Gadeon vs LinkLab

Gadeon runs Qwen3.5 and QwenPaw hybrid language models entirely on your device's Apple Neural Engine, using Core ML as its only framework dependency. There's no C, no Python, and no cloud calls: every token is generated locally, keeping your conversations private. Under the hood, these models mix Gated DeltaNet recurrent layers with a smaller number of attention layers. Gadeon compiles that hybrid architecture directly onto the Neural Engine's dataflow accelerator, so both prompt processing (prefill) and token generation (decode) run on-chip instead of falling back to the CPU or GPU. On an Apple Silicon Mac, prefill measures roughly 2 to 2.4 times faster and about 4.4 times more energy-efficient per token than a Metal GPU baseline, while leaving the GPU free for other work. Conversations in Gadeon are append-only: each turn extends the model's fixed-size recurrent state instead of re-processing the entire chat history, so response latency doesn't grow as a conversation gets longer. The app can also roll back to an earlier point in a conversation, dropping reasoning traces or replaying tool calls in a clean form without reprocessing everything from scratch. For ternary-quantized models, Gadeon switches to a pure-Swift Metal GPU backend behind the same session and tokenizer interface, so both compute paths feel identical from the chat UI. Gadeon is a proof of concept for bringing Gated DeltaNet hybrid models to the Neural Engine through Core ML - built to explore what's possible on Apple's own silicon.

Links end up everywhere. A page you meant to read, sitting in a browser tab. A booking confirmation buried in an email. A recipe someone sent you in a group chat. A form your bank told you to fill in later. They are scattered across a dozen apps, and by the time you want one back, you can't remember which app it was in. LinkLab is one place for all of them. Share a link into LinkLab from whatever app you're in, give it a name you'll actually recognise, and it's there when you need it — on your device, offline, in a list you can search. SAVE A LINK FROM ANY APP, IN TWO TAPS Hit Share in Safari, Mail, Messages, or anywhere else a link turns up, pick LinkLab, and the save sheet opens right there. Add a label, drop it in a category, done — no app switching, no copying and pasting a URL into a note you'll lose. FIND IT AGAIN IN SECONDS Everything you save lands in one list, newest first, searchable by name or by URL. Group links into categories you name yourself — Recipes, Work, To Read, Bills, whatever fits — and filter the list down to one of them with a tap. Save the same link twice and LinkLab updates the entry you already have instead of piling up duplicates. A QR CODE FOR ANY LINK Generate a QR code for any saved link with one tap. Share it, or save it to Photos. It's the quickest way to get a link onto someone else's phone — the check-in page, the shared album, the menu — without spelling out a URL or making them type anything. OPEN ANYTHING, NOT JUST WEB PAGES Tap a saved link and LinkLab hands it to whatever app on your device can open it — a web page to your browser, an address to Maps, an email to Mail. It handles custom app links too, the myapp:// kind that Safari won't touch, which makes it a genuinely useful scratchpad if you build or test apps for a living. If nothing on the device can open a link, LinkLab says so plainly instead of doing nothing. RUNS ENTIRELY ON YOUR DEVICE LinkLab has no network code at all. No server, no account, no sign-up, no analytics, no third-party SDK — nothing your links could be sent to, because there is nowhere to send them. QR codes are generated on your device. Photos access is add-only and requested only at the moment you choose to save an image. For anyone whose links are scattered across too many apps, and for the developers and testers who have been using it as a deep link scratchpad since day one.