Writing from
Purplelink.

The Purplelink stats endpoint read event blobs one at a time, and the code even included a comment noting it would fail at higher volume. The failure arrived through latency rather than traffic spikes.

Many journal submission portals require a diff PDF alongside a revised manuscript. Here is how a researcher uses Purplelink's LaTeX Diff tool to produce one without a local LaTeX installation.

The Purplelink cover letter generator takes only your abstract, not the full manuscript. Here's the capability and privacy reasoning behind that design choice.

Two Purplelink tools that compute entirely in the browser had no usage data, because instrumentation relied on intercepting backend calls. Fixing that required defining what a run actually means, and the answer differed by tool.

Double-blind peer review requires more than removing your name from the title page. Grant numbers, IRB protocols, citation prose, and named software all identify you to a reviewer who knows your field.

OpenAlex stores paper abstracts as inverted indexes rather than strings. Building the Daily Digest's paper fetcher surfaced this, and the reconstruction has limits.

Paper Review scans your manuscript before submission with four adversarial passes: figure integrity, citation cross-check, a multi-persona reviewer debate, and a prioritized checklist. Here is what using it looks like from a researcher's side.

The Purplelink Daily Digest harvests 400-800 items daily and picks 20 to publish. Here's why the design uses a single AI curation call instead of scoring items one at a time.

Academic researchers who work across multiple fast-moving fields have no good daily briefing built for them. Here's why the existing infrastructure falls short, and what I built to fix it.

What caught my eye today at the intersection of cybersecurity, AI, research, and entrepreneurship. Curated every morning.

Most BibTeX errors don't come from researchers making mistakes. They come from the bibliography managers and export tools researchers trust to get it right.

PDF text extraction breaks on tables and equations. Here's how a researcher uses Purplelink's PDF to Structured Data tool to get clean, LLM-ready text out of an academic paper.

Compile to PDF, convert to and from Word, diff two versions, validate and build BibTeX, and render equations - no account, no install, nothing stored. They're live now.

Every travel tracker on the App Store solves a piece of the problem. Most miss the same things. Here's what we built differently - and why.

Your health data is the most personal data you have. Here's why we decided early on that none of it would leave your phone - and what that decision cost us.

TeXShop is from 2000. Overleaf is a web app. Neither was designed around the manuscript workflow researchers actually live in. ModernTex is.

Why I formed a software LLC, what I'm building with it, and the philosophy behind making software that's meant to last - not just ship.