Hacker News: mattw1810

Show HN: ContextFS – virtual file system abstraction for agent context

mattw1810 — Sat, 10 Jan 2026 18:05:50 +0000

ContextFS is a library that lets you construct a virtual filesystem using a router-style API. It's mountable via NFS or `just-bash`, and useful for providing agents with hierarchical, explorable context defined in code.

  // Define file system structure
  const vfs = createFileSystem({
    issues: {
      ":id": {
        [list]: () => getIssues().map(i => ({ id: i.id })),
        [read]: (c) => getIssue(c.params.id).body,
      }
    }
  });

  // Mount with NFS
  await mount(vfs, { mountPoint: "/tmp/issues" });
  await fsp.readFile("/tmp/issues/123");

  // Explore with just-bash
  const env = new Bash({ fs: new ContextFS(vfs) });
  await env.exec("ls /tmp/issues");

Key features: - Express-style dynamic paths (:id, :slug.md) with full TypeScript inference - Read and optional write support - Symlinks to represent relationships - Browser-safe core library - Pure TypeScript NFS v3 implementation, no native deps

Comments URL: https://news.ycombinator.com/item?id=46568246

Points: 2

# Comments: 0

New comment by mattw1810 in "OpenAI Progress"

mattw1810 — Sun, 17 Aug 2025 07:05:15 +0000

All the same they choose to highlight basic prose (and internal knowledge, for that matter) in their marketing material.

They’ve achieved a lot to make recent models more reliable as a building block & more capable of things like math, but for LLMs, saturating prose is to a degree equivalent to saturating usefulness.

New comment by mattw1810 in "OpenAI Progress"

mattw1810 — Sat, 16 Aug 2025 17:03:28 +0000

On the whole GPT-4 to GPT-5 is clearly the smallest increase in lucidity/intelligence. They had pre-training figured out much better than post-training at that point though (“as an AI model” was a problem of their own making).

I imagine the GPT-4 base model might hold up pretty well on output quality if you’d post-train it with today’s data & techniques (without the architectural changes of 4o/5). Context size & price/performance maybe another story though

New comment by mattw1810 in "A critical look at MCP"

mattw1810 — Sat, 10 May 2025 21:29:06 +0000

This is no different from a web app though, there’s no obvious need to reinvent the wheel. We know how to do this very very well: the underlying TCP connection remains active, we multiplex requests, and cookies bridge the gap for multi-request context. Every language has great client & server support for that.

Instead we ended up with a protocol that fights with load balancers and can in most cases not just be chucked into say an existing Express/FastAPI app.

That makes everything harder (& cynically, it creates room for providers like Cloudflare to create black box tooling & advertise it as _the_ way to deploy a remote MCP server)

New comment by mattw1810 in "A critical look at MCP"

mattw1810 — Sat, 10 May 2025 21:18:07 +0000

I think some of the advanced features around sampling from the calling LLM could theoretically benefit from a bidirectional stream.

In practice, nobody uses those parts of the protocol (it was overdesigned and hardly any clients support it). The key thing MCP brings right now is a standardized way to discover & invoke tools. This would’ve worked equally well as a plain HTTP-based protocol (certainly for a v1) and it’d have made it 10x easier to implement.

New comment by mattw1810 in "A critical look at MCP"

mattw1810 — Sat, 10 May 2025 20:57:15 +0000

MCP should just have been stateless HTTP to begin with. There is no good reason for almost any of the servers I have seen to be stateful at the request/session level —- either the server carries the state globally or it works fine with a session identifier of some sort.

New comment by mattw1810 in "Google Gemini has the worst LLM API"

mattw1810 — Sun, 04 May 2025 09:12:08 +0000

Their patchy JSON schema support for tool calls & structured generation is also very annoying… things like unions that you’d think are table stakes (and in fact work fine with both OpenAI and Anthropic) get rejected & you have to go reengineer your entire setup to accommodate it.

New comment by mattw1810 in "Act on Press"

mattw1810 — Thu, 23 May 2024 21:14:15 +0000

Funnily enough I recently noticed that the X app on iOS started doing this for me… on ads. If I place my finger on an ad while scrolling down, without fault it opens the ad overlay sheet. I guess that’s one way to increase CPC revenue

New comment by mattw1810 in "The perfect configuration format? Try TypeScript"

mattw1810 — Wed, 17 Nov 2021 19:52:42 +0000

The Rust implementation has it as an experimental extension (https://github.com/facebookexperimental/starlark-rust/blob/m...)

New comment by mattw1810 in "Luxembourg set to make all public transport free"

mattw1810 — Thu, 06 Dec 2018 00:24:03 +0000

To be fair, in my experience ticket checking in Luxembourg was (at least in the city buses) so incredibly lax already that not a lot of revenue will be lost by this. All the same it's a nice policy though.