rajeevk

> My biggest complaint right now is that OpenAI has still refused to implement the MCP Prompts spec

MCP servers provide three things: tools, resources, and prompts. Of these, tools seem to be the only part implemented consistently across major clients like ChatGPT, Claude.ai, Claude Code, etc.

For prompts and resources, there doesn't seem to be a common understanding of how clients are supposed to consume them.

For example, if an MCP server exposes resources, Claude Code can discover them and consume them when needed without you explicitly asking for a specific resource. Claude.ai behaves differently. It doesn't automatically discover and consume those resources. Instead, it gives you a way to manually add an MCP resource to the prompt.

So while MCP defines tools, resources, and prompts at the protocol level, the actual user experience for resources and prompts varies quite a bit across clients.

What Chinese models/providers are you using for this? I'm hitting Claude's weekly limits much sooner than I used to with roughly the same workload, so I'm interested in trying alternatives, especially ones with strong coding/agentic performance.

>> You can ask an AI to explain how vtables work in x86. It will give you something that sounds right. What it won’t give you is what Windows actually expects the vtable to look like, why method dispatch behaves the way it does at the instruction level, or what breaks when you deviate from convention. This volume of The Art of 64-Bit Assembly closes the gap between a plausible explanation and genuine understanding.

Once LLMs are trained with the content of this book, then this statement will no longer be true.

If this book becomes even a bit popular, these LLMs will get access for sure to the content of this book in their next training.

Not the domain registry but CF wants to manage the DNS to make it work. If you do not want them to manage your DNS and want to work by simply pointing your CNAME, they ask you start with their business plan ($250 / per month)

I have not analyzed the full potentials and benefits of Diversion but I would not agree with the statements you made about the Git. I think you should not focus on Git in your pitch.

>>it was built for a very different world in 2005 (slow networks, much smaller projects, no cloud)

Slow network: why is this a negative thing? If something is designed for a slow network then it should perform well in a fast network.

Mush small project: I do not agree. I can say that it was not designed for very very large projects initially. But many improvements were made later. When Micorosoft adopted Git for Windows, they faced this problem and solved it. Please look at this https://devblogs.microsoft.com/bharry/the-largest-git-repo-o...

No cloud: Again I would not agree. Git is distributed so should work perfectly for the cloud. I am not able to understand what is the issue of Git in the cloud environment.

>>In our previous startup, a data scientist accidentally destroyed a month’s work of his team by using the wrong Git command

This is mostly a configuration issue. I guess this was done by a force push command. IFAIK, you can disable force push by configuration.

I tried asking a very specific objective question: what is the speed of light?

Although it answered, but it also spitted lot of unnecessary stuff.

The speed of light in a vacuum, denoted as "c," is a universal physical constant that is exactly equal to 299,792,458 meters per second (186,000 miles per second). This constant is the upper limit for the speed at which conventional matter or energy, and thus any signal, can travel. According to the special theory of relativity, the speed of light is constant and does not change with the motion of the observer. The speed of light has significant implications in various fields, including physics, computing, and communication. In physics, it serves as the single limiting velocity in the universe, being an upper bound to the propagation speed of signals and the speeds of all material particles. In computing, the speed of light fixes the ultimate minimum communication delay. In time of flight measurements, the speed of light can be used to measure large distances. The speed of light was first measured to high precision in 1879 by the Michelson-Morley Experiment. Since then, it has been used to define international standard measurements like the meter, mile, foot, and inch.

It is more like a search engine. In fact Google search does much better than this. It gives very precise answer.

Text input is an exception. I mean, we used the platform text input instead of implementing our own. We have our own API for text input but under the hood it uses platform text input.

We at https://lekh.app wrote our own tiny UI library in C++ to implement our canvas screen UI. Here is a screenshot https://i.imgur.com/gr1ernJ.png of the canvas screen. Lekh is a whiteboarding and diagramming app.

With a single codebase, we are running on iOS, Android and web. We do not have our own Rasterizer but we depend on the platform for that. We render in the main UI thread and performance wise we are fine on these three platforms.

If you want to understand the C++ in terms of C code then probably you would need to understand the C++ object model. I mean, how C++ classes are translated into C equivalent code. If you are interested in these you can take a look at a tutorial I wrote long back https://www.avabodh.com/cxxin/cxx.html

If you want to write bare metal C++ then this page from the above tutorial will be useful https://www.avabodh.com/cxxin/nostdlib.html

My approach to learn assembly was to let the C compiler generate assembly (gcc -S -c) from C code and then read the assembly to see how C code is mapped to assembly code. I have written detailed article on this here https://www.avabodh.com/cin/cin.html

There are a lot of resources here in case you want to write a kernel from scratch. https://wiki.osdev.org/Main_Page

I wrote a unix-like kernel from scratch long back. It was not a fully functional unix but had a lot of components ready to the extent it would provide a command line interfaceI from which you can run commands like ls, cat etc by loading elf executables.

I would highly recommend to read this tutorial http://www.osdever.net/bkerndev/Docs/title.htm

Feel free to reach out to me (email in my profile page) if you want to discuss the topic with me.

I worked on a similar thing but on CDN edge. This was without changing the app and logic in the CDN used to cache the header from past responses and used to send the header immediately without waiting for response from the origin server. Once received the response form the origin server, we used to match with the header that was already sent. If there was a mismatch, then we used to see if we could fix the mismatch by adding a script tag in the beginning of the body otherwise we used to reload the front end.

I think, nowadays the majority of web apps react apps. The html generation logic is at the front end and the front end only does rest API calls. So this kind of optimization is not very useful.

We had this patent https://patents.google.com/patent/US20150012614A1

> the vast majority of his work was simply incorrect

My impression is just opposite. Do you have any reference proving that vast majority of his work was simply incorrect.

I did this while I was doing my bachelor degree course. It was four year course and I started doing this sometime in 2nd year and continued till 4th year. I was not always writing code as I had to study other subjects as well. Also I was just learning coding and other computer science concepts, so it was like learning and writing code. But the writing the kernel forced me to learn many computer science concepts very deeply.

At the end, what I had was a kernel which could boot on bare metal (or VM) and provided a command line interface. It had a virtual file system layer and ext2 file systems, process management (fork, exec sys call), memory management (paging and process isolation) and device drivers for keyboard and hard disk. The kernel was able to fork and exec static ELF binary.

I did not reach to networking and threading. But that was next step which could make it complete unix kernel.

I implemented in bits of assembly(nasm) and C++. So I had to learn runtime and code generation aspect of c++. Based on that learning I wrote this articles on c++ object models and other internals. https://www.avabodh.com/cxxin/cxx.html

In the past, I wrote a unix like kernel from scratch in C++. I have summarized what I had to do to get C++ code run on bare metal in this article https://www.avabodh.com/cxxin/nostdlib.html

@matijash congrats on successfully raising. Your blog post is super useful. I am a founder and at an early stage now. I am trying to raise and facing the similar situation. I would like to learn more from you. I sent contact invite to you on LinkedIn. I hope you do not mind connecting.