's one form of self-improvement, but that's not the only form.
But that's an example of how It's like a human.
Once they're working with you more and more, they know what you want.
They want the way their style is.
So therefore, they can be a lot more effective at helping you, your tasks or what you want to do.
Right here, it could include agents capable enough to build and train models themselves with little or no human involvement, right?
So it's the whole life cycle.
The whole process is controlled by agents.
That's another form of self-improvement.
So there's many forms.
Something I would add here is that larger context windows don't eliminate the context problem.
So I think agents still need to decide what information is relevant, what can be discovered, what should persist as memory, what information is authoritative.
So I think context engineering is becoming almost like a memory management for this AI system.
So giving an agent more context is not necessarily better if that context is noisy or is stayed or is contradictory.
Okay.
Yeah, let's kind of switch gears slightly.
And Arthur, you mentioned about you are working on multi-agent systems.
I mean, definitely that's another very interesting area.
So whatever you can share with the audience, you know, can you tell us, you know, how these multi-agent systems work?
What are some good use cases?
Actually, that's a bit of a sensible topic because when you think about multi-agent systems as a topic, it's been there for a few decades already, right?
Before agents.
People from the academic community are seeing whatever comes with LLM agents and after transformers.
And it's related to the third multi-agent systems, kind of trying to reinvent things that were there in the past.
So my research area is on classical multi-agent systems, so before LLM, which doesn't mean that you can't incorporate or use LLM systems as part of it.
But these days I've been working on belief revision.
So this is a big part.
of solving two problems actually in the AI landscape.
One related to hallucination and the second one related to information accuracy and provenance after being processed by the agent.
It's kind of between computer science and philosophy, a lot of formal logic, but yeah, it's really interesting.
One thing that we've been working with, I think, beginning of this year or the end of last year.
Formal logic-based profile definition that could be used.
It's kind of similar that we can use with the concepts around JAD that we were discussing before.
So to have a search input, a search output, and how we can make the AI, not the agent stuff.
I was reframed from using agent, but actually working from here to there.
Thanks, Arthur.
A quick one back to you here.
You mentioned about spec-driven development and software factories.
And I've been seeing this in a couple of companies.
They're doing this.
They're creating.
dedicated teams to kind of create this software factory model.
Can you share more about that, you know, spec-driven development, how it works, especially for those who may be new to it, and what's happening in this space?
This probably started maybe four or five years ago.
And in terms of like with the coding agent now is, you know, pretty powerful.
But as you build a project, typically you want to have a plan in advance in terms of what you want, the capabilities and whatnot.
and not jump to the coding agent immediately to build.
So it is kind of borrowing what we know is a good practice to do, but now codified it in a form for the coding agent that you work with.
So if you build a brand new, like a Greenfield project, for example, you build a spec of what is it about, what are the kind of capabilities that you want, and then you kind of have a brainstorming session with your coding agent.
to provide a context that it needs before you actually go and build it.
I think there's a lot of good practices there that people have built around this.
And I think most people recognize that's a good practice.
I don't hear much more about it now because people are actually practicing it.
And the AI factory is another that's probably in the last few months.
of really how to infuse AI into the various steps of your software development lifecycle.
The typical use case is, hey, somebody raised a bug.
How do you get AI to take that from beginning to end of actually take the PR, do an analysis, trying to recreate a problem, come up with a fix, submit a PR.
And then you have going through the whole rest of the steps of review and testing and so on.
That is, as we all have too many things to do, too many tasks.
And I think if we can help automate that, that will accelerate another dimension to accelerating the software development lifecycle.
The AI agents fix the bugs they created in the first place.
I want you to ask this question about multi-agent system.
When you say you use multi-agent system, which pattern do you go for?
So, for example, is this one orchestrator?
Is it spin-up multiple workers?
Or do you have different agents working in parallel?
And if that's the case, then how do you deal with communications across different agents working on the same complex task?
Basically, the formal definition goes back to agents working together.
in a social capacity in terms of their interaction with each other to solve a common problem.
