Cursor@Home

How to make your own cursor - ChatpGPT

hal

see also

  • How to code Claude Code in 200 lines of code
  • La bulle IA est monstrueuse et cela va tout péter !!
    • on paye 5x a 10x plus chere l’IA sur le cloud
    • les bulles explosent toujours plus fort et plus tard qu’anticipé.
    • l’IA ce n’est pas que des LLM sur le cloud
    • renforcer la resilience des entreprise grace a l’IA est une vraie valeur ajoutée
    • l’automatisation par l’IA n’est pas forcement meilleurs que via les RPA
    • mais cela ne marche que sur des solutions complexes (qui sinon ne sont pas rentable)
    • ex le passage des avions a helices au jet n’a pas été juste un changement de moteur (repensé le process: vol pressurisé en haute altitude / controleur aerien / securité).
    • la detection de fraude est usage pertinent
  • 10000 Euros par mois grace à l’IA ! Oui et alors ? - a propos de N8N
    • le marché de l’automation est déjà saturé et surévalué
    • on ne sait pas remplacer talent humain par de l’IA
    • on sait faire que les mauvais deviennent moyen.

# Skills

# grill-me ⮺

---
name: grill-me
description: Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".
---

Interview me relentlessly about every aspect of this plan until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies between decisions one-by-one. For each question, provide your recommended answer.

Ask the questions one at a time.

If a question can be answered by exploring the codebase, explore the codebase instead.

# Price 💲

Coding LLM Prices Comparison: My 5 Takeaways , per prompt:

Model Rank Cost
Sonnet 4.6 15 $0.47
Composer 2.5 17 $0.18
deepseek-V4-Pro 9 $0.11
Qwen 3.6 Plus 6 $0.07

# VSCode

Cline vs Continue

Aspect Cline Continue
Core idea Autonomous AI agent AI coding copilot
Autonomy Very high Low–medium
Can run terminal commands ✅ Yes (with approval) ❌ No
Can edit many files at once ✅ Yes ⚠️ Limited
Can create/delete files ✅ Yes ⚠️ Mostly edits
Interaction style Task-driven agent Inline assistant
Risk level Higher (powerful) Lower (controlled)
Best for Big refactors, automation Day-to-day coding help

# continue.dev ⮺

Using:

# Cline ⮺ / github / 👤

Cline is an open source AI coding agent that brings frontier AI models directly to your IDE. Unlike autocomplete tools, Cline is a true coding agent that can understand entire codebases, plan complex changes, and execute multi-step tasks.

see Cline vs Cursor

Long session are coslty because of the increased input token.

Rank Model Leaderboard ⮺   Plan Act Price Comment
★★★★☆ deepseek/deepseek-v4-pro 2026-05-17   Context: 1.0MInput: $0.43/MOutput: $0.87/M cheap
? deepseek/deepseek-flash       cheapest (free)
  z-ai/glm-4.6 2026-04-03      
  moonshotai/kimi-k2.6 2026-05-17   Context: 262K Input: $0.73/M Output: $3.49/M  
  x-ai/grok-code-fast-1 2026-04-03      
☆☆☆☆☆ Gemma4 12B Coder 2026-04-03     16GB Local LLM setup

# Setup

  • Native tools:
  • Background edit: enabled

# Terminal

  • background execution mode (avoid messing & be stuck within VSCode)

# CLI ⮺

prerequesite: needs Node 20+, see isntallation Getting Started

cline

  • config - Opens an interactive configuration view
$ npm install -g cline
$ cline auth

# pi.dev ⮺

An open-source, terminal-based AI coding agent designed around a minimalist philosophy - (github)[https://github.com/earendil-works/pi#pi-agent-harness-mono-repo] / HN.

Pi provides:

  • AI coding assistance directly in the terminal: it can read files, edit files, write code, and run shell commands inside a project.
  • Broad model support: it connects to 15+ providers and hundreds of models, including OpenAI, Anthropic, Google, local/self-hosted models, and others.

see also

# Hermes ⮺ 🔱

6 Hermes Agent use cases

  • /goal
  • Kanban

# Project 💭

Run it on raspberry-pi

# Dashboard

Running as service
# /etc/systemd/system/hermes-dashboard.service
[Unit]
Description=Hermes Dashboard
After=network.target

[Service]
Type=simple
User=YOUR_USERNAME
WorkingDirectory=/home/YOUR_USERNAME
ExecStart=/usr/local/bin/hermes dashboard --no-open
Restart=always
RestartSec=5

[Install]
WantedBy=multi-user.target
# setup
$ sudo systemctl daemon-reload
$ sudo systemctl enable hermes-dashboard
$ sudo systemctl start hermes-dashboard

# test
$ sudo systemctl status hermes-dashboard
$ journalctl -u hermes-dashboard -f

see also

# Claude

Did you know that it is named Claude in reference to Claude Shanon the man that defined anthropie of information content (and the company is named antropic…)

see also

  • Claude Code Is Steganographically Marking Requests / HN - claude code todays date differently depending on local configuration to tweek behavior depending on users
    • Claude Code now allows Anthropic to remotely inject system prompts
    • To summarize what they’ve already been doing:
      • filtering out people from the wrong side of “all humanity”, years before it was demanded by the government
      • downgrading their models in arbitrary ways (later saying “sorry but not really”)
      • actively sabotaging the replies, as in covertly modifying them to feed the users incorrect results
  • Claude Conway - instead of a normal chatbot that waits for you to type something, Conway would act more like a persistent background assistant that keeps running and can continue working over time. supposed to spy on you;

# Codex ⮺

# Alternatives

# Model provider

# Deepseek ⮺ 👤

Written on December 17, 2025, Last update on April 18, 2026
LLM agentic-AI at_home vscode project