ContextliContextli

Researchers ยท Experiment Run Log

Log an experiment run before the next one blurs it.

Who this is for

Researchers, grad students, and ML engineers running back-to-back experiments who need a timestamped record of exactly what they changed and saw.

The moment this saves you

I bump one variable, watch the result, and swear I'll remember, then by the third run I cannot recall whether it was the learning rate or the dropout that moved the number.

See it work

Messy spoken thought in. A clean, structured artifact out.

What you said

Run twelve, I bumped the learning rate to three e minus four and added dropout at point two, accuracy went up about two points but it's clearly overfitting after epoch twenty, the train loss keeps dropping but val flattens. Next run I want to keep the learning rate where it is but add weight decay instead of leaning on dropout.

experiment-run-log.md

Run 12

  • Changes: learning rate to 3e-4; added dropout 0.2
  • Result: +2 points accuracy; overfitting after epoch 20 (train loss falls, val flattens)
  • Interpretation: real gain, but regularization isn't holding
  • Next run: keep learning rate, swap dropout for weight decay

The workflow

1

Record a voice note

Hit the hotkey and talk, no formatting, no typing.

2

Tag it with this context

Contextli shapes your words into the structured output above.

3

Find it later

Everything's searchable and organised by context.

4

Pull it into Claude or ChatGPT

Bring your contexts straight into your AI tools with the Contextli MCP.

Your raw recording and transcription stay on your device, so you can always go back to the original.

The prompt behind this context

system prompt

I'm going to narrate one or more experiment runs out loud. Turn each into a structured log entry. For each run, output a bold Run label (use the number or name I give, otherwise "Run"), then bullets for: Changes (what I varied this run), Result (what I observed, with any numbers I said), Interpretation (only if I offered one), and Next run (the change I want to try next, if I said it). Keep my real numbers and variable names exactly. Never invent metrics, hyperparameters, or conclusions I didn't state. If I describe several runs, output one block per run in the order I said them. Output only the log, no preamble.

Make it your own. This is a starting point. Once it's in Contextli, tweak the instructions so the output comes out exactly how you like it.

Use this context

One click copies it and shows you exactly how to drop it into Contextli.

Next, open Contextli, go to the Contexts page, click Import, choose From JSON, paste, then Import Context. It is ready to use.

Make it your own. This is a starting point. Once it's in Contextli, tweak the instructions so the output comes out exactly how you like it.

Your raw recording and transcription stay on your device, so you can always go back to the original.

Related contexts

Questions people ask

Questions researchers ask about Experiment Run Log

How do I take study session notes without losing the thread of what I am learning?

The most effective approach is to take it in fully first, then speak a summary immediately after the study session ends while it is still fresh. The Experiment Run Log context structures your spoken summary into a study note with key points, questions, and takeaways. You retain more because you summarized in your own words instead of transcribing.

What is the best way to capture takeaways from a study session so I remember them later?

Speak a structured summary using the Experiment Run Log context immediately after the study session ends. The context formats your spoken words into a study note with the main ideas, anything worth keeping verbatim, and open questions. Speaking a summary in your own words is one of the most effective recall techniques, and Contextli handles the formatting so the result is readable later.

How do I take study session notes by voice without typing?

Add the Experiment Run Log context to Contextli, then speak your summary. The context produces a study note in plain text you can paste into your notes system. The recording stays on your device.

What should a study session note include to be useful later?

A study session note is most useful when it covers the source and date, the main argument or thesis, three to five key points or insights, anything worth quoting, and your own reactions or questions. The Experiment Run Log context structures your spoken debrief to capture all of these, so you do not have to remember the template while speaking.

How do I add this context to Contextli?

Copy the context on this page, then open Contextli and go to the Contexts page. Click Import, choose From JSON, paste it into the Import from Clipboard window, and click Import Context. It is ready to use in under 30 seconds. If you do not have Contextli yet, you can download it for free first.

Is my voice recording private? Does Contextli send it anywhere?

Your voice recording and the transcription are stored on your device only. Contextli processes your audio locally and does not send your recordings or transcription text to any server. The structured output it produces is text you control, and you decide where it goes.

Can I change what the output looks like?

Yes. Every context in Contextli is a starting point you can edit. Open the context in the app, change the instructions to adjust the structure, tone, or fields, and save. The next time you use it, the output reflects your changes. You are not locked into the default format.

Do I need to install an app to use this context?

Yes. Contextli is a free app. Download it, then copy this context and paste it into the Import from Clipboard window on the Contexts page. The whole process takes about 30 seconds.

Browse more

Experiment Run Log