Oct 14, 2026 · 23 min · 10 segments
Jacob Beningo explains why agentic firmware development is already here and how embedded teams can move beyond ad hoc AI prompting. He covers skills, agents, RAG databases, and MCP servers, shows how…
Jacob BeningoHost
And what I mean by that is that they are good at opening up Copilot, Cloud Code, or ChatGPT and having a conversation with it back and forth to build or get something up and running, solve a problem, debug their system, or whatever it happens to be that they're trying to accomplish.

And as I mentioned, this is where you start to leverage skills, agents, RAG, and MCPs so that you can repeatedly get agents your agent orchestrator or whatever LLM it is that you're using to reliably develop your system essentially over and over again.

I don't want to necessarily say over and over like it's deterministically creating code, but the idea here is that you might have a certain coding style that you use within your company.


And then you generate an agent that has the skills and access to the RAG database so that it can actually go through and do a code review for you.

And then every code review that it does, it will know what to do every time you say, hey, can I go and have you review this code for me? Okay.



We hear about people who are developing, you know, stuff online, who are doing web-based stuff, who are, you know, really steeped deep into artificial intelligence.

There certainly are some, you know, I'll kind of use myself as an example, someone who's been exploring how do we take this intersection between AI and the process or the industry that I am involved in the most, which is the embedded systems industry.

How do we leverage AI to build more robust embedded systems? How do we make that scalable? How do we help accelerate the way that we develop embedded systems? I'm not necessarily a fan of trying to provide AI with a spec for what I want to build and click a button and have it produce the whole widget out the other side.

I view AI as a tool that every embedded developer can use to make their systems more robust, more repeatable, accelerate the development cycle, and deliver something a lot faster for a lot less cost.

Ultimately, what I have seen in my 20 or 30 years of developing embed systems, and today as I go out and help teams, nobody has enough engineers to be able to develop a product, right? They all are short on budget, time to market, resources.


It doesn't necessarily have to take away the fun parts of our jobs, the stuff that we'd like to do on a day-to-day basis, but it can do, I think, the stuff that maybe we find to be boring, that we don't want to do, that then frees us up to focus on actually innovating and creating new features that our customers want versus maybe all of the low-level infrastructure that really takes a lot of time to get in place and get it right that prevents us from doing what we're best at.

And what I mean by that is that they are good at opening up Copilot, Cloud Code, or ChatGPT and having a conversation with it back and forth to build or get something up and running, solve a problem, debug their system, or whatever it happens to be that they're trying to accomplish.

And as I mentioned, this is where you start to leverage skills, agents, RAG, and MCPs so that you can repeatedly get agents your agent orchestrator or whatever LLM it is that you're using to reliably develop your system essentially over and over again.

I don't want to necessarily say over and over like it's deterministically creating code, but the idea here is that you might have a certain coding style that you use within your company.


And then you generate an agent that has the skills and access to the RAG database so that it can actually go through and do a code review for you.

And then every code review that it does, it will know what to do every time you say, hey, can I go and have you review this code for me? Okay.



We hear about people who are developing, you know, stuff online, who are doing web-based stuff, who are, you know, really steeped deep into artificial intelligence.

There certainly are some, you know, I'll kind of use myself as an example, someone who's been exploring how do we take this intersection between AI and the process or the industry that I am involved in the most, which is the embedded systems industry.

How do we leverage AI to build more robust embedded systems? How do we make that scalable? How do we help accelerate the way that we develop embedded systems? I'm not necessarily a fan of trying to provide AI with a spec for what I want to build and click a button and have it produce the whole widget out the other side.

I view AI as a tool that every embedded developer can use to make their systems more robust, more repeatable, accelerate the development cycle, and deliver something a lot faster for a lot less cost.

Ultimately, what I have seen in my 20 or 30 years of developing embed systems, and today as I go out and help teams, nobody has enough engineers to be able to develop a product, right? They all are short on budget, time to market, resources.


It doesn't necessarily have to take away the fun parts of our jobs, the stuff that we'd like to do on a day-to-day basis, but it can do, I think, the stuff that maybe we find to be boring, that we don't want to do, that then frees us up to focus on actually innovating and creating new features that our customers want versus maybe all of the low-level infrastructure that really takes a lot of time to get in place and get it right that prevents us from doing what we're best at.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.