Topic
13 articles found
In April 2026 Google shipped Gemma 4, a multimodal model with a native audio path. I wanted to add it to Parlotype, my .NET 10 dictation app, as a second speech engine alongside Whisper. Four runtime paths got cut before I landed on llama.cpp's llama-server as a child process. This post walks throug
This is a submission for the Gemma 4 Challenge: Build with Gemma 4 TL;DR: Vitreus is a spreadsheet agent that lets you ask natural-language questions of CSV and XLSX workbooks, then uses Gemma 4 to return an auditable JSON action manifest. The manifest can highlight rows, write values, and add formu
This is a submission for the Gemma 4 Challenge: Write About Gemma 4 When local AI conversations happen online, they tend to sound like this: "I ran the 70B model on my dual-GPU workstation." or "You only need 64GB RAM and a 24GB graphics card." Meanwhile, I'm sitting with an Intel i5, 16GB RAM, inte
This is a submission for the Gemma 4 Challenge: Build with Gemma 4 scribe-check is a local-first command-line tool that reads a Markdown article and a folder of source documents, and reports every concrete claim in the article that isn't corroborated by the sources you handed it. It checks five cate
This is a submission for the Gemma 4 Challenge: Write About Gemma 4 Most developers still think AI limitations are mainly about intelligence. They are not. The real bottleneck has quietly been memory. Not RAM. Context. For years, using AI coding assistants felt like working with a brilliant engineer
This is a submission for the Gemma 4 Challenge: Write About Gemma 4 The easiest mistake to make with a capable local model is to treat the model call as the whole application. I almost made that mistake while building with Gemma 4 E2B. My project was a local dementia-care assistant called RememberMe
This is a submission for the Gemma 4 Challenge: Build with Gemma 4 I Built a local-first AI CCTV assistant Using Gemma 4, Frigate, and Home Assistant Most CCTV notifications are noisy and not actually useful. Typical camera alerts are mostly: Motion detected at Front Door That tells me something hap
This is a submission for the Gemma 4 Challenge: Bondmap is a social relationship network where anyone The core idea: you add people and define how you know them. The problem it solves: Most people have no clear picture Key Features: Interactive D3.js network map with color-coded connections (purple
This is a submission for the Gemma 4 Challenge: Build with Gemma 4 CureNet AI is an ABDM-native, offline-first Health Intelligence platform built to unify fragmented medical records securely under the Ayushman Bharat Digital Mission (ABDM) and FHIR R4 standards. In rural India, reliable internet is
This is a submission for the Gemma 4 Challenge: Build with Gemma 4 Inspiration Legal documents are everywhere—job contracts, rental agreements, terms & conditions—but most people don’t fully understand what they’re signing. Legal language is often complex, intimidating, and inaccessible to non-exper
This is a submission for the Gemma 4 Challenge: Write About Gemma 4 There is a district in Uttar Pradesh called Sonbhadra. It sits in the southernmost corner of the state, surrounded by forests and hills. I am a student at IIT Jodhpur. Sonbhadra is where I come from. Every time I go back home, I car
This is a submission for the Gemma 4 Challenge: Build with Gemma 4 Guild is a free, locally-running autonomous coding agent that works while you're away and backs off when you're present. The problem: cloud-based AI coding agents (Copilot, Cursor, Claude Code) require paid APIs, constant babysitting
This is a submission for the Gemma 4 Challenge: Write About Gemma 4 Most coverage of Gemma 4's multimodal capabilities stops at images. That's understandable — image input is the most obvious thing to demo. But Gemma 4 E2B and E4B ship with two more input modalities that are genuinely novel for a lo
