Typeahead 2.0

Private AI autocomplete for every app on your Mac

Productivity

Writing

Artificial Intelligence

Visit WebsiteSee on Product HuntTwitter ⧉

One BIG red flag is a lack of a trial! If if it's the perfect tool, I wouldn't spend a penny on something that doesn't offer a trial!

Congratulations on the launch! The "writes with you, not for you" is great message. One thing I'm curious about - how do you handle personalisation across very different writing styles? I write professionally in English but it's not my first language, so my natural voice has its own rhythm and quirks. Does Typeahead adapt to that over time, or does it tend to pull suggestions toward more "standard" phrasing?

One idea: a quick personalisation guide could be really helpful - what to feed Typeahead early on to help it learn faster and adapt to a specific style. Even a simple "start here" checklist would make the onboarding feel much more intentional. Or maybe you already have something like this?

I think this hits a nerve for me. I've been looking for something like this for some time, but nothing ever sat well.
Typeahead becomes the mini PA we all need.

The RAM and syntax questions are covered, so mine is about when it decides to suggest, not what. With inline autocomplete the misfires that actually hurt aren't wrong words, they're suggestions that pop mid-thought and pull my attention off what I was about to type. A correct completion at the wrong moment still costs me the sentence I had in my head. Does the model gate on anything like typing pace or a pause, or does it fire on every keystroke and let Tab sort it out?

Congrats on the launch. This is a good app from the utility standpoint! :)

Are the insights shareable? Maybe a share card like how Aqua Voice generates showing how much we saved and what was the typing speed will help create word of mouth.

Good stuff, Hiten! Got this from Rohan's community and definitely keen on trying.

Love the new updates. I am curious how are you able to keep it on a LTD while other similar tools are charging for a subscription?

Does Typeahead learn from our writing over time, or is the behavior fixed based on instructions?

Does Typeahead has ability to change writing styles across work emails, Slack messages, and other apps?

Hello! The app looks great. I'm very excited about it. I couldn't find any information on how it compares to Cotypist. Is it intended to do the same? Any pros over Cotypist?

running local on-device is the right call for something that sits in every text field you type in, that's exactly the kind of tool where sending keystrokes to a server would be a dealbreaker for me. how big is the local model and does it noticeably eat into battery on a laptop over a full day of typing

@samasante how does the app specific style adaptation work? Does it learn automatically from my past writing in that specific app, or do we set preferences manually?

Love that everything stays on-device. Are there any plans to support local open-source models, or is the autocomplete engine already fully self-contained?

Just bought it and it works great. Just curious: how much power is Typeahead drawing from the overall Mac battery, given the local model running, etc.? Let's say on an M5 Pro, 48 GB RAM- using the Gemma 4 model.

Congrats on 2.0, Sam! The per-app writing style switching is the standout feature for me — most autocomplete tools either annoy you in Slack or sound too casual in client emails. Solving that automatically is a smart move.

Question: with the local model running per-app context, how do you handle apps where the user pastes in a lot of external content (long threads, docs, etc.) — does it adapt on the fly or need a moment to pick up the tone shift?

One-time pricing with 8 free updates already delivered is a strong trust signal too. Rooting for you 🚀

the per-app writing style that switches automatically is the 2.0 headline — keeping Mail formal while Slack stays casual removes a whole layer of mental overhead. does it switch on app focus alone or does it read the window/thread title too?

The per-app opt-in for terminals and code editors is the detail I would actually use, since most autocomplete is all-or-nothing. Two things on the local model: roughly what size is it and what is the RAM/battery cost of keeping it resident, and in code-editor mode does it parse the buffer as code (syntax-aware) or just treat surrounding text as plain context? Trying to gauge whether it suggests something useful mid-function or only autocompletes prose.