And in that sense, they can have common goals, shared goals, distributed goals, and they can use orchestrators, they can use different architectural patterns.
So for multi-agent systems per se, the definition is very broad in terms of how it can appear.
But in practice, what we see is mainly related to orchestration, orchestration and communication.
So since they have to work together, And they have to do that in a social manner, which means communicating.
So FIPA standards and things like that.
Everything kind of revolves around orchestration communication.
So I have different protocols for communication, different architectures for orchestration.
A lot of work on just these two.
Yeah.
So I'm very interested in agent-to-agent communications, right?
So actually, recently I use multi-agent a lot.
So I have some complex tasks.
I would get a bunch of agents to work together.
So I actually built this protocol to let them, like, okay, when you start going on a task, it's a part of big systems.
So, like, as an agent working on the same task, so we can message them to make sure that you guys go, like, you can coordinate and, like, don't overlap.
So what happened after that is that, like, my task didn't get done at all.
I kept getting a lot of messages, and it turned out that one agent started, like, forwarding tasks to the other agent.
And then that agent forwarded that to another agent, and then none of them did the work.
And I'm spending like half an hour just like checking, okay, where does this go?
We end up doing it and nobody did it.
So I feel like communication is one issue, one thing, but like for agent to realize, okay, this task is relevant in my lane and I should do it.
Versus like, okay, it's just going to form it out to a bunch of other agents.
So you need to be aware of what other agents are doing and the scope of that.
How do you manage that?
You try, you don't do it, you know, you try to do that.
Because the thing is, depending on how you're tackling the problem, for example, one of the simplest ways of doing that is using a blackboard kind of messaging to see what everyone's doing, so they're just posting back and have the shared, let's say, context for that.
But what if you have two multi-agent systems talking to each other?
The problem starts that you have different IDs, you can have different ID schemas for agents.
And you have different vocabulary.
You have different definitions for the tasks.
You don't have a shared state.
You don't have shared states, period.
And then you have research.
You have a lot of things.
So for the past decades, when you see all the work that was produced, they're trying to tackle specific problems in specific setups.
So, for example, how do you tackle the problem of one multi-agent, so have shared environment, share everything, trying to...
You have to delegate a task that you can't do for some reason.
So an agent started doing something and for some reason wasn't able to keep doing that.
You have to find an agent that's willing to do that.
Just for that, you need to set up.
Are agents benevolent?
Do you have to put a premium on that?
No?
Let's take the happy path.
All of them are benevolent, are trying to help.
So here comes the big question.
How do you define...
the equivalence of a task to achieving a goal.
So I want you to get from point A to point B.
Okay, so getting there by car or getting there by bus, it should be the same or not, the half-time restriction.
One thing that you'd find when you go deep and deep and deeper on that is that if you don't have a very specific shared setup, a very specific shared conditions, the difficulty of the problem increases exponentially.
So one of the problems of the multi-agency systems area that we have and we know that we have is that we have a lack of standards.
So we have standards for communicating, for example, have messaging standards.
We do have some protocols in place, but you don't have a standard architecture or a standard state description or a standard environment description for that sense.
That's part of the problem.
You have two multi-agent systems with two different setups, two different structures, data structures, descriptions, whatever.
And then you have to start tackling the particulates.
So how do you do that?
And then going back to your problem, how do you solve that?
You have to start by compartmentalizing your problem.
Like, okay, this is my box.
Okay, so I'm trying to solve this problem in this very specific box, in these very specific conditions and setup and descriptions.
And good luck from here.
As long as you have an alien agent coming up or a different multi-agent system, so you have different clocks.
So start over again.
That's the usual path.
Not sure to actually answer your question, but it is what it is.
No, it's a very hard problem.
I'm wondering if any of you, like, as we love to hear from Ian and Daniel, like, have you used the box orchestrations ability of different models?
First of all, like for in CLOT or in codecs, where you can ask an agent, okay, spin up my workers to do this task to make things faster.
Have you ever used it?
