Nicheloom

Market intelligence for builders — see what's gaining traction before it's crowded.

I run a vision model on every screenshot, locally, on a 4GB GPU

Details

External ID
48522417
Source
HN
Company
—
Product
I run a vision model on every screenshot, locally, on a 4GB GPU
Website domain
github.com
Launched
June 13, 2026
Cohort
—
Upvotes
36
Upvotes percentile
0.8135245901639344
Tags
—
Fetched at
Sept. 7, 2026, 9:26 p.m.
Updated at
Sept. 7, 2026, 9:26 p.m.

Enrichment

Theme
gpu compute and acceleration tools
Vertical
Horizontal
Function
Model & infra
Audience
Developer
AI stance
AI-native
Project type
Hobby / open-source project
Normalized one-liner
vision model inference on screenshots locally
Manually corrected
False

Could you build this?

Yes A background script that captures the desktop screen on an interval and sends frames to a lightweight local vision model via Ollama or Hugging Face is readily vibe-codeable in a weekend.

Discussion

7 comments analyzed.

Competitors mentioned: Rewind app, Microsoft Recall

Concerns raised: Cloud-based versions caused backlash over privacy, Data being uploaded to servers where companies can monitor user activity, Small language models (gemma-e4b) too limited for text generation

Feature requests: Open-ended image generation capability

Competitors

Other products that read as similar to this one — 480 launches clear the similarity bar, closest 8 shown.

Attention rank: #92 of 481 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).

Launched 224 days after the earliest competitor.

Other launches for this product

Same idea, different domain

Nobody's really built a model & infra tool for Fintech yet.