Savindu Thathsara
Author ProfileClaim This Author Profile
Prove ownership by publishing #HashtagPLUS and this profile link on your author page or an article under your byline. A moderator or admin will review the request before it merges into your real HashtagPLUS username.
🌐 dev.toSource
Savindu Thathsara checked Hearthlen on the ThinkPad. 13.0 GB was still free. The bilingual speech model and Qwen 3B were the delay.
🌐 dev.toSource
Savindu Thathsara tried an offline AI assistant for an 8GB laptop. The 16GB ThinkPad still had 13 GB free, and a real answer took 3017 ms.
🌐 dev.toSource
Savindu Thathsara gave Hearthlen one sentence on a ThinkPad. In 48881 ms the local agent wrote a form, opened it in Chromium, and filled both fields.
🌐 dev.toSource
Savindu Thathsara asked Hearthlen to fill a form on a ThinkPad. It stopped to ask what to type, in 2486 ms, then filled both fields in 18983 ms.
🌐 dev.toSource
Savindu Thathsara lowered a clipping microphone on the ThinkPad. hey Hearthlen then woke the listener, and Win+H opened a chat that takes a command.
🌐 dev.toSource
Savindu Thathsara's RAM budget for an offline AI assistant on 8GB: a tiny always-on model, then whisper, llama.cpp, and a voice that exit when the reply is done.
🌐 dev.toSource
Savindu Thathsara explains an on-device AI assistant that learns routines from use, stores them locally, and lets you read, delete, or pause every habit.
🌐 dev.toSource
Savindu Thathsara's build-in-public roadmap for an offline AI OS: baseline the 8GB laptop, measure local models, then skills, voice, training, and safe updates.
🌐 dev.toSource
Savindu Thathsara is building a lightweight Linux OS with a local AI assistant. No cloud model API. It learns your routines and runs on 8GB of RAM.