#aider
61 items
Same 9B Qwen weights: 19.1% in Aider vs 45.6% with a scaffold adapted to small local models (www.reddit.com) Show HN: Which public repos are friendliest to an AI coding agent? (www.agentfriendlycode.com via hn) Public leaderboard ranking GitHub, GitLab, and Bitbucket repos by how agent-friendly they are for Claude Code, Cursor, Devin, GPT-5 Codex, Gemini CLI, Aider, OpenHands, and Pi — per model, with AGENTS.md / CLAUDE.md, CI, tests, and dev-env…
Show HN: Thurbox – A tmux-based TUI and CLI for local AI agent orchestration (github.com via hn) Thurbox Run several coding agents at once — Claude Code, Codex, Antigravity, opencode, aider, or any CLI you describe yourself — side by side in one terminal. Each gets its own persistent tmux session and its own git worktree, so they neve…
A directory of AI agents, MCP servers and agent skills, cross-linked (aiagentslisting.com via hn) 5 Agentic AI Coding Tools Compared for Developers A curated list of five agentic AI coding tools — Claude Code, Cursor, OpenAI Codex CLI, Aider, and OpenCode — and how to choose one for your workflow. 6 min read A curated directory of AI a…
New to agents, mcp , etc how do I get to a point where i can lay back and let my agents do the work (www.reddit.com) Currently working on some projects. I have some agents and chrome scrap tasks id like it to do.
Show HN: Zerminal – a terminal-first Zed fork for AI coding agents (zerminal.dev via hn) A terminal-first development environment for agentic coding. Use Claude Code, Codex, Aider, and other CLI agents in a focused workspace.
The Benchmark Gap: 1,472 runs show coding-agent context changes outcomes (github.com via hn) The Benchmark Gap 1,472 OpenCode runs reveal what vendor scores don't tell you about AI coding tools. Plus a 60-run supplemental Aider/Cline probe, a 48-run installed-CLI smoke, and a 10-row current-OpenCode context probe to test whether t…
Honey, I Shrunk the Coding Agent (itayinbarr.substack.com via hn) Honey, I Shrunk the Coding Agent Coding Agent Adaptation Lets a 9B LLM Outperform 10x Larger Models on Aider Polyglot Benchmark Small local language models often appear weak when they are placed inside coding agents designed around frontie…
Faster Gemma 4 on MLX with multi-token prediction (ollama.com via hn) Gemma 4 is now significantly faster in Ollama 0.31 on Apple Silicon via multi-token prediction (MTP), powered by MLX. Performance is now up to 90% faster when used with coding agents, as measured using the Aider polyglot benchmark.
Supervising AI Agents (github.com via hn) AI Agent Control Checklist A practical checklist for supervising AI coding agents across branches, worktrees, reviews, approvals, and human intervention points. The problem AI coding agents -- Claude Code, Cursor, Codex, Aider, OpenCode, a…
Show HN: Fabrika, from babysitting agents to software factories (fabrika-ai.com via hn) Software factory running on your computer. Fabrika turns CLI coding agents — Claude Code, Aider, any command — into a managed team.
Anyone evaluated the difference between Qwen Code for the local qwen models vs another harness? CC, OC, LC, Aider etc.. (www.reddit.com) For me, opencode doing fantastically but was wondering if qwen code would be more native and have better functionality, since idk which agentic harness they used to get their benchmark results
Sequencer: Visual multi-agent workflow pipelines. (www.reddit.com) I built Sequencer, an open-source visual prompt-to-agent chaining engine. When I build apps with AI tools, I break the project into bite-sized prompts, then copy-paste each one into Cline or Aider and wait.
Making AI coding sessions persistent across agents (github.com via hn) 🌐 English · 日本語 · 简体中文 · 繁體中文 drift_ai Vendor-neutral handoff for AI coding tasks — between Claude, GPT, Gemini, DeepSeek, local LLMs. Reads from Claude Code, Codex, Cursor, Aider.
