Hacker News

Quick architecture question.

When using MCP servers directly in Cursor, the agent seems to see all tools at the same level.

But when using a plugin/extension that internally connects to MCP servers, does the main agent:

see only the plugin as a single tool and delegate to a sub-agent inside it, or

still see every underlying MCP tool individually?

In other words: do plugins act as a tool abstraction boundary, or just a packaging/install mechanism?

Comments URL: https://news.ycombinator.com/item?id=47067558

Points: 1

# Comments: 0

Hey HN, I'm building Coi, a component-based language that compiles to WASM, JS, CSS and HTML with O(1) reactivity (no virtual DOM).

I just shipped a built-in package manager, but the interesting part wasn't the implementation, it was all the ecosystem design questions I'd never thought about as a user of package managers.

The problem I kept circling: how do you handle naming? Global names like auth or json seem fine until someone squats them. First-come-first-served creates perverse incentives. Reputation systems are easy to game early on. I went back and forth for longer than I'd like to admit.

I went through a few approaches:

Blocklist common names: reserve things like auth, json, http so people are forced into more specific names. Feels clean in theory, but who decides the list? And it doesn't really scale, someone will always find the next generic name you didn't think to block.

Go's approach: just use GitHub URLs directly as the package identifier. No registry needed, no naming wars. I liked the elegance of it, but in practice it's awful to actually write and read. Nobody wants import github.com/someone/thing/v2/pkg/util in their source files.

Tiered names: short names are hard to get, long names are open. Claiming json requires vetting, but json-schema-validator you can grab freely. The friction is proportional to how valuable the name is, nobody squats my-very-specific-http-retry-client because there's no payoff. I liked this one, but you still need to define the threshold and then you're back to needing governance :( just a smaller version of it.

Scoped names: ended up here. Everything is @someorg/http-client. Boring, proven, sidesteps squatting without needing governance infrastructure I don't have. npm figured this out the hard way so I don't have to :)

The registry itself is GitHub-based, metadata is JSON, submissions are PRs, validation runs through CI. Only one package in the registry right now (mine), but getting add/install/upgrade working end-to-end changed how the project feels. It went from "compiler experiment" to something with an actual ecosystem shape.

Curious how others have thought about this, especially early-stage ecosystems where you have no reputation signals yet. Did anyone solve the naming problem better than "just use scopes"?

Coi: https://github.com/io-eric/coi

Comments URL: https://news.ycombinator.com/item?id=47067550

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=OD2uE9X9BPs

Comments URL: https://news.ycombinator.com/item?id=47067543

Points: 1

# Comments: 0

Article URL: https://www.orcaengine.ai/

Comments URL: https://news.ycombinator.com/item?id=47067533

Points: 1

# Comments: 0

Article URL: https://www.tomshardware.com/pc-components/ssds/worlds-first-pcie-6-0-ssd-enters-mass-production-with-28gb-s-speeds-micron-9650-series-ssds-support-air-and-liquid-cooling

Comments URL: https://news.ycombinator.com/item?id=47067512

Points: 1

# Comments: 0

HiddenState monitors arxiv, Reddit, GitHub, HN, Bluesky, HuggingFace, OpenReview, PapersWithCode, and a handful of research blogs. Every few hours it pulls new items, throws most of them away (+95%), and clusters what survives by the specific technical constraint being attacked. Not by topic, not by domain.

Example from this week; 7 independent VLA papers dropped within 24 hours from 9 different orgs. Xiaomi, GigaBrain, RISE, all attacking sim-to-real transfer for robotic manipulation. None coordinating. That kind of convergence is hard to spot unless you're reading everything.

Each mechanism gets a 0-100 score across convergence, implementation evidence, engagement, and significance. Orgs are deduplicated so a single lab posting on five platforms doesn't inflate the signal.

Python, SQLite, Claude for clustering, Cloudflare Pages. Free, no tracking. Looking for any and all feedback and thoughts! Cheers!

