<55
Early signals, worth watching.
Research intelligence prototype
Atlas Signal
A research paper library app designed for opportunity hunters, people who track emerging trends across science and technology to spot where the next opportunities are forming.
Why I built this
I started reading more research papers while redesigning games at work, but extracting the useful ideas from them took a lot of effort. I built Atlas Signal to explore how research could be distilled into something easier to scan, while also surfacing signals and potential opportunities I might want to investigate further.
What it is
Atlas Signal is a research paper library I use to make sense of the papers I come across, styled like a newsletter or intelligence brief.
Each paper is scored and filtered so I can quickly judge what's worth reading, while surfacing signals and potential opportunities worth investigating further.
Key concepts
Signal scoring. Every paper gets a score from 0–100 indicating how relevant or impactful it might be:
Signal system
0–100 relevance index
One score, four review states
The tier color, chip, range, and meter all follow the same signal cascade used across the Atlas interface.
55+
Gaining traction across the feed.
70+
Well-evidenced and actionable.
85+
High-confidence, act on this.
Collections. Papers are organised by domain: Energy & Materials, Health & Bio, AI & Compute, Human Interface, Workplace & Cognition. Each collection works like a folder you can filter by.
Brief view. Each paper has a structured brief with an abstract summary, top application areas with sub-scores, and key takeaways, distilled so you don't need to read the full paper to understand its relevance.
Application threads. Papers are mapped to real-world industries and use cases (e.g. "Grid transmission 96 · MRI imaging 91") so the bridge from research to opportunity is explicit.
Who it's for
I built this for myself. Staying current across multiple domains of AI without reading hundreds of papers is a real problem I had, and the usual workaround of skimming abstracts and conclusions felt like it was losing too much along the way. Atlas Signal is my attempt at something better: get a sharp enough overview to know if a paper is worth your time, then go deep when it is.
This version is focused on CHI research papers in Explainability AI (XAI) and human-computer interaction. It's become a personal reading tracker for the work I'm doing on AI transparency and interpretability, which is where most of my thinking lives right now.
Design System
Approach
Built with Claude and Codex, then shaped by my own design taste. The intention was for it to feel like an academic paper, structured and credible, but easier to digest before committing to the full research.
Visual systemThe palette is warm paper tones with a terracotta accent. IBM Plex runs across all type roles. Tier colours cascade from a single CSS variable pair, so one class modifier updates the signal pill, chip, sidebar, and row header at once.
ReferenceAll sixteen components are extracted into a standalone reference page: live HTML with real data, annotated with the class names and the reasoning behind each decision.