Workspace

My Dream Research Workspace

We almost had a perfect research tool without realizing it.

Let me explain. And how I'm rebuilding it for my PhD research.

Software engineers have this tool called an IDE.

A place where they write and run code, building out the full software codebase. VSCode used to be the most popular one. When AI came along, there was a renaissance of this tool, like Cursor, and we ended up with a UI like this:

urban-heat · draft.md · Beeblio

EXPLORER

URBAN-HEAT

literature

chen2024-nature.pdf

refs.bib

data

stations.csv

figures

draft.md

notes.md

draft.mdnotes.mdrefs.bib
urban-heatdraft.md
1# Urban heat islands
2
3## 2. Results
4
5Daytime ΔT averaged **2.31 °C** across
6214 stations [@chen2024, tab. 2].
7
8| site | ΔT (°C) | p |
9|-----|---------|-------|
10| LIS-14 | 2.31 | 0.001 |
11| JKT-07 | 1.88 | 0.013 |
12
13Peaks at 15:40, see
14`figures/fig2_trend.png`.
PROBLEMS OUTPUTTERMINALPORTS

➜ urban-heat git:(main) python analysis.py

loading data/stations.csv … 214 stations

fitting diurnal model … done (r² = 0.82)

figures/fig2_trend.png

➜ urban-heat

ASSISTANT

auto
Check the ch.2 claims against chen2024-nature.pdf
read literature/chen2024-nature.pdf

3 of 5 claims check out. The mean ΔT matches Table 2, but two p-values were transposed.

+ [@chen2024] reports 2.31 °C mean UHI

+ intensity for mid-latitude cities

AcceptReject
Ask about your draft…
main* 0 0Markdown

A left panel for the file explorer

A main area for writing code

A right panel for AI interactions

This was almost a perfect UI for research, because:

urban-heat · draft.md · Beeblio

EXPLORER

URBAN-HEAT

literature

chen2024-nature.pdf

refs.bib

data

stations.csv

figures

draft.md

notes.md

draft.mdnotes.mdrefs.bib
urban-heatdraft.md
1# Urban heat islands
2
3## 2. Results
4
5Daytime ΔT averaged **2.31 °C** across
6214 stations [@chen2024, tab. 2].
7
8| site | ΔT (°C) | p |
9|-----|---------|-------|
10| LIS-14 | 2.31 | 0.001 |
11| JKT-07 | 1.88 | 0.013 |
12
13Peaks at 15:40, see
14`figures/fig2_trend.png`.
PROBLEMS OUTPUTTERMINALPORTS

➜ urban-heat git:(main) python analysis.py

loading data/stations.csv … 214 stations

fitting diurnal model … done (r² = 0.82)

figures/fig2_trend.png

➜ urban-heat

ASSISTANT

auto
Check the ch.2 claims against chen2024-nature.pdf
read literature/chen2024-nature.pdf

3 of 5 claims check out. The mean ΔT matches Table 2, but two p-values were transposed.

+ [@chen2024] reports 2.31 °C mean UHI

+ intensity for mid-latitude cities

AcceptReject
Ask about your draft…
main* 0 0Markdown

Center · Editor

At the center, the human is still the main author of the document. We researchers need to be accountable for everything we write and publish.

Left · Files

Everything is reachable on the left. A big research project has to organize a bunch of literature, data, figures, and other files.

Right · Assistant

And when we need help, the assistant is just one glance away, on the right.

Bottom · Terminal

Plus, we can run any data analysis or visualization right in the terminal, at the bottom.

But then everything changed when the Chat UI took over.

We're discarding that perfect UI in favor of chat composers everywhere:

Plan, Build, / for skills, @ for context
“As a PhD student, I dream of having a tool like that, instead of a black-box chat interface where the AI just hands me a final document.”

So I'm building exactly that dream tool, and I plan to make it open.

Here's the design.

And here's the roadmap.

I'm halfway there.

I said the IDE was almost perfect.

The IDE still isn't a rich text editor. So we're closing the gap:

A Native Word Processor

We researchers still need a rich text editor like Word or Google Docs. The IDE editor isn't one.

Citations & References

The IDE still lacks a citation system and a reference manager, built in rather than duct-taped on the side.

The best part

Cloud Storage & Compute

It lives in the cloud, so we can bring our research anywhere, without fear of losing our data.

Support & follow the journey.

The tool is built in the open, and every milestone lands in public.
Feedback is welcome. We're also exploring ways to sustain this project.