Blink-Edit – Cursor-style next-edit predictions for Neovim (local LLMs)
Hey HN, we built a pure-Lua Neovim plugin that brings Cursor-style next-edit predictions to Neovim, running entirely on local models. Tab to accept, Esc to reject. GitHub: https://github.com/BlinkResearchLabs/blink-edit.nvim We're Blink Research Labs (https://github.com/BlinkResearchLabs) - an open research collective building AI coding tools in the open. Our philosophy is simple: if it's not open, it's not research; if it's not fast, it's not usable. We think the best AI coding tools shouldn't be locked behind $20/month subscriptions or closed-source…
What it does
In the maker’s words, at launch
Hey HN, we built a pure-Lua Neovim plugin that brings Cursor-style next-edit predictions to Neovim, running entirely on local models. Tab to accept, Esc to reject. GitHub: https://github.com/BlinkResearchLabs/blink-edit.nvim We're Blink Research Labs (https://github.com/BlinkResearchLabs) - an open research collective building AI coding tools in the open. Our philosophy is simple: if it's not open, it's not research; if it's not fast, it's not usable. We think the best AI coding tools shouldn't be locked behind $20/month subscriptions or closed-source walls. We saw the Sweep model release a few days ago and realized the Neovim ecosystem deserved a complete, local-first, AI coding solution. So we built one. What makes this different: - Pure Lua, no external dependencies — No servers, no Node, no Python. Just Lua talking directly to your model backend. This matters when you're waiting for predictions on every keystroke. - Multiple providers — Built-in support for Sweep (1.5B, optimized for next-edit) and Zeta (7B, from Zed Industries). Adding a new provider is ~50 lines of Lua. - LSP-aware context — We fetch definitions and references for the symbol under your cursor and include them in the prompt. The model knows what foo() does before suggesting changes to it. - Backend-agnostic — Works with llama.cpp, Ollama, vLLM, or any OpenAI-compatible server. Bring your own inference. The plugin sends context-aware prompts based on your cursor position, recent edits, and (optionally) LSP symbols. Predictions render as ghost text inline. We handle all the edge cases: blink.cmp/nvim-cmp menu conflicts, debouncing, streaming, health checks. Getting started takes 30 seconds: llama-server -hf sweepai/sweep-next-edit-1.5b-GGUF --port 8000 require("blink-edit").setup({ llm = { provider = "sweep", backend = "openai", url = "http://localhost:8000" } }) The Sweep 1.5B model runs at 200+ tok/s on M-series Macs and fits comfortably on a 4GB GPU. For those with more VRAM, Zeta (7B) gives noticeably better predictions. This is alpha software - we're iterating fast and want feedback. If you're a Neovim user who's been jealous of Cursor's tab-completion, give this a shot and tell us what breaks.
Does the same job
all alternatives →
Locofy: design-to-code agentsJun 2026 · ▲112Agentic frontend layer between Figma and Cursor & Claude
- SSSpaceNeovim – Spacemacs for Neovim2016 · github.com · ▲105
- LALunarVim – An opinionated, extensible, and fast IDE layer for Neovim2021 · lunarvim.org · ▲93
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
Launched alongside, January 2026
the whole month →- IN
Hey HN! I wanted to share something I built over the last few weeks: isometric.nyc is a massive isometric pixel art map of NYC, built with nano banana and coding agents. I didn't write a single line of code. Of course no-code doesn't mean no-engineering. This project took a lot more manual labor than I'd hoped! I wrote a deep dive on the workflow and some thoughts about the future of AI coding and creativity: http://cannoneyed.com/projects/isometric-nyc
AI · Jan 2026 · cannoneyed.com
Automatic AI-powered code reviews the moment you open a PR
Dev tools · Jan 2026 · kilo.ai
