Robin Clowers
robinclowers.com
Robin Clowers
@robinclowers.com
Software engineer at Poggio.
I'll check it out, seems really cool! Thanks for responding
December 10, 2025 at 10:28 PM
I did read the post, it's not clear to me how useView avoids the probably I described. From the PostView example it looks like the post is not nullable, so it must suspend until the request completes, right? So if you have useView calls deeper in the tree, does that not create a waterfall?
December 10, 2025 at 10:10 PM
How do you avoid client waterfalls? The post mentions using a batch fetch, but if a component high in the tree suspends until it gets data, and it has children that also request data, you still end up with a waterfall, right?
December 10, 2025 at 4:17 PM
True, you don't need severless, but it's cheaper and easier to manage IMO
October 25, 2025 at 5:30 PM
Sounds like your goal is to learn by doing it yourself though, which totally makes sense
October 25, 2025 at 5:30 PM
Yeah, staying simple makes sense. This is why so many people host next on Vercel, they handle all of this for you. Also having preview builds on PRs is a game changer, especially working on a team. Though frankly even for my personal photo album app it's useful.
October 25, 2025 at 5:28 PM
How do you plan to handle deployment, some sort of blue/green? Or just a hard cutover?
October 25, 2025 at 5:11 PM
Severless gives you a lot of benefits though, like being able to run multiple versions of your RSCs at the same time, so clients continue to work without a hard refresh.
October 25, 2025 at 5:10 PM
I think Chris's point is that the LLM could just call the tool directly, rather than making a tool call through the MCP server
August 27, 2025 at 2:21 PM
How would you call a CLI tool over the network though? Having a standardized transport and auth mechanism is valuable
August 27, 2025 at 2:19 PM
We could have built a rest api and published a spec, but then you would have to manually set up the context with the LLM. For an engineer, that's easy enough, but for non-technical folks it's not straightforward.
August 26, 2025 at 5:29 PM
Right, but MCP provides a consistent interface for many different types of tools. Maybe a concrete example would help, I work at poggio.io, and we're working on an MCP server, which allows a non-technical user to add our product to whatever LLM product they happen to be using.
Poggio | Tailored POVs for your people and agents
Poggio is an AI-powered account intelligence platform that arms go-to-market teams with an exec-level point of view for every account.
poggio.io
August 26, 2025 at 5:28 PM
Sure, but MCP is designed specifically for LLMs, so it's opinionated. MCP supports stateful persistent connections and a rich response format. In theory these are all things you could do with an openapi spec, but having these things standardized makes it much easier to build
August 26, 2025 at 2:46 PM
MCP is for discoverability and interoperability, not human in the loop use cases.
August 26, 2025 at 3:09 AM
No idea what you mean by this, but go search "bboy" on YouTube and then tell me breakdancing went away.
July 27, 2025 at 1:51 AM
Came here to say this too, never seen raw jalapenos on a Seattle dog
May 25, 2025 at 9:25 PM
My wife convinced me to get rid of a bunch of cables a while back. A few weeks later I realized I needed a coax cable, but I had recycled them all 🥲
May 2, 2025 at 3:15 PM
They are called rules for a reason!
April 30, 2025 at 7:31 PM
Do you have a citation for the claim that no more are coming? That's pretty hard to believe. Clearly the tariffs are having a huge impact, but I'm sure some things will still be imported.
April 27, 2025 at 3:22 PM
Also just use a template for your resume! Then you don't even have to do any layout yourself.
March 11, 2025 at 2:05 PM