Hacker News: fwystup

New comment by fwystup in "Evaluating AGENTS.md: are they helpful for coding agents?"

fwystup — Wed, 18 Feb 2026 12:34:57 +0000

My opinions about the study:

- Don't state the obvious: I wouldn't hand a senior human dev a copy of "Clean Code" before every ticket and expect them to work faster.

- File vs. Prompt is a false dichotomy: The paper treats "Context Files" as a separate entity, but technically, an AGENTS.md is just a system prompt injection. The mechanism is identical. The study isn't proving that "files are bad," it's proving that "context stuffing" is bad. Whether I paste the rules manually or load them via a file, the transformer sees the same tokens.

- Latent vs. Inferable Knowledge: This is the key missing variable. If I remove context files, my agents fail at tasks requiring specific process knowledge - like enforcing strict TDD or using internal wrapper APIs that aren't obvious from public docs. The agent can't "guess" our security protocols or architectural constraints. That's not a performance drag; it's a requirement. The paper seems to conflate "adding noise" with "adding constraints."

New comment by fwystup in "I put a real-time 3D shader on the Game Boy Color"

fwystup — Mon, 09 Feb 2026 01:36:41 +0000

Really interesting Project, thanks for sharing. Reminded me a bit of my time coding assembly on the C64 (yeah, I'm old). For 3D (wire-frame) we also needed to find creative ways around hardware limitations, especially the lack of a multiply instruction.

New comment by fwystup in "Running Claude Code dangerously (safely)"

fwystup — Tue, 20 Jan 2026 16:54:27 +0000

Posted almost at the same time about Kata. I'm trying to use Kata as replacement for the standard docker runtime (since I already have a tool based on docker).

The idea is to simply use the runtime flag (after kata install):

docker run -d --runtime=kata -p 8080:8080 codercom/code-server:latest

Hope this works, with this I could keep my existing docker setup.

New comment by fwystup in "Running Claude Code dangerously (safely)"

fwystup — Tue, 20 Jan 2026 14:56:54 +0000

I'm currently building a Docker dev environment for VSCode (github.com/dg1001/xaresaicoder) usable in a browser and hit the same issue. Without docker-in-docker it works well - I even was able to add transparent proxy in the Docker network to restrict outbound traffic and log all LLM calls (pretty nice in order to document your project). For docker-in-docker development and better security isolation, I'm considering Kata Containers instead of Vagrant. Which gives me real VM-level isolation with minimum perf overhead, while still be able to use my docker stuff. Still on my TODO list though. Has anyone actually run Kata with vs code server? Curious about real-world quirks - I've read that storage snapshot performance can be rough.

New comment by fwystup in "Claude Haiku 4.5"

fwystup — Thu, 16 Oct 2025 11:25:14 +0000

Does anybody also see this with claude code and haiku 4.5 (tried to set the env var, no change): "API Error: Claude's response exceeded the 8192 output token maximum. To configure this behavior, set the CLAUDE_CODE_MAX_OUTPUT_TOKENS environment variable."?

New comment by fwystup in "Show HN: I analyzed why my post got 0 votes and built this"

fwystup — Sun, 10 Aug 2025 23:58:43 +0000

Glad to hear it. Your NLTK approach sounds interesting — would love to hear more about it. BTW, I’m planning to improve my project’s documentation. Funny enough, under “consideration” it flagged itself: "Documentation quality is not explicitly high. The effectiveness of the LLM’s scoring criteria is subjective and not deeply explained." Sadly, that’s true.

New comment by fwystup in "Show HN: I analyzed why my post got 0 votes and built this"

fwystup — Sun, 10 Aug 2025 23:50:16 +0000

Got it, thanks for clarifying. Makes sense that participation matters more than raw karma. I’ll keep that in mind and try to contribute more broadly, posting less about my own projects.

New comment by fwystup in "Show HN: I analyzed why my post got 0 votes and built this"

fwystup — Sun, 10 Aug 2025 23:26:28 +0000

You're right — I totally missed that. Fixed it, please reload.

