Founder · Atlas
Jet New
I’m the founder of Atlas, an AI-powered knowledge workspace for visual, source-grounded research. I trained as an AI researcher at NUS (Turing Research Programme), then worked as a data scientist at Indeed and director at AgentScale AI before starting Atlas. I design and run the benchmarks on this blog and personally evaluate every tool we compare.
What I work on
Atlas turns scattered PDFs, web clippings, and notes into a connected knowledge workspace with cited answers and auto-built mind maps. I lead product, write the blog, and run head-to-head comparisons against competitors. Every claim about Atlas on this site is grounded in measurable behavior, not marketing.
Testing & benchmarks I’ve published
- Best Document AI Tools (2026), Hands-on test of 6 document-AI tools on 100+ research papers.
- Best AI Research Assistants (2026), Tested 7 research assistants across discovery, extraction, synthesis.
- NotebookLM vs Claude Projects, Side-by-side workflow test on real research scenarios.
- Atlas 2026 PDF AI Benchmark, First-party measurement protocol for PDF/research-AI tools.
How I test
For every comparison, I run each tool on the same corpus (typically academic PDFs spanning humanities and STEM), score across fixed axes (citation accuracy, answer correctness, source coverage, latency, price-per-query), and lock the rubric before testing. Atlas is included with full disclosure and ranks where the data places it on each axis, the rubric is fixed before testing and not weighted to favour Atlas. The full methodology lives at /research/2026-pdf-ai-benchmark.
Map your next paper with Atlas.
Understand deeper. Think clearer. Explore further.
