Hacker News: ibash

New comment by ibash in "Whats Up with Claude Lately?"

ibash — Tue, 03 Mar 2026 04:44:20 +0000

Aha! Same for me. A telling sign to me is it will take him 2-3 turns of “that’s not right” before he’ll say “let me do this properly…” and do what I original requested.

New comment by ibash in "Very Important People"

ibash — Wed, 10 Dec 2025 08:18:37 +0000

It’s not celebrity worship, it’s trying to show the absurdity of celebrity worship.

New comment by ibash in "Claude Advanced Tool Use"

ibash — Mon, 24 Nov 2025 23:36:40 +0000

Better than rest is a low bar though. Ultimately agents should rarely be calling raw rest and graphql apis, which are meant for programmatic use.

Agents should be calling one level of abstraction higher.

Eg calling a function to “find me relevant events in this city according to this users preferences” instead of “list all events in this city”.

New comment by ibash in "Do not put your site behind Cloudflare if you don't need to"

ibash — Tue, 18 Nov 2025 19:09:22 +0000

Kamal + cloudflare tunnel is a neat setup.

New comment by ibash in "Alchemy"

ibash — Sun, 16 Nov 2025 17:31:20 +0000

AI is having the same effect on art as the iPhone did on photography.

There’s a lot more photos now, most of them mediocre, but some exceptional.

It does become harder to filter great photography from noise.

New comment by ibash in "Collaboration sucks"

ibash — Thu, 13 Nov 2025 02:55:37 +0000

Except, they have one person in the ear. Not 4-5, not people giving opposite opinions, not drive by takes.

By the time a race engineer is communicating with a driver all of that has been shaken out. Specific concrete options are given to the driver, and usually only one.

New comment by ibash in "Most Americans say 'Arabic numerals' should not be taught in school (2019)"

ibash — Fri, 07 Nov 2025 00:38:28 +0000

This is one of those meaningless trivia questions.

Its not bigotry nor is it clever. It’s just word play.

New comment by ibash in "US may cut air traffic 10% by Friday without shutdown deal, sources say"

ibash — Thu, 06 Nov 2025 01:52:04 +0000

This is very French!

New comment by ibash in "Ask HN: What's the best way to build agents in Oct 2025"

ibash — Fri, 31 Oct 2025 23:53:04 +0000

Pydantic ai is pretty good. There were some growing pains as it moved towards 1.0, but going forward the API should be mostly stable.

It does have slightly more complexity than I’d like, but most of the design decisions are sound and the code is easy to read.

New comment by ibash in "Composer: Building a fast frontier model with RL"

ibash — Wed, 29 Oct 2025 16:17:13 +0000

Very cool, congrats!

New comment by ibash in "UIs Are Not Pure Functions of the Model – React.js and Cocoa Side by Side (2018)"

ibash — Wed, 29 Oct 2025 14:09:37 +0000

Influencer marketing is a big part of the problem here.

The best practices engineering influencers shill are geared towards shipping a todo list quickly, and no thought is given towards maintainability.

New comment by ibash in "UIs Are Not Pure Functions of the Model – React.js and Cocoa Side by Side (2018)"

ibash — Wed, 29 Oct 2025 13:56:55 +0000

But if you advocate for this now, newer engineers think you’re doing it wrong.

New comment by ibash in "Using AI to negotiate a $195k hospital bill down to $33k"

ibash — Tue, 28 Oct 2025 16:24:17 +0000

The thing is the prices are all made up anyway. The hospital hallucinates prices, so they don’t blink an eye when an llm does the same.

New comment by ibash in "Journalists turn in access badges, exit Pentagon rather than agreeing new rules"

ibash — Thu, 16 Oct 2025 09:03:30 +0000

Because it's interesting. We don't need to censor everything.

New comment by ibash in "What is “good taste” in software engineering?"

ibash — Mon, 29 Sep 2025 07:45:57 +0000

I disagree. The ability of someone to read code doesn't grow exponentially, after a few years of experience everyone hits the same plateau. More years of experience does not mean you can understand more complex code.

That is to say, if you target "readable to the majority of engineers with 3-4 years of experience, without them getting confused" then you've hit the mark.

New comment by ibash in "Less is safer: Reducing the risk of supply chain attacks"

ibash — Sat, 20 Sep 2025 03:58:02 +0000

> Obsidian plugins have full, unrestricted access to all files in the vault.

Unless something has changed, it's worse than that. Plugins have unrestricted access to any file on your machine.

When I brought this up in discord a while back they brushed it aside.

New comment by ibash in "React is winning by default and slowing innovation"

ibash — Tue, 16 Sep 2025 06:10:49 +0000

Disagree. Hooks took one problem and reshaped it, they didn’t actually solve the problem.

With hooks you still need to think about lifecycle, side effects, and state, but the syntax changes.

The real solution is overall application design and thinking through architecture, but unfortunately that’s higher effort than “just use hooks bro”.

Show HN: MMO MCP – A multiplayer game you play through Claude Code and Cursor

ibash — Tue, 02 Sep 2025 15:46:06 +0000

Hi HN! I built a multiplayer text game where players connect through Claude Code, Cursor, and other MCP clients.

I've been enjoying Claude Code and wondered what would happen if I made a multiplayer game as an MCP tool. It's a weekend experiment to see what happens when AI becomes the interface to multiplayer games.

Every player is a real person (or autonomous AI) leaving permanent marks on a shared world. Break a chair, it stays broken for everyone. Write on a wall, others will find it. Claude/Cursor is just your interface - like using a very smart terminal to play a MUD.

Try it: No signup needed. Just make up any player_id and password:

- Live at: https://www.summon.app

- Claude Code: `claude mcp add game "https://mcp.summon.app/mcp?player_id=&password=" --transport http`

- Cursor: there's an "add to cursor" button on the website

The world persists between sessions. Use 'play' to create a character, then explore. You can 'conjure' objects and 'do' arbitrary actions (interpreted by LLM).

No real-time messaging yet, but you can leave messages by conjuring signs and writing on them.

Technical: Built with fastmcp and runs a small LLM server-side to interpret actions. The interesting part is how different the game feels when your interface is an AI that can creatively interpret your intentions.

Would love to have folks join and explore together.

Comments URL: https://news.ycombinator.com/item?id=45104635

Points: 1

# Comments: 0

New comment by ibash in "AI for Coding: Why Most Developers Are Getting It Wrong"

ibash — Mon, 28 Jul 2025 16:58:39 +0000

That’s not quite right.

The tools are going to get better, and it’s going to lift everyone while doing so. But it will vary across task.

The bottom tier of engineers will have improvements to their code.

The top engineers will move faster. Top engineers will still be great at what they do.

—-

The same thing is happening in music. Everyone gets lifted but on different dimensions.

New comment by ibash in "AI for Coding: Why Most Developers Are Getting It Wrong"

ibash — Mon, 28 Jul 2025 16:54:07 +0000

I don’t think it’s a craft thing, I think it’s a speed thing. Senior engineers are faster in their existing workflows.

I’ve found ai tools most valuable for:

1. Quick “how to do x in y” language

2. Large scale refactorings that are mostly mechanical.

This still takes a bit of guidance to get the right output (and breaking down the refactoring an into multiple steps). But it does speed things up when I would touch 40-some files. I still review all the code.