NotebookEntry 06
Lumen
A fluorescence learning bench. Pick a protein, a laser and an emission filter and see how much of the light you would collect, then watch photobleaching wear down the signal-to-noise ratio.
- Interactive app
- FPbase reference spectra
- Notes stay on this device
- Made with Codex
Four sections. The spectral bench compares six fluorescent proteins using twelve numerical reference spectra from FPbase. The signal lab models photobleaching and signal-to-noise for a simple camera. Reading and recall has four paper guides, six practice questions and a glossary. The notebook keeps your predictions and saved setups. It is a teaching model, not an instrument calibration; the formulas and their limits sit under each chart.
Your notebook stays on this device. Notes, saved setups and quiz progress are kept in this browser’s local storage. Nothing is uploaded, so I never see them. Each browser keeps its own notebook, and clearing this site’s data deletes it, so use Export notebook for a backup. Used Lumen as a file on your own computer before? Export the notebook there and import it here.
YOUR FLUORESCENCE LEARNING BENCH
Follow the light.
Explore how a protein, a laser, and an emission filter work together.
Change one thing.
Notice what moves.
EXCITATION → EMISSION
See the overlap.
THE SMALL PRINT MATTERS
A pocket protein library
Reported excitation/absorbance and emission peaks are shown below. Molecular brightness describes a protein under reported conditions. Cellular performance also depends on expression, maturation, environment, and your optics.
| Protein | Ex/Abs → Em (nm) | ε (M⁻¹ cm⁻¹) | Quantum yield | ε × QY / 1,000 | Sources |
|---|
What this bench calculates
PHOTONS HAVE A BUDGET
More light.
More information?
Explore expected signal, shot noise, and bleaching in a simple camera model.
Try doubling exposure.
Does SNR double too?
80 CONSECUTIVE FRAMES
Watch the signal fade.
Expected detected electrons per frame, integrated over each exposure. No gaps between frames.
Signal-to-noise ratio
Model, units & assumptions
A SHORT ROUTE INTO THE FIELD
Read a little.
See a lot more.
A 30-minute starting path, then a few questions to test your intuition.
Learn the ideas.
Bring better questions.
Four starting points
Words worth knowing
Sources, data provenance & limits
KEEP THE OBSERVATION
Your lab before
the lab.
Save comparisons, record a prediction, and bring one good question to your mentor.
Stored in this browser.
Yours to export.
Moving over from a copy of Lumen saved on your computer? That copy keeps its own notebook. Choose Export notebook there, then Import notebook here: your notes, saved setups and reading progress merge into this one.
Trash & local storage
Notes live in this browser on this device. Export a backup before clearing browser data or moving the app. Opening a different copy or address may use a separate notebook. To move a notebook, export it from the old copy and import it here; if the same note is in both, the newer version is kept.
How it was made
Lumen was made on 10 October 2026 with Codex, OpenAI’s coding agent, which downloaded the reference spectra and wrote the code and its 22 automated tests. Claude, Anthropic’s AI model, then fitted it into this site. The spectra are FPbase’s numerical reference curves, rounded to four decimal places and otherwise unchanged: no smoothing, shifting or invented curves. The sources for each protein, the download hashes and the processing notes are in the data provenance notes.
Lumen is an independent educational project. It is not a product of FPbase or UC San Diego.