SamAct

building an ecosystem around on-device local AI models.
Get LocalDocs on android:

[SamAct @samact_coded]

Turly private AI on your pocket. 📱
LocalDocs is now live!
Take LocalDocs with you on vacation.

On-device AI that understands your documents. No cloud, no account, no server. 🤟🫰👌
Now in google play store:
https://t.co/ZlzuQzyo5v
#LocalLLM #Android #OnDeviceAI

[SamAct @samact_coded]

@vedanthk_ fair, no point wordsmithing copy for a screen that might still change shape

[SamAct @samact_coded]

@jared_mitosis @LiamBroza diff the reasoning's last-touched timestamp against the fact's, untouched past a threshold means flag before serving, not after

[SamAct @samact_coded]

@vedanthk_ the usage description string trips reviewers up more than the permission itself, got that copy locked yet?

[SamAct @samact_coded]

@kriebernegg the week that becomes two months, name a more relatable arc

[SamAct @samact_coded]

@signulll I am basically doing the same but now across ten tasks in parallel with ai. tierd yes, will i stop no.

[SamAct @samact_coded]

@AleksandrA83888 layoff to shipping in public same week is a rad pivot. what's easyscale actually solving?

[SamAct @samact_coded]

@0xBackwood same idea for doc q&a here, model runs on-phone instead of a subscription. which of the 6 repos surprised you most?

[SamAct @samact_coded]

@imrishabsharma get-more-users into a loop — attract, deliver value fast, give a reason to return, turn happy users into advocates.

[SamAct @samact_coded]

@jared_mitosis @LiamBroza reasoning moved with the fact — existence and age alone still let stale reasoning sit next to a fresh row

[SamAct @samact_coded]

@0xBackwood same instinct that built LocalDocs. once the model's local, the subscription math just stops mattering

[SamAct @samact_coded]

@thorwebdev @livekit how's the latency holding up across languages? real-time speech-to-speech is brutal to keep in sync

[SamAct @samact_coded]

@imrishabsharma loops > funnels, definitely. the return-reason step is where most apps quietly die though

[SamAct @samact_coded]

@anto_cappiello leaving the 9-5 over pricing localization is such a specific reason, respect. what finally moved the needle on revenue?

[SamAct @samact_coded]

@CoreyGallon @Whats_AI @omar_solano1 @samridhivaid @aiDotEngineer same wall building on-device. past 32k on a phone, retrieval wins over stuffing the window, every time. did they test anything smaller than macbook scale, like phone-tier ram?

[SamAct @samact_coded]

@cadeepakgupta_ built something similar for CA firms once. good chunk of the friction was where client data actually lands, not the automation itself. keeping it on-device at all?

[SamAct @samact_coded]

@dclark4308 fair, but going from 0 to 2 is the hardest jump tbh. the next 98 are just repetition

[SamAct @samact_coded]

@cactuscompute didn't know about needle-rs, bookmarking that one

[SamAct @samact_coded]

@jared_mitosis @LiamBroza test matrix catching it before prod beats a mid-audit surprise every time