How do I use gemma4 on 5090 gpu for coding? (www.reddit.com) I'm trying to replace openai codex which i used for development all the time, with gemma4 on 4090, small tasks it solves quite impressively, but i need to have some agent. So I tried to connect 31b to cline and to aider and it didn't reall…
Ask HN: Anyone using dictation with coding agents? (news.ycombinator.com) Curious if folks are using dictation with coding agents for speach to text? Anyone completely hands-free with spoken responses from the agent as well?
Show HN: Stream – Unlimited DeepSeek v4 Flash 0731 (camelai.com via hn) A flat-rate DeepSeek API for Hermes, OpenCode, OpenClaw, Aider, and other high-volume coding agents.
Husk – a desktop workspace for terminal AI agents (github.com via hn) Husk wraps claude, copilot, codex, aider, or any other terminal-based AI agent in a clean Electron window with a real PTY, drag-drop file context, voice output, session resume, and a one-glance dashboard. The reasoning, thinking format, an…
Show HN: Dejavu, stop showing coding agents the same command output twice (github.com via hn) Dejavu Stop showing coding agents the same command output twice. Dejavu is a PATH shim for Claude Code, Codex, Cursor agent, opencode, Aider, Gemini CLI, and other terminal-based coding agents.
Show HN: Leakproof – local secret-egress firewall for Claude Code/Cursor/aider (github.com via hn) leakproof Local-first secret firewall for AI coding assistants. Your security team banned Claude Code or Cursor over data egress.
We built Branchless, a desktop app for running parallel dev sessions with agents, terminals and editors, without switching branches (www.reddit.com) Hey everyone, We have been building Branchless, a desktop app for Mac, Windows and Linux. The basic idea is simple: we wanted a way to work on multiple tasks at the same time without constantly switching branches, stashing changes, opening…
10-gate security audit SKILL for web apps (www.reddit.com) There are a few security focus SKILLs. We are working another new one for web app.
Show HN: Ait – Claude, Codex, and Aider as a team, on your laptop (github.com via hn) ait Local control plane for multi-agent AI coding Run Claude Code · Codex CLI · Aider · Gemini CLI · Cursor as a team on the same task — context handoff, review gate, attempt ledger — all on your machine. English · 繁體中文 60-second walkthrou…
Two flat-fee agent endpoints, no token meter: OpenClaw chat ($7/mo, 128K ctx) + All You Can Code ($19/mo, 256K ctx). OpenAI v1. (www.reddit.com) For anyone running agents (coding or otherwise) who'd rather pay a flat fee than meter tokens. Two tiers, both flat fee, both unlimited: OpenClaw ($7/mo) - Nemotron-3-Nano-Omni-30B-A3B - 128K context - For general-purpose agents: research,…
Why does GitHub Copilot feel less accurate compared to Agentic/Autonomous AI tools ? (www.reddit.com) I'm looking for a solid solution to bridge this gap. How can we actually use these tools properly for complex development?
We built a free tool that generates a DESIGN.md from any live URL, keeps AI coding agents on-brand (www.reddit.com) The Google Labs DESIGN.md spec launched last month, it's a machine-readable markdown file your AI coding agent reads to understand your design system. This tool automates creating it.
corrigé code apk (www.reddit.com) bonjour a tous je travaille sur un projet apk et j'ai rencontré quelques problème au niveau des notifications ect...si quelqu'un peut m'aider a corriger les code et faire fonctionner apk . je utilise en ce moment Android studio.
SafeSandbox – infinite undo for AI coding agents (Cursor, Claude Code, Codex) (github.com via hn) SafeSandbox Infinite undo for AI coding agents. SafeSandbox is a local-first developer tool that automatically creates snapshots and checkpoints while AI coding agents (Cursor, Claude Code, Codex, Aider, etc.) modify your repository.