And do you have any impressions on them?
I haven't done that extensively.
Maybe Daniel.
No, what I do is I work with them like an ecosystem of specialized agents.
So, for example, you could have one agent which is responsible for coding.
We can have another which is responsible for testing.
We can have another which is responsible for security, another for deployment.
And perhaps you can have a supervisor agent coordinating all of them.
So at that point, all of the challenges start looking like a familiar to a distributed systems engineer.
So you have an identity, you have permissions, you have observability, you have orchestration, you have failure recovery, and that communication between the components, between all those agents.
So that's why I think agent engineering will increase becoming more like a system engineering discipline rather than just prompt engineering or orchestrate those agents individually.
Yeah.
So when you mention that, maybe think about like the sandbox system.
One thing I did see from the trend from this year and last year is a lot of sandbox companies, right?
And I've been in sandbox for Asians.
So do you own like using sandboxes for Asians?
For example, each Asian has to work in an isolated work tree or something like that, or is it dependent on the task?
Yes, for me, it's mainly dependent on the task.
So something I've been wondering about.
So we have compute architecture, right?
So AWS, we have GCP, we have Azure.
And usually a company is a shop, right?
Hey, I'm a GCP shop.
So one of my things is run out on Google Cloud or something like that.
I wonder if the same thing happens to AI?
Because a lot of companies have subscriptions or have a contract with multiple providers.
Do you see a future where like...
people would just stay with one provider?
Maybe OpenAI would have owners this model?
Or do you think that people would have a multi-provider system?
Yes.
In my opinion, I think it's going to happen the same that is happening with cloud computing.
I mean, you are not tied to just one cloud provider.
You work with multi-cloud providers.
So with AI, it's going to happen the same.
You are going to be tied to multi-AI providers working on different things of your workflow, different tasks.
But I don't see just like being...
attached to only one provider.
Like you said, for example, I work with AWS, but I have multi-cloud providers working on different states of my workflow with GCP and ASIO.
So with AI, I think it's going to happen the same.
You can say, okay, I prefer Antrofic, but in the end, we'll end up working with OpenAI, with Grok, with any other providers.
So that's going to happen.
Thank you.
Yeah, Lex, I would like to touch a couple more topics and we can wrap up.
It's been a great discussion, but I still have a couple more topics.
I want to make sure that we share our thoughts with the audience.
The first one is the evaluation frameworks.
So definitely this has been a big topic this year.
I even heard in one of the conference presentations that 2026 is the year of AI evaluations.
So there are a lot of different ways of doing this.
There are harnesses and the LLMs being used as judges, LLMs as judge, right?
So how do you measure AI?
effectiveness, agentic effectiveness in general?
Like what kind of metrics do you see in your own experience?
Like how do we make sure that the end-to-end process is being, you know, has checks and balances?
Maybe here and starters off on that, like what are you seeing in terms of agentic evaluation space?
This is a very important topic for companies that are bringing their AI agents to production, particularly the one that for their customers, not so much about internal ones.
If you look at example from DoorDash, they build an AI systems for ordering food.
You have companies, other companies that are building for customer support agents for external users.
And we know, you know, we all know that LMs are not deterministic and building agents, they can, they can take the wrong turn.
They can make the wrong tool calls and stuff like that.
So, evaluations is kind of a way to help us to make sure that the agents do the right thing for the right tasks at the highest level.
And without evaluation, how would you know that it's gone rogue or do something that's unexpected, not according to your company's policies and all that stuff like that?
And there's like methodology behind that.
And we have going to, by the way, we're going to have a session on that at QCon SF.
One of the experts on evaluation is going to be there.
And companies not only like have to do this, but they have to incorporate this as one of their core steps in building AI agents.
And a lot of companies ending up building something called evaluation harness, which helped drive the whole evaluation flow from getting the data.
running the evaluation or tests, as we all know tests, right?
And then assessing the quality at the end.
And that helps drive the decision whether it's a go or no go for release.
