In-IDE assistants
Best when the model can see your repo, run diffs and stay inside your editor flow.
- Use for refactors inside one service or module
- Keep PR-sized changes so review stays human
- Pair with tests before merging
AI with programming
AI for code you can review.
Practical programming workflows for scoped changes: set acceptance criteria, inspect the diff and run the relevant tests.
Workflows
4 common coding patterns
Chat for debugging
Fast for regex, SQL, one-off scripts and explaining unfamiliar stack traces.
- Paste minimal reproducers, not whole repos
- Ask for trade-offs, not just the first answer
- Verify against official docs
Local models for code
Useful for private snippets, air-gapped machines or low-latency autocomplete experiments.
- Start with Ollama or LM Studio on hardware you already own
- Expect weaker reasoning on large refactors
- Good for boilerplate and docstrings
CLI and agent sessions
Hand off multi-file refactors, test loops and PR prep to Claude Code, Copilot cloud agents or Codex when chat is too shallow.
- Start with permission prompts enabled
- Keep tasks PR-sized for review
- See the coding agents guide for product choice
Guides
Developer-ready tutorials
Local vs Online Agents
Choose between local AI agents and cloud agents for work, study and automation, with setup paths for Ollama, LM Studio and major cloud assistants.
Open guideAI Coding Agents
Pick one coding agent for daily work — AI-native IDE, extension, terminal agent or desktop Codex — without subscribing to every developer tool at once.
Open guidePrompting That Saves Hours
Concrete prompt patterns for meetings, reports, code review, client emails and BIM document tasks that reduce rework instead of creating more editing.
Open guideDev stack
Tools worth testing first
Coding and development
Freemium
Cursor
Edits code in-repo with context instead of copying snippets into a separate chat window.
Official sitePaid
GitHub Copilot
Autocomplete, agent mode in supported IDEs, and GitHub cloud agents for issue-to-PR workflows.
Official sitePlan-dependent
Claude Code
Terminal, IDE and Slack agent that reads your repo, runs commands and can prepare pull requests after permission prompts.
Official sitePlan-dependent
OpenAI Codex
Software-development mode in the ChatGPT desktop app for repo-aware technical tasks separate from Chat and Work.
Official siteFreemium
ChatGPT
Fast drafting for scripts, regex, SQL and one-off debugging when IDE integration is not required.
Official siteLocal AI agents
Free
Ollama
Runs open-weight models on your own PC or server with a simple CLI and local API for private chat and scripting.
Official siteFree
LM Studio
Desktop app for downloading, chatting with and serving local models without sending prompts to a cloud API.
Official siteFree / self-host
Open WebUI
Browser UI for local models with chat history, document upload and team-friendly self-hosting on top of Ollama or compatible backends.
Official siteFree
Jan
Offline-first desktop chat for local models when you want a simple app without running a server stack.
Official siteDeveloper news
Latest for builders
BotPsy: On-Device Personal Assistant with a Sub-Billion-Parameter Agentic Language Model
Open the original announcement for confirmed product details, availability and pricing.
Read Hugging Face Blogllm-gemini 0.36
Release: llm-gemini 0.36 Gemini 4 Carbon was announced two weeks ago for "a subset of cyber defenders", but today it's available to everyone else . This is Google's best ever Gemini, priced at $2/million input and $10/million output at launch (that's GPT-6.1 Sol / Claude Sonnet 5.5 prices) but scheduled to double in price to...
Read Simon WillisonThe indie game community is in disarray over AI and Big Walk
Some Independent Games Festival (IGF) jurors published a statement yesterday saying they would not vote for games that use generative AI in "any capacity." Soon after, the developer of indie hit Big Walk has published a statement saying it "sparingly" used AI in the "final stretch" of the game's development to "help work...
Read The Verge AIIntroducing Our Smallest, Most Efficient Model: Muse Ember
Introducing Our Smallest, Most Efficient Model: Muse Ember Meta AI just launched their entry in the Claude Haiku/GPT Luna category: Muse Ember, which they describe as: [...] our smallest and most efficient model designed for complex reasoning execution and the work agents repeat all day: tool calls, coding subtasks, routing,...
Read Simon WillisonQuoting Prompting Claude Fable 5.2
Claude Fable 5.2 responds closely to what the prompt says about who owns the code it works on. In Anthropic’s testing on small coding tasks, when a prompt said nothing about ownership or presented the code base as someone else’s, the model tended to add to the existing structure, even when that repeated existing code.
Read Simon WillisonKeep calm and call clone with Rust
Ryan sits down with Josh Triplett, member of the Rust Leadership Team at the Rust Project, to chat about how the borrow checker delivers compile-time memory safety without garbage collection overhead, why Rust consistently ranks as developers' most loved language by providing psychological safety for complex systems, and how...
Read Stack Overflow Blog