New comment by fwystup in "Show HN: I analyzed why my post got 0 votes and built this"

fwystup — Sun, 10 Aug 2025 21:12:25 +0000

Thanks for the feedback — definitely not meant as spam. It’s open source, not a commercial blog. Just wanted to surface some fresh ideas that might otherwise get missed. Really appreciate the tip!

Show HN: I analyzed why my post got 0 votes and built this

fwystup — Sun, 10 Aug 2025 19:35:18 +0000

Maybe you've had this experience too: You build something you're proud of, post it on HN with your low-karma account, and... crickets. Zero votes, zero comments. That's what happened to me last Monday. I posted my coding tool (XaresAICoder - an open-source browser IDE) that I'd built with AI assistance. In my mind it was revolutionary. On HN? Completely ignored. Then I wondered: How many other potentially great projects suffer the same fate? What "hidden gems" are we missing because they come from low-karma accounts? So I built hn-gems (with help from Claude and my own XaresAICoder). It works in two stages:

Continuous scanning: Analyzes all new HN posts from accounts with <100 karma, scoring them for technical merit, originality, and problem-solving value AI curation: Every 12 hours, an LLM deep-dives into the top 10 candidates, checking GitHub repos, documentation quality, and actual utility

The result is what you see at the link - a curated list of overlooked quality posts that deserve more attention. The interesting part: I barely wrote any criteria. I just told Claude "open source good, pure commercial bad, working demos good" and let it figure out the scoring. The AI assessment varies slightly each run, which actually makes it more interesting. GitHub: https://github.com/DG1001/hn-gems

Is this useful? Do you have ideas how to improve this tool if necessary?

(And yes, my XaresAICoder that got 0 votes? The AI thinks it's actually pretty good. I'll take that as a win.)

Comments URL: https://news.ycombinator.com/item?id=44857620

Points: 5

# Comments: 8

New comment by fwystup in "How I code with AI on a budget/free"

fwystup — Sun, 10 Aug 2025 18:28:31 +0000

Qwen3-Coder-30B-A3B-Instruct-FP8 is a good choice ('qwen3-coder:30b' when you use ollama). I have also had good experiences with https://mistral.ai/news/devstral (built under a collaboration between Mistral AI and All Hands AI)

Show HN: XaresAICoder – Open-Source Browser IDE with Claude, Gemini, Aider

fwystup — Mon, 04 Aug 2025 06:54:48 +0000

I recently gave a talk about AI coding tools at the Java Forum Stuttgart. While preparing demos and switching between different AI assistants, I realized how tedious it is to constantly juggle environments and tools. That sparked the idea for XaresAICoder. The problem: Many developers can't install tools on their corporate notebooks but want to experiment with AI coding. Others waste time managing ports, environments, and worrying about AI agents accidentally modifying files outside project boundaries. My solution: XaresAICoder provides isolated browser-based development environments with VS Code (via code-server) at their core. Each workspace comes pre-configured with Claude Code, Gemini CLI, OpenCode SST, and Aider - allowing you to test different LLMs (including locally-hosted models with Aider) without installation hassles.

Key features:

One-click project creation with Python/Flask, Node/React, or Java/Spring templates Sandboxed execution - no port conflicts or accidental file modifications Integrated Forgejo Git server for automatic version control Runs with docker on WSL, Linux (local or server), macOS likely works too (can someone confirm?) Container-level password protection for workshops/internal networks

This isn't another "vibe-coding" browser tool - it's a real development environment for professional developers who need proper tooling but face installation restrictions or want safer AI experimentation. All Open Source and on your machines. And: made mostly with AI (Claude); really great when you just sit and watch how claude code fixes docker config problems and do tests runs.

I'm curious: What's your biggest friction point when working with AI coding assistants? Happy to discuss the architecture!

Comments URL: https://news.ycombinator.com/item?id=44782782

Points: 2

# Comments: 0