So anytime that you're doing serious customer-facing agent, you have to incorporate and spend effort and time and effort of doing evaluation right.
If you build an internal agent, you're probably not going to spend a whole lot time on it.
But if you're building a real serious agent, for your company, then definitely a must.
And you see many examples across companies from LinkedIn, DoorDash, Ubers and such.
And there are many cases study on that.
Thanks, India.
Definitely.
I think this area is going to see like more it all comes together, right?
So any other analysts, any other thoughts from anybody else?
How do you kind of evaluate your own AI systems?
Yes, I think I completely agree that.
with the idea that this year is becoming the year of evaluations for models and agents.
So once an agent can perform an action, evaluating only the model is not enough.
So you have to evaluate the whole system.
You have to evaluate the model, the tool, the context, the agent loop, the environment.
So here is where the metrics change as well.
So it's not only whether the answer was correct.
So it's whether the agent completed the task, how much cost, how long it took, how many interventions were necessary.
and whether it did something unsafe along the way.
So I think all these evaluations were just starting to see how are we going to evaluate all this workflow regarding agents and regarding LLMs and regarding the interactions that they do in our job.
Yeah, good point.
All those metrics are more relevant now, right, with the overall.
Yeah, one more topic I definitely want to highlight in this discussion is the physical AI.
Daniel, you've been looking at this for a while.
We talked about this last year as well.
How do we bring the, intelligence into the real world, right?
So if AI is into the software system, how do we bring that into the hardware, whether they're like robots or what now they're being called collaborative robots, cobots, right?
And then auto, even like self-driving cars.
Daniel, maybe you can start us off on this.
Like, what do you see here?
What's happening?
And what we should be paying attention to?
Basically, it's saying, how do we bring AI into the physical world, no?
So over the last few years, we have spent the time putting intelligence into the software, now we're starting to connect that intelligence to machines that can perceive and act in a physical world.
So what you said, we are starting to see robotics, we're starting to see autonomous vehicles, drones, and industrial systems are now part of this transition.
What I think makes physical AI fundamentally different from software agents is that mistakes have physical consequences, no?
The physical AI.
So one thing is that in software, an agent can make a mistake generating their own file, but a robot can move their own equator.
So it's going to be a whole different level of mistake.
So the reliability, simulations, safety, verification, evaluations are going to become more and more important in the physical world.
So based on that, all the foundation models for robots that are happening right now, another important shift.
is that robotics are starting to benefit from more of the same foundational model approach that transform the language and vision.
So instead of programming every individual behavior manually, we're increasingly seeing models that combine vision, language, perception, actions.
So is that the robot can potentially understand instructions like, I don't know, pick up that object and put that on the table and translate that into a high-level intent of a physical world.
So as you mentioned, there are a lot of projects regarding the physical model, like NVIDIA, like Google Robots, where all these multimodal models are taking language and visual information and producing action for robots.
So that connection between language, perception, actions is what makes the space of robotics and physical AI more interesting.
So we're going to start seeing things like simulation for digital twins robots, all the things that are going to happen for open-source robotics.
all the things that are going to happen from AI agents to physical agents, all of this is just going to start moving faster and faster over the next year because they, I mean, right now the AI is here.
Now we have to put it in our physical world.
Yeah, that's going to be definitely one of the applications that's going to see a lot more attention, a lot more developments, right?
Whether it's manufacturing or any other physical industrial use cases.
Any other thoughts?
robotic robots and vision language models?
How do you see them?
Yeah, I think it's very, very exciting.
Definitely a lot of people say that it's the next frontier to get the models into the real world.
And we see a lot of progress.
And I think the only challenge is that there are several challenges.
So people keep saying that robotics is like cars, right?
Like LM to say, okay, we know we already have a lot of robots in the real world.
We have Wemos everywhere.
We have jokes.
I'm not sure how to look this out.
I have never driven one, but I see the...
I see the advertising for it.
So we see a lot of robots in the real world.
So people keep telling me that, oh, we know how to train massive LOMs, multi-model stuff.
