Just for fun and educational purpose: this is a simple MCP server in shell that helps understanding how it works under the hood without third party libraries.

Just a toy, and very incomplete, but hopefully useful.

Awesome.

I wanted to try to implement this for months. You did a really good job.

Thank you! Still a WIP, but a very fun learning / inspiration project. Got a bit of Rust jammed in there, bit of device constraint dancing, bit of multiple LLM api normalization, bit of spacial vision LLM education, etc.

At some point I wanted to turn goMarkableStream into a MCP server (model context protocol). I could get the screen, but without “hack” I couldn’t write the response back.

The trick here is to inject events as if they came from the user. The virtual-keyboard works really reliably, you can see it over at https://github.com/awwaiid/ghostwriter/blob/main/src/keyboar... . It is the equivalent of plugging in the reMarkable type-folio.

Main limitation is that the reMarkable drawing app is very very minimal, it doesn't let you place text in arbitrary screen locations and is instead sort of a weird overlay text area spanning the entire screen.

I wrote those articles without AI (well a bit of AI to fix grammar and syntax). Even the schema are handmade :) Hope it is useful.

Second part: https://blog.owulveryck.info/2025/01/27/mcp-part-ii-implemen...

It is a three part article from the concepts of Agentic, MCP to the implementation of a server and the implementation of a host.

The Chief Data Officer should expose what it means for a complant to be « data-driven »

This article presents a model for activating a data-driven value flywheel within organizations, emphasizing the interplay between business and technology through a structured, four-phase approach. However, as with any framework, its utility and effectiveness depend on real-world application and feedback.

I welcome the Hacker News community to share their insights and experiences with this model. Do you find it useful in your work? What aspects resonate most with you, and what improvements or alternatives would you suggest?

I wanted to play with data contracts to understand what value they can bring. It ended in a tiny data mesh implementation that fits in a blog post.

I wanted to test the bitol standard. So I invented a use case based on a RAG. I put most of the code (CUE, Python and DuckDB) so you can try it yourself.

I hope that it will be useful.

I think that in Europe, that will soon be prohibited without declaration and explicit consent.

If that's true, it's pretty ridiculous. I wish legislation would get at the core issue, which is not processing, but that collection happens at all. Normalizing surveillance, but asking companies to please not do anything with that data for minors, is such a far cry from what we deserve.

But Discord must store the messages, to make searching possible. And I assume the model works based on the writing style, topics discussed, etc. What would you propose, automatically remove all messages after a few weeks? Even then Discord technically can process the messages and build the same model. So yeah, in this case the processing is the problem since I don't want Discord to know my gender and age (unless I explicitly tell it).

> What would you propose, automatically remove all messages after a few weeks?

This but unironically.

If people or bots or the NSA want to keep records, let them scrape and cache 'em.

I actually think it should all be P2P, but that would kill the thin shell of a business model that exists.

I would hate such a model, having persistent history is one of the killer feature of Discord vs, say, IRC.

I dunno, the search on Discord is trash. The one time I thought it would be useful to go back and find something, it was completely incapable.

This is not my experience, I use the search heavily and find it an invaluable tool to prevent asking repetitive questions.

At least under GDPR, collecting personal data is already considered processing. Processing is only allowed under specific conditions and purpose.

It would be perfectly fine if Discord stored messages from users (for brevity, assuming that's personal data), made it searchable and available as long as the user exists on the platform, provided that they specify this as the purpose of processing, and the user explicitly gave informed consent. Should Discord want to perform any other processing besides that designated purpose, they would (again) have to receive the user's explicit, informed consent. Otherwise, it's simply not allowed.

No need to remove messages in such a case. Of course, technically they could process that data for a different purpose than specified, but that's illegal and may eventually result in a fine that can be based on a percentage of total global revenue.

Not profit or loss, revenue. Then suddenly it becomes a really interesting business decision to do other than defined things with the data at hand.

(and something something purpose limitation, data minimization etc.)

If you suspected you had underage users using your service illicitly. Could you not use such methods to find them?

I wonder at what point analytics, queries, or ML to determine age gets in the way of actually keeping them from getting into things they shouldn't...

Under age users aren't going to just tell you their age honestly, so you'd have to use some method to derive their age.

> If you suspected you had underage users using your service illicitly. Could you not use such methods to find them?

You could use something like that and it would be counted as “legitimate interest” under GDPR, but:

- you wouldn't need (and then would be forbidden) to do segmentation between more categories than “too young”/“old enough”. This is not what they are doing here.

- you'd still need to tell the users that you're doing it, and provide a way for the users to modify that data upon request

> Under age users aren't going to just tell you their age honestly, so you'd have to use some method to derive their age.

You don't have a legitimate interest of knowing their age, only for knowing if they are underage or not.

Under the DMA's requirements to protect children I think it will be basically mandatory to do this.

In the UK it already is under the Online Safety Bill.

Diving into RAG (Retrieve, Augment, Generate) systems? Check out my article where I explore the engineering aspects of creating a RAG from scratch—it's not another overview article nor is it about understanding Langchain. Instead, it's an attempt to pinpoint where exactly the engineering ideas lie in the process. Learn the challenges and insights from my experience!