News for hackers. Stuff that matters.

Gusarich

Gusarich on Hacker News

Re:Claude Opus 4.6 on Thursday February 05, 2026 @12:41PM (#46902259)
not out yet

Show HN: Pitaya – Orchestrate AI coding agents like Claude Code

Posted by Gusarich on Sunday August 31, 2025 @12:33PM
from the show-hn dept.

https://github.com/tact-lang/pitaya

Pitaya is a local, open-source orchestrator for AI coding agents (Claude Code, Codex CLI). It runs many agents in parallel, isolates each in Docker with its own git branch, supports pluggable Python strategies, and persists state so runs are resumable. Quickstart + short demo are in the README.

( Read More... | no comments | 4 points | on HN )
Re:A "Livestream" Dashboard for Hacker News – Newest Story and Live Co... on Monday August 18, 2025 @02:05AM (#44937833)
cool but i'd rather have it as a browser extension for hn and not as a separate website
Re:Is AI Hitting a Wall? on Sunday August 17, 2025 @01:26AM (#44929104)
it's not hitting a wall anytime soon. just look at gpt-5.
Re:LLMs' "simulated reasoning" abilities are a brittle mirage on Tuesday August 12, 2025 @01:59AM (#44872901)
The article already seems outdated on the first day. The key points about SFT are irrelevant in the era of RL.
Re:Ask HN: Claude Code vs. ChatGPT Pro? on Monday July 28, 2025 @12:22AM (#44707220)
Get both ChatGPT Pro and Claude Max if you can. Otherwise, it's a tough choice and depends on how much of your AI time is spent coding versus chatting. If you mostly code, go with Claude Max; otherwise, I'd prefer ChatGPT Pro + Claude Pro.
Re:Ask HN: What are you working on? (July 2025) on Monday July 28, 2025 @12:14AM (#44707181)
A configurable and extensible orchestrator for Claude Code (and other agents). Turns out it can be effectively used not only for coding, but also for reviews, testing, and other tasks. https://github.com/Gusarich/orchestrator
Re:Ask HN: Best Practices for Writing Tests? on Friday July 18, 2025 @07:24AM (#44603457)
That's too generic a question. There are many different techniques, frameworks, and approaches, so there's no single universal answer. But assuming you're asking about testing general-purpose software without specific requirements, I'd recommend the following approach:

1. Basic tests that ensure your software runs without crashing or breaking.

2. Specific test cases for various inputs, settings, and invalid scenarios.

3. Edge-case tests covering extreme inputs or unusual combinations.

4. If your software's behavior is predictable or you can easily verify correctness, consider randomized testing to validate thousands of random inputs automatically.

Re:4o Image Generation on Wednesday March 26, 2025 @06:03AM (#43480510)
It produces amazing results for me! But the wow effect would have been greater if they had released it a few months ago.