MCP has been all the hype for the last 1.5 years. Kind of, at least. But you can safely ignore it as a developer. Agent Skills are the way better alternative.
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
As a developer, using AI efficiently can be quite got a bit easier because you can safely ignore MCP, the Model Context Protocol, now, and I'll explain why. The Model Context Protocol was already introduced in 2024 behind it sounds pretty good. The idea is that large language models on their own are dumb. They're just producing tokens, always will. So they're producing tokens. They don't know anything. And the reason why ChatGPT, for example, now much better than it was when it was initially released is not only
because models got smarter or better but mostly because ChatGPT now has tools included, like web search or the capability to execute code. And the idea behind MCP simply was to make it easier to expose new tools to large language models. Sounds like it makes sense, right? Yeah, sounds like it does but it doesn't unfortunately because one is, or two big problems with MCP actually are that you can't give them a lot of tools. If you start adding, like dozens of MCP servers to your
AI application, chances are that it will just not use any of these tools or use them in the wrong way because polluting that context with all that MCP information because all these MCP servers, of course, need to expose information about the tools they provide and about the resources they provide and that can quickly lead to a situation where you're making your AI application dumber instead of smarter. So that's one problem. Another problem is that you have to hope that the AI uses certain tools or you have to kind of tell it, "Hey, please use that MCP
server to get a certain task done," which, eh, works but is not really ideal and then there's another maybe even bigger problem when it comes to MCP and development, on here. As a developer when it comes to tools like Cursor there only so many things you want your AI to do. That might be spinning up a browser looking at the website. That might be querying a database to see if some data has been inserted. That might be to reach out to some
documentation. Maybe there also is some Git-related stuff and that's it then. Now the bad news for the MCP fans are you don't need MCP for any of that. Cursor and all these other AI coding tools already have web search built in or the capability to send a request to some website and fetch the content to read it so they can already visit websites. No MCP needed for that. They already have browser use implemented. Cursor has at least. It can spin up a browser and view it and work with it and
navigate it for you. They can use the command line interact with Git which allows them to interact with a can do all these things already and that's the reason why I don't really use any MCPs and never felt like anything's missing except for one thing. One problem I do have as a developer is providing the right context or it's not a problem but it's the most difficult and important thing. For example, I know that there are a lot of modern browser APIs and CSS features I would like to use in my
project. I don't want to write CSS code like it's but that's what AI typically does as a default and I don't want that so I typically remind AI to please use relative colors, for example, or please use the pop-over API and that's, course, annoying but that's something where I this new Skills standard which is emerging and evolving. The idea behind Skills is that in your project you can add a Skill folder and right now where you add that
folder still kind of depends on which tool you're using. Cloud code, you would have a Skills folder in the and then for each skill you would add a subfolder in that Skills folder and add a skill.md file. That is the bare minimum. Now what goes into that skill.md file? In the end just some context, just a prompt you could say. This is simply how such a skill.md file could look like. You have some metadata which is just the name and description a- as a minimum and the description is the part that will help the AI decide
whether to use the skill because the idea behind these Skills is that the coding agent essentially loads all these names and descriptions into its context whenever you start a new chat and that's obviously just a few tokens so that's no problem. It's not polluting your context but it has all that information. It has a system prompt, of course, that makes sure that this is prominently embedded and that it considers all these skills whenever tackling a task and it would then
take a look at the entire file, open that entire file but only then and then it therefore gets the rest of this content which is just an extra prompt, just some extra context that's loaded on demand and that's the important part. Now a Skill can be more than that. You can add subfolders to each skill to even add scripts so your own code that should be executable so that you can add your own tools in the end connected to a skill that can be used so for example you could add code that opens and reads a PDF document so that you have your own parsing logic in
place. You can add extra references like extra you could therefore, for example, build a skill that helps the AI write better React components by making it aware of the fact that it might not need useEffect because AIs love to use useEffect but they might not need to. There is a full article on the official docs that you might not need an effect. You can put that into your references, skill related to that, maybe also add some other React component related information in there and then whenever you ask the AI
to do something that's related to a React component there's a high chance of it loading that skill and using that extra knowledge and context so that you don't constantly have to repeat yourself or hope that the AI does things the way you want it to do and that's why I find this Skill spec quite interesting and why I'm using Skills in my projects to make AI aware of all the modern syntax I'd like it to use and why I personally don't find MCP super interesting when it comes to development with AI.
GPT 6 Astra was released. It's a good model, I have to say. Especially for 3D work it seems to be a real game changer. But due to all the hype, you may be disappointed...