joshualarosa.ai
ResourcesPodcastNewsletterWork with me
← ResourcesAI Tools

Any research paper. A live app in minutes.

The exact Claude + Higgsfield App Builder system that turns any research paper into a live, shareable interactive web app — the setup, the universal prompt, and the worked example you can click right now.


Don't cite the famous paper. Rebuild it.

Every student cites Ebbinghaus. Almost nobody has rebuilt him as software. This system takes any research paper and turns it into a live, interactive web app — the famous chart recreated with the original data, plus a working tool visitors can use on themselves.

See the finished example — a 140-year-old memory experiment, now a working spaced-repetition study tool:

🧠 forgetting-curve-1885.higgsfield.app →

The worked example — Ebbinghaus (1885), live right now

The stack: Claude reads the paper and does the thinking. Higgsfield's App Builder (connected through MCP) builds and deploys to a public URL. You paste one prompt.

The setup — connect Higgsfield to Claude (100 free credits)

⚡ Sign up + get your Higgsfield MCP link →

Start here — 100 free credits for new users

Promo code: HIGGSFIELD_HQMVB9R9 — enter it when you sign up.

  • Copy the MCP link — sign up through the button and copy the connector link.
  • Open Claude's connectors — Customize → Connectors.
  • Add a custom connector — paste the link, confirm.
  • Toggle green — Claude can now build and deploy real web apps through Higgsfield.

Pick your paper — the three rules

  • It needs numbers you can plot. One strong chart beats ten weak ones.
  • Visitors should be able to run the experiment on themselves. Stroop test, reaction time, loss-aversion bets. If a visitor can generate a data point in under a minute, the app spreads.
  • Old classics are ideal. Pre-1930 = public domain, famous names, simple methods — and nobody has seen them as software.

Starter shelf: Ebbinghaus 1885 (forgetting curve → flashcards) · Stroop 1935 (interference → play-it-yourself) · Yerkes-Dodson 1908 (arousal curve → focus tracker) · serial position effect (→ try-it-live) · Kahneman & Tversky loss aversion (→ betting game).

The universal prompt — part 1: recreate the results

Two brackets decide everything: [CORE RESULT] — the one finding worth recreating — and [HANDS-ON TOOL] — a way for visitors to use the finding on themselves.

THE PROMPT · PART 1 OF 2I've attached [AUTHOR]'s [YEAR] paper "[PAPER TITLE]". Read the full paper, then use the Higgsfield MCP connector and its App Builder to create and deploy a live interactive web app that recreates [CORE RESULT — the paper's famous finding, e.g. "the forgetting curve"] and turns it into [HANDS-ON TOOL — a working tool visitors can use, e.g. "a spaced-repetition study tool"].

Workflow — use Higgsfield for everything:
1. Start by loading Higgsfield's website creation instructions so you follow their app builder flow correctly.
2. Create the app as a new Higgsfield website project named "[PROJECT-SLUG, lowercase-with-dashes]".
3. Build all features listed below, then deploy through Higgsfield and give me the live shareable URL.
4. I'll be watching in the preview — after deploying, confirm the site is live before finishing.

Part 1 — Recreate the paper's results:
- Extract the original data behind the core result from [SECTION/TABLE where the numbers live]: [LIST THE KEY VARIABLES]. If exact values aren't in the paper, use the standard published figures and say so on the page.
- Plot the result as [CHART TYPE]: x-axis "[X LABEL]", y-axis "[Y LABEL]".
- Mark and annotate each original data point on the chart with its actual value.
- Add a toggle that overlays [THE SECOND INSIGHT — e.g. "the spacing effect"].
- Include a short "About the experiment" panel explaining the method: [2–4 METHOD CONCEPTS] — in plain language a college student would understand.

Part 2: make it usable

THE PROMPT · PART 2 OF 2Part 2 — Make it a usable tool:
- Let users [CREATE/ENTER THEIR OWN DATA — flashcards, habits, guesses, reaction-time trials].
- Apply the paper's actual model to that input: [THE MECHANIC — schedule reviews on the paper's intervals, score against the published baseline, predict their result].
- After each use, plot the user's personal results on top of the original [YEAR] data.
- Show a simple dashboard: [3 LIVE STATS that matter].

Design:
- [THEME — dark/light], clean modern typography, smooth chart animations.
- Mobile-friendly, since [AUDIENCE] will open this on their phones.
- One-page layout: result visualization on top, interactive tool below.

Deliverable:
A single live, deployed Higgsfield app with a public URL I can send to anyone. All data stored locally in the browser (no login required). When the link loads, the recreated [YEAR] result should render immediately with the annotated original data points.

Paste both parts as ONE message, paper PDF attached, Higgsfield connector on. Claude reads, builds, deploys, and hands you the live URL.

The worked example — the answer key

  • Paper: Ebbinghaus, Memory (1885)
  • [CORE RESULT]: the forgetting curve · [HANDS-ON TOOL]: spaced-repetition flashcards
  • [SECOND INSIGHT]: the spacing effect · [MECHANIC]: review intervals pulled from the curve itself
  • [3 LIVE STATS]: cards due today · current retention estimate · next scheduled review
  • Result: forgetting-curve-1885.higgsfield.app

Put it on your resume

Use the classic student format — Education → Technical Skills → Experience → Projects — and write every bullet with the X-Y-Z formula: "Accomplished X (result), by doing Y (skills, actions, tools), as measured by Z (numbers, scale, users)." Your Projects entry, pre-filled:

PROJECTS ENTRY · X-Y-Z FORMAT[Your App Name] | Builder & Developer
Link: [your-app].higgsfield.app

• Recreated [author]'s [year] [core result] as a live interactive web app (X), by prompting Claude with the Higgsfield App Builder MCP to extract, model, and deploy the paper's original data (Y), producing a public tool with [N] visitors and every original data point annotated (Z).

• Designed a hands-on [tool type] applying the paper's actual model to user input (X), using [the mechanic] (Y), measured by [live stat, e.g. retention estimates per session] (Z).

Straight talk: check the app's numbers against the paper before you share it broadly — AI can misread a table, and the whole flex is that YOUR recreation is faithful. Five minutes of checking protects the credibility that makes this impressive.

Keep reading

Get the rest of “Any research paper. A live app in minutes.” + the PDF — free.

Drop your email to unlock the full guide and download it.

joshualarosa.ai

Quiet editorial · siblings with building leverage

InstagramTikTokYouTubeSubstack