LeonidBugaev

It is all useful, and depends on the distribution channel and surface complexity. I use both MCP and CLI in my case.

Having the CLI allows you to have the huge application with the big surface be available to the AI agent and so they'll be able to learn it on demand. For example my app has more than 1,000 help pages. There are no other ways to load all this information into the CLI context and to be frank it will be quite stupid. Instead I use progressive discovery.

It first reads the original help message to understand which stage of the flow it is right now. It auto-discovers the topics through the error messages and through the various hints. It has an Elasticsearch-like search inside its own help command so it's a full self-contained application with self-discoverability, which has a graph, ontology and the full developing flows inside of it. All of those parts are automatically given to the AI only when it actually needs them.

And to be frank if you look at the latest YCombinator batch, all of the companies are building the custom harness. What I'm calling above CLI is actually a custom harness, which is running inside someone else's agent group.

Having remote MCP allows me to give limited read-only surface to clients which not supposed to run any CLI commands. And while MCP has similar patterns, like bundling docs and prompts, it is not as flexible as CLI and does not allow huge scale.

So much energy in my life was spent making compromises between quality and speed, or pushing technical dev priorities. Constantly growing backlog, and guilt for the choices I had to do. What if we could fix the bug problem with AI? I mean, completely.

When you are measuring how intelligent model is, in fact you mostly measure how good it was following the plan, or question, based on your own intelligence.

Difference is subtle, but crucial. And it is so easy to be confused, while measuring small model performance, like various Flash variants or local models.

Smaller models became quite good at doing the work under the well specified goal, like fixing the well specified bug. But can it find the bug, without you giving it any clues?

In my benchmark the answer is mostly No. Being able to invent the questions - thats the real intelligence - and thats where the big models show true strength.

About 2 years ago I have created a small project https://helpwanted.dev OSS community to match people who looking for contribution with projects who actually ask for the help. It is very active and actual as ever! So a few reflections on whats the current pulse in the OSS community. What works and what not. Big projects are closing down. And now we have so many new people being enabled to fulfil their own ideas and it is renaissance of the OSS not the end of it! Bigger projects will find a way how to fix contribution inflow. But we should look at it as opportunity to find the people with the same interests.

"Install XERJ (docs: https://xerj.org/llms.txt), Allows you to write prompts like "index this project's sources, and set up reference coding: clone and index the open-source repos closest to what we're building, and search how they solved a problem before writing code.". Saves up to 5x tokens.

Not only attackers. I have to switch to Kimi or GLM even in cases of basic issue triage on my own projects! Current guardrails are ridiculous.

I recently had experience working with one of the top security audit firms that collaborate with OpenAI on their cybersecurity tools. What struck me is that they found a lot of various issues (not all of them ended up to be legit, but a lot is) and filled the backlog of the project. However every collaboration with that

person was essentially a proxy for some security-related agentic prompt: - Hey, agent, open the issue! - Hey, I clarified it with the agent! - Agent disagrees! - Agent agrees!

Thats exactly same as Claude Code offer: https://code.claude.com/docs/en/quickstart

Thankfully Google has some basic protection for it. I accidentally commited my google api token, as part of some OTEL trace JSON file, and within a few minutes my key was automatically locked by google, and marked as leaked (with exact link pointing where it has happened).

I had a few useful examples of this. In order to make it work you need to define your quality gates, and rather complex spec. I personally use https://github.com/probelabs/visor for creating the gates. It can be a code-review gate, or how well implementation align with the spec and etc. And basically it makes agent loop until it pass it. One of the tips, especially when using Claude Code, is explictly ask to create a "tasks", and also use subagents. For example I want to validate and re-structure all my documentation - I would ask it to create a task to research state of my docs, then after create a task per specific detail, then create a task to re-validate quality after it has finished task. You can also play around with the gates with a more simple tooling, for example https://probelabs.com/vow/

Hope it helps!

Nice one! I had my own spin on this issue as well, but from the other angle https://github.com/probelabs/maid

Getting AI to generate valid mermaid diagrams on scale extremely hard. With maid i'm hitting 100% accuracy.

Maid is basically built from scratch mermaid parser, without any dependnecies, which knows how to auto-fix common AI slop diagramming issues.

It does not implement the Auth :)

(mcp auth is terrible btw)

You should check https://probeai.dev/ too. Thats one of those building blocks which makes AI trully understand the code.

To put it simple:

A2A is for communication between the agents. MCP is how agent communicate with its tools.

Important aspect of A2A, is that it has a notion of tasks, task rediness, and etc. E.g. you can give it a task and expect completely in few days, and get notified via webhook or polling it.

For the end users for sure A2A will cause a big confusing, and can replace a lot of current MCP usage.

Hello HN!

I'm building Probe https://probeai.dev/ for a while now, and this this docs-mcp project is showcase of its capable. Giving you a local semantic search over any codebase or docs without indexing.

Feel free to ask any questions!