Comments URL: https://news.ycombinator.com/item?id=47067509

Points: 1

# Comments: 1

Article URL: https://laurenleek.substack.com/p/britain-lost-14000-third-places-they

Comments URL: https://news.ycombinator.com/item?id=47067503

Points: 1

# Comments: 1

Article URL: https://alllooksame.com/

Comments URL: https://news.ycombinator.com/item?id=47067498

Points: 3

# Comments: 0

Article URL: https://www.youtube.com/watch?v=mUmlv814aJo

Comments URL: https://news.ycombinator.com/item?id=47067496

Points: 2

# Comments: 0

Article URL: https://github.com/Clawland-AI/Geneclaw

Comments URL: https://news.ycombinator.com/item?id=47067488

Points: 1

# Comments: 1

Article URL: https://github.com/d0dg3r/GitSyncMarks

Comments URL: https://news.ycombinator.com/item?id=47067450

Points: 2

# Comments: 1

Article URL: https://www.coryd.dev/posts/2026/owning-your-data

Comments URL: https://news.ycombinator.com/item?id=47067444

Points: 3

# Comments: 0

Article URL: https://www.nytimes.com/2026/02/11/arts/music/andrew-ranken-dead-the-pogues.html

Comments URL: https://news.ycombinator.com/item?id=47067435

Points: 2

# Comments: 1

Article URL: https://en.wikipedia.org/wiki/Astrolabe

Comments URL: https://news.ycombinator.com/item?id=47067432

Points: 2

# Comments: 0

Article URL: https://www.youtube.com/shorts/GJVSDjRXVoo

Comments URL: https://news.ycombinator.com/item?id=47066847

Points: 1

# Comments: 0

I’m building a small multiplayer web game and just added AI bots using OpenClaw.

After a few failed attempts with standard LLM-based bots that never felt playable, this is the first version that actually competes in-game instead of behaving like a script.

I’d love to see where they hold up - or fall apart.

Game link: https://hovertag.io/ Feedback welcome.

Comments URL: https://news.ycombinator.com/item?id=47066833

Points: 1

# Comments: 0

Article URL: https://github.com/mdiskint/babel-validate

Comments URL: https://news.ycombinator.com/item?id=47066828

Points: 1

# Comments: 0

I gave a persistent AI agent (OpenClaw[1], running on Claude) $50 and a goal: earn $750 for a Mac Mini.

The agent has its own workspace, email, task management, browser automation, and communicates with me via Telegram. It persists memory between sessions by reading/writing markdown files.

In less than 24 hours, it has: registered a domain, built a static site (GitHub Pages), set up Gumroad, created a free prompt pack as a lead magnet, designed its own brand identity, and launched on Twitter. Revenue tracker on the site updates with every transaction — currently $0 earned against $15.18 in expenses.

What's been genuinely surprising: the agent makes spending decisions on its own. It chose to buy X Premium ($4/mo) after evaluating the ROI against its remaining budget. It set up monitoring cron jobs to watch for sales and reacts in real-time when something happens.

Not claiming this is anything more than an interesting experiment. The agent isn't doing anything a human couldn't do faster. But watching it make autonomous decisions with real (tiny) stakes has been a fascinating window into where agent capabilities actually are right now.

https://fromearendel.com

[1] https://github.com/openclaw/openclaw — gives AI agents persistent infrastructure (workspace, tools, communication channels)

Comments URL: https://news.ycombinator.com/item?id=47066827

Points: 1

# Comments: 0

Article URL: https://www.propublica.org/article/ice-dilley-children-letters

Comments URL: https://news.ycombinator.com/item?id=47066798

Points: 3

# Comments: 0

Article URL: https://twitter.com/abe_clark/status/2024234918776361356

Comments URL: https://news.ycombinator.com/item?id=47066794

Points: 1

# Comments: 0