Coding Agent Harness Comparison 2026: Claude Code, Codex, Amp (techstackups.com via hn) Coding Agent Harness Comparison 2026: Claude Code, Codex, Amp, OpenCode, Gemini CLI, Pi, Command Code, Factory, and Aider In 2023, there was one serious terminal coding agent: Aider. By May 2026, there are at least nine, representing every…
Benchmarking Local LLM/Harness Combinations (neuralnoise.com via hn) I’ve been running a small benchmark, harness-bench , that pairs local LLMs (served via llama.cpp ’s llama-server ) with agent harnesses (Aider, Claude Code, OpenCode, Pi, Qwen CLI) on 16 software-engineering tasks across Python, PyTorch, J…
Lessons from building a coding agent for 8k context windows: token budgeting, parallel executors, and per-file isolation (www.reddit.com) Most AI coding tools (Cursor, Aider, Claude Code) assume you have a 200k-token model. If you're running local LLMs through Ollama or LM Studio, or hitting free-tier cloud APIs like Groq or OpenRouter, you've got around 8k tokens to work wi…
Show HN: An agent that remembers across sessions (no chat history) (github.com via hn) Hi HN — I built this in my off-hours over the last 3 months. Sharing now because I just filed the provisional patent yesterday (US 64/050,345) and the repo is freshly public.
Little-coder: A coding agent optimized to smaller LLMs (github.com via hn) little-coder A coding agent tuned for small local models, built on top of pi. The research story behind all this — why scaffold–model fit matters, how a 9.7 B Qwen beat frontier entries on Aider Polyglot, and what the load-bearing mechanis…
Has anyone managed to use gemma 4 e4b in Open Code/other agentic TUIs? (www.reddit.com) Hi everyone, as a power user I hit Claude Code's usage cap too often I wanted to set up my own local model, however I only have RTX 5070 with 12 GB of VRAM so the only realistic option was Gemma 4 with effective 4B params. When I tried to…
I traced the agentic calls. Here's where the token consumption comes from (www.reddit.com via reddit) I expected an agentic coding assistant to use more tokens than a simpler tool. I didn't expect the difference to be this large.
How to replicate Claude Code's agentic workflow without the crazy rate limits? (www.reddit.com via reddit) It’s frustrating to see usage get choked down this hard, especially given the company plans to IPO soon. I am on a pro plan that on an avg gave 3-4 hrs a day which worked for me until recently.
This one helped me a lot, so if anyone needs it (www.reddit.com via reddit) on devtoolsniff.com they give your AI IDE a verify.sh + one rule and it stops declaring broken code "done" for example: verify.sh: #!/usr/bin/env bash pnpm typecheck && pnpm lint && pnpm test --run rule in .cursorrules / CLAUDE.md: "After…
Benchmarked the free API tiers you can point a coding agent at - half of them now want a card (www.reddit.com via reddit) I run aider and Cline against free tiers instead of paying per token, and my setup broke twice this month when model IDs disappeared under me. So I stopped guessing and measured what's actually left.
I’ve written software for about 30 years. I've been a heavy coding agent user for the past 1+ year. What practical coding-agent questions can I help answer? (www.reddit.com via reddit) I've been mostly hands on coding professionally for 20+ years. I have taken time in between to lead teams, run product management or run enterprise pre-sales.
Agents Don't Paginate: First-Chunk Selection for LLM Tool Responses (arxiv.org) Coding agents built on large language models (LLMs), such as Claude Code, Cursor, OpenAI Codex, GitHub Copilot, and Aider, receive tool responses that routinely exceed the agent's per-turn token budget. The standard remedy, pagination, is…
Trying to run Claude Code / coding agents for free: tried proxy failovers and self-hosting, but hit walls. How are you accessing frontier Claude models for free? (www.reddit.com via reddit) Hey everyone, I’ve been trying to set up a reliable workflow to run terminal coding agents (like Claude Code and Aider) for my development projects without running into hard blocks. Here is what I’ve tested so far: OmniRoute / Multi-Provid…
I kept losing track of my own agent backlog, so I built a git-native orchestrator and ended up adding full local-model support along the way (www.reddit.com via reddit) Hey everyone, TL;DR: Built a git-native orchestrator for my coding-agent backlog to safely run parallel agents overnight. It isolates every ticket in its own git worktree so they can't step on each other, and it pools local models (Ollama…
Les barrières de claude... (www.reddit.com via reddit) Je suis pas dev, ni informaticien. juste je m'y interresse un peu.
Qwen 3.8 27B Aider score (www.reddit.com via reddit) I ran the Aider benchmark on Qwen 3.8 27B FP8 with FP8 KV cache 256K context vLLM. The score: 72.9 This matches Gemini 2.5 Pro from 2025-04-12 which also scored 72.9.