We know how to train robots, like serving cars.
We should know how to train robots to work with real world.
But I don't think it's quite the same.
So one big difference between robotics models and LOMs is that the challenge of backtracking, right?
For LOMs, when you train things on a virtual, like on the computer, if it makes a mistake, you can run back, right?
You can reset the state.
and continue so you can do a lot of like explorations and pick the best state you can't really do that with robotics right like you can't get a robot okay there's a squawk now that's there the robot fell falls and it breaks you can't just roll it back and get the same robot as before so it makes exploration extremely extremely hard and to set another challenge in data and so it was been looking into a lot of like robotic setups so a lot of them have raised a lot of money uh so here's my favorite of these companies raised like what like a billion dollars i don't remember off the top of my head I think like Yanli Kun is a company.
We have a bunch of like general, like company on the general list of physical intelligence, Sunday, skilled AI, which I'm like very excited about.
If you look into it, I don't like looking into it.
And the main difference is the approach on like to data.
So I think that people are still betting on like what kind of data would help robot models to scale.
Some something tricky with AI models is that like sometimes when it's small, they don't quite work.
a model only work really well when they're big enough, right?
For a long time, we thought the language models were like, they're generally trash, and they suddenly boom.
Like, you have GD3 and GD4, and it's like, it's suddenly everywhere.
So a lot of the bet is like, okay, the models are not working well yet.
Is this because the approach is wrong, or is it because we just do not have enough data yet?
So people have to make a bet on like, will this kind of like architecture, when this kind of training algorithm, when this kind of data be the right approach to scale?
So we see a lot of this company making a big bet.
So there's some companies that say, okay, we need real robotics data.
So in China, we have like a Chinese government is investing a lot of money into like have like robot training center, right?
We have like humans operating like real robots and just collect a lot of this data.
And it's extremely expensive.
So a lot of people were like, okay, we do need this kind of data.
Maybe you can just go with some kind of video.
Like if humans can learn from watching video.
can robots do it, right?
So it was collecting a lot of this in videos.
We have like people paying people in different countries, like just like wearing like the camera on their head and collect videos.
So it's very interesting.
But since a lot of companies go into middle ground, like physical intelligence, like I think it's generalist, like on Sundays, they have some kinds of clubs.
So they were like, okay, like videos are not put in for like haptic information, right?
Because a lot of time you can't judge from a video, like how much force it should put into like squeezing an egg.
or like a strawberry.
So you need to have some kind of gloves, like people have wear them and I charge with my manipulate object and get data to train the robot.
So it's kind of like the middle way because you don't have the phone robots.
It's a lot cheaper, but it's also not just videos.
But anyway, I think it's a very exciting space.
Thanks, Chip.
Yeah, definitely we are kind of running.
shorter on the time remaining in this meeting.
So I know there are a lot of topics.
We have talked about the topics that at least that we think are definitely important to be aware of.
But I know there are other topics that we can probably, we won't have time today, but I just want to quickly mention, you can follow up offline as separate podcasts or articles.
So there's AI governance platforms that are coming up and then there is AI security and regulations is another important topic.
AI observability is an evolving topic as well.
And definitely the AI SRE and AIOps are also getting a lot of attention.
So do you see anything else happening at a very high level before we wrap up?
Arthur, do you have any comments?
Yeah, I think we could regret relating multi-agent systems and AI engineering because it's not as hopeless as that might have seen when it's on a ship, you know?
But there's a lot of things that we could bind together and bring together to correlate to how we solve problems that we have using work that has already been developed.
So that would be a great topic.
We can start wrapping up.
So this is kind of my favorite part of the podcast to request each panelist to make a prediction for next year.
And we'll take a checkpoint next year and see how these predictions have kind of been realized.
So yeah, we can start off with Chip.
What do you see?
What's your prediction if there is one thing?
you'd like to see happen or you think it's going to happen in the AI space, what would it be?
I really want to see in US robot companies that have robots ready to be bought.
