Logical (YC F25): a local-first proactive desktop AI copilot
Hey HN! My co-founder and I have been building Logical, a proactive desktop AI copilot that watches what you're doing (locally), understands the context of your workflow, and surfaces helpful actions before you prompt it. - Quick demo: https://www.loom.com/share/090a065315934aa7b36a7676f9394d1f - Try Logical: https://trylogical.ai/signup Logical lives on your desktop, infers what you're trying to do across apps – email, meetings, documents, PDFs, terminals – and: - Gives you a reply suggestion when you hit "Reply" on an email thread - Offers to "Check…
What it does
In the maker’s words, at launch
Hey HN! My co-founder and I have been building Logical, a proactive desktop AI copilot that watches what you're doing (locally), understands the context of your workflow, and surfaces helpful actions before you prompt it. - Quick demo: https://www.loom.com/share/090a065315934aa7b36a7676f9394d1f - Try Logical: https://trylogical.ai/signup Logical lives on your desktop, infers what you're trying to do across apps – email, meetings, documents, PDFs, terminals – and: - Gives you a reply suggestion when you hit "Reply" on an email thread - Offers to "Check schedule" when you open a message asking for a quick chat - Automatically extracts to-dos during calls and from pretty much anywhere on your screen (and reminds you to follow-up) - Suggests a formula in Excel as you work that you can apply with one click - Explains terms of research papers as you highlight them No prompting. No switching context. No copying text around. * Why we built this * Despite big progress in LLMs, the dominant UX is still: User does work –> realizes AI could help –> stops –> writes a prompt. But your computer already has the context of what you're doing. It knows what window you're in, what text you're reading, which script just errored, and what meeting you're sitting in. We wanted an AI that uses this ambient context to proactively assist – more like a real teammate than a chatbot. * Privacy and data handling (something we deeply care about) * Right now: - We offer a technical guarantee that no user data ever touches Logical servers. - Context is sanitized locally (our local pipeline strips PII / sensitive text before anything is sent off). Long-term, we aim to move everything on-device as small language models and consumer AI chips mature. We've seen interest from founders, researchers, engineers, and privacy-sensitive users who want AI benefits without cloud exposure. * What's under the hood * - A context engine that digests signals and user data from apps (both local, and if you choose, cloud-based services). - A sanitization pipeline that removes identifiable or sensitive details before model usage. - A local vector store + lightweight knowledge graph for immediate retrieval. - An intent engine that infers "what you're trying to do" in real time and surfaces actions at the right moment. * What's next * - Windows support. Logical is currently Mac only. - Letting developers plug into the context engine and intent engine to offer richer experiences on their apps. At least until desktop MCP is good enough. - Fine-tuned integrations with more apps and workflows. Would love your feedback: If you're interested in: proactive AI; OS-level context awareness; on-device AI; privacy-preserving AI; building AI that actually reduces friction instead of adding more prompts Happy to chat in the comments! [email protected] is always open for feedback.
Does the same job
all alternatives →
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, November 2025
the whole month →
- IB
Life & fun · Nov 2025 · bitsnpieces.dev
- BBoing▲782
Life & fun · Nov 2025 · boing.greg.technology