After 18 months with Claude, sometimes I forget just how incredibly cool being able to build purpose built tools just for me is (www.reddit.comhttps) At this point, I fear I sometimes take for granted just how incredibly powerful Claude Code is. I've been using AI for development work since the first month Aider dropped back in 2023, I wrote early MCP support in Roo Code (RIP) that rema…
Echec du démarrage de l'espace de travail de Claude (www.reddit.com via reddit) Coucou, j'ai ce petit message depuis hier : Échec du démarrage de l’espace de travail de Claude.VM connection timeout after 60 secondsRedémarrer Claude ou votre ordinateur résout parfois ce problème. Si cela persiste, vous pouvez réinstall…
sandbox-cli is now in public beta 🚀 (www.reddit.com via reddit) Run Claude Code, Codex, Gemini, Cursor, Aider and 10+ other coding agents with full autonomy — inside a disposable Docker container. Only your project is mounted.
Hit your usage limit mid-session? I built a free CLI that moves your session between Claude Code and 8 other agents (open source, Rust) (www.reddit.com via reddit) You're deep in a session with Codex or opencode, and boom: usage limit. You wait, or you start over in Claude Code from scratch.
Managing parallel agent sessions without context-switching collisions (www.reddit.com via reddit) I've been experimenting with running multiple parallel Claude Code and Aider sessions on the same codebase, but I ran into a major headache: agents rewriting the same files and colliding. I solved this by setting up a local task board wher…
What's the best harness for Fable 5? (one premium model + cheap frontier models doing the rest) (www.reddit.com via reddit) Sharing my own model-routing setup and the conclusions I've reached so far, because I've been tuning this for a while and want to pressure-test my reasoning against people who've done the same. My constraint: Opus and Sonnet 5 are my workh…
I code from my phone now: Claude Code runs on a VPS, one command sets it up (www.reddit.com via reddit) For the cost of a coffee a month I run Claude Code on a small cloud box instead of my laptop. I close the laptop and it keeps working.
If you run coding agents unattended or in parallel, how do you verify the run actually worked? (www.reddit.com via reddit) I run a lot of agent loops (Claude Code / Codex / aider), sometimes overnight or several at once. My recurring headache: when I come back, I can't quickly tell whether a run actually did the right thing, quietly broke/regressed something,…
Building an open source context management layer for coding agents — looking for honest feedback (www.reddit.com via reddit) If you've used Cursor, Aider, or Claude Code on a long session you know the problem — context either bloats with irrelevant history or gets silently truncated at the worst moment. Building a Python library that gives you precise, explicit…
agentsweep: a CLI that finds & redacts the secrets your AI coding agent saved to disk in plaintext (www.reddit.com via reddit) Every time you paste an API key, DB URL, .env file, or (worst case) a crypto wallet seed phrase into Claude Code, it gets written to a local history file in plaintext. And it doesn't just sit there — these agents re-read their own history…
Does Cursor ever create more cleanup than expected because the task was too loosely scoped? (www.reddit.com via reddit) The AI literally deleted everything on my computer and I was left staring at a frozen screen (www.reddit.com via reddit) This isn’t a horror story. It actually happened to me.I was using an AI agent to automate some tasks.
Grok Build CLI, agents-cli, and the CLI coding tool gold rush (www.reddit.com) xAI dropped Grok Build CLI. Google has agents-cli.
Deepseek tui alternatives, when do you jump from single model terminal agents (www.reddit.com) Been using Deepseek-Tui for days. solid for v4 workflows.
I wanted to know small local LLM code and made a personal projects. (www.reddit.com) I Started toying around LLM about sometimes ago. I think Qwen3.5 came out after like a month.
Here is what most people get wrong about saving tokens with AST tools (www.reddit.com) I spent the last day benchmarking codebase context tools against a real AI agent. Not synthetic token counts.
Agentic Guardrails: 4 markdown workflows to improve the output quality of AI coding agents (github.com via reddit) Agentic Guardrails Reusable workflow templates that keep AI coding agents from shipping sloppy code. These are markdown-based instructions that any AI coding agent can follow — Cursor, Claude Code, opencode, Aider, Gemini CLI, or anything…