I feel like, I know it's just like very painful.
I feel like in China, you can just buy a robot in the US.
You just cannot get from an American company yet.
So if we don't want to see the shipping, I feel like we'll be waiting for OpenAI and Antropic IPOs for a while.
I think they delay.
Very curious to see like how...
that's going to impact the market, right?
Because like, if the IPOs go well, the market can go well.
But if the IPOs don't go well, what implications would it have for the funding and like the AI market?
I wouldn't expect to see a lot of like cheaper and faster models.
And there's a type safe, type startup model.
It would be very interesting.
Thank you.
How about you, Daniel?
What's your prediction?
I think we're going to start seeing limitations on AI.
So for example, the first is going to be the energy.
So all the AI infrastructure is becoming extremely power hungry.
So I think the access to electricity, capacity, cooling data center locations, that is going to become a biggest constraint for the growth of AI.
So we may start seeing like an energy crisis for AI.
The second prediction that I would like to share is like a futuristic prediction, but I think it's going to happen that all this AI infrastructure is going to start moving beyond the AI.
the traditional data center and it's going to move into the space.
So I think putting AI into the space is going to be more efficiently.
Probably it's going to take a little longer, but it's going to be more efficiently than keeping the AI heat in the earth.
And finally, I think the AI competition between United States and China is going to increasingly look like a technological Cold War.
So it's not about who has the best model anymore.
who has the best chips.
So it has to be with who has the manufacturer, who has the energy, who has the data center, who has the open models, the supply chains, and the access to compute.
So we are already seeing that semiconductor access and domestic ship productions are becoming strategic issues in that relationship.
So I think that for 2027, some of the biggest AI stories may actually happen outside the model themselves.
And it's going to be more about the energy, the chips, the infrastructure, and the geopolitics.
And Ian?
Just have two things.
Every AI agent will incorporate some form of the Jeff model into the AI agents.
And more and more of the communications going to be between agent and agents moving forward as we figure out how to build, operate, and bring all these agents to productions.
And Arthur?
Let's go back to what I said at the beginning.
I think and I expect that the most successful AI deployments that we'll see will be much narrower than what we see right now in terms of rather than one agent with access to everything, we'll see a lot of things using smaller specialized agents for mixed models, for well-defined tasks.
And the big difference will be in the surrounding system.
So you have trusted data, leave the permissions, tasks, a way to see, and there's not back to your visibility, but a way to see what the agent did and whether it actually helped.
Teams can not only do that, but can actually measure the cost and quality of completed tasks.
You can have ROI for the task, not just the number of tokens of agents that use it.
I think that this will be a great leap.
It aligns well with what he said on Jad and what Daniel said on the...
shifting away from big models.
But I think that's where I put my chips in.
Thanks, Arthur.
Yeah, I think I'll take a quick one minute to kind of wrap up my comments.
Yeah, AI is a significant technology shift.
And as we all know by now, after almost four years, it is not a fad or some temporary blip on the technology radar.
It is here to stay.
It is one of the major technology waves in the last, I don't know, 20 years, right?
So my advice and request to our audience is We should take advantage of this technology transformation, not be afraid of it.
Find ways to use AI tools to automate any workflows in your day-to-day life, whether in the office or in our personal life.
This is a tool that we can use to find more time to do the important things in our lives.
Let the agents and the tools do the routine tasks.
So my prediction for next year is definitely I agree with all of you.
AI will be even a bigger part of the daily lives.
And I always say this as a prediction, like, and I would like to see whenever we can, I don't talk about AI anymore, it has become part of the fabric of our lives.
I think then we, we can say we have really arrived in.
So I think there'll be more evolution, more development, but I think it's going to the niche areas will be settled, you know, where AI will be really valuable and, and where the human intelligence is still required.
Right.
So, so thank you all for your time.
We took a little bit longer time, but there are so many topics and we can go on forever, but.
I think we need to wrap this up and we will follow up with any other topics later.
So thank you for joining today.
Have a great one.
Until next time.
Thank you.