The Breaking Point
It was 2 AM on a Tuesday when Dr. James Martinez hit his limit.
His 287-page physics thesis was due to his committee in 72 hours. He'd just made a small change to Chapter 4—fixed a typo in an equation—and clicked compile on Overleaf. The progress bar appeared.
And stayed there.
23 minutes later, the compile timed out. Again. For the fourth time that night.
"I knew my thesis was complex—lots of Feynman diagrams, TikZ figures, a 400-entry bibliography," James recalls. "But I'd been dealing with slow compiles for months. That night, with the deadline looming, I finally snapped."
The Problem: Death by Compile Time
James's thesis wasn't unusually large for a theoretical physics PhD. But it had accumulated complexity over four years:
- 287 pages across 8 chapters
- 47 TikZ figures including complex Feynman diagrams
- 12 full-page equations with aligned derivations
- 412 citations in the bibliography
- Custom notation macros used throughout
On Overleaf's free tier, compilation took 8-12 minutes on a good day. During peak hours—which seemed to coincide exactly with when James needed to work—it stretched to 20+ minutes. Timeouts became routine.
The math was brutal:
- Average compiles per writing session: 15-20
- Average compile time during crunch: 15 minutes
- Time lost to compilation per session: 4-5 hours
"I was spending more time watching progress bars than actually writing," James says. "And every timeout meant losing my train of thought completely."
The Search for Solutions
James tried the obvious fixes:
Splitting the document — He already used \include for chapters, but Overleaf still had to process everything.
Draft mode — Helped somewhat, but he needed to see the actual figures to verify his work.
Upgrading to Overleaf Pro — The 4-minute compile limit was better, but still painful with complex documents.
Local LaTeX installation — He tried TeXLive on his laptop, but configuration was a nightmare. Different machines had different results. Collaboration with his advisor became impossible.
Then a labmate mentioned local-first editing.
The Switch: Skepticism to Salvation
"I was skeptical," James admits. "How could compiling in a browser be faster than Overleaf's servers?"
The answer: no network latency, no queue times, no shared resources.
After exporting his thesis from Overleaf (a 2-minute process) and importing it into a local-first editor, James ran his first compile.
47 seconds.
He ran it again, thinking something was wrong.
43 seconds.
"I literally laughed out loud," he says. "My roommate thought I'd lost my mind."
The Numbers Don't Lie
James tracked his productivity for two weeks before and after the switch:
| Metric | Before (Overleaf) | After (Local-First) | Improvement | |--------|-------------------|---------------------|-------------| | Average compile time | 14.2 minutes | 48 seconds | 94% faster | | Compiles per session | 8 (limited by time) | 25+ | 3x more iterations | | Timeout failures | 4.2 per week | 0 | 100% eliminated | | Productive writing time | 2.1 hours/session | 4.8 hours/session | 129% increase | | Time to submission | On track to miss | Submitted 2 days early | Saved his deadline |
But the quantitative improvements only tell part of the story.
The Qualitative Transformation
Recovered Flow State
"The biggest change wasn't the compile time itself—it was my mental state," James explains. "When a compile takes 15 minutes, you can't just wait. You context-switch to email, Twitter, whatever. Then the compile finishes, you've forgotten what you were doing, and it takes another 10 minutes to get back into it."
With sub-minute compiles, James could stay in flow:
- Make a change
- See the result
- Adjust if needed
- Keep writing
"For the first time in months, I was actually writing for hours at a stretch instead of in fragmented 20-minute chunks."
Fearless Experimentation
Long compile times had made James conservative. Try a different figure layout? That's 15 minutes to see if it works. Experiment with equation formatting? Another 15 minutes.
"I'd been avoiding improvements because the cost of experimentation was too high," he says. "Once compiles were fast, I could actually try things. My thesis got significantly better because I wasn't afraid to iterate."
Offline Independence
The week before submission, James's apartment building had an internet outage. Pre-switch, this would have been catastrophic.
"I just... kept working," he says. "Didn't even notice for the first hour. When I realized the internet was out, I panicked for a second, then remembered—everything was local. I worked the whole day without a problem."
The Technical Setup
For researchers facing similar challenges, here's what James's final workflow looked like:
Document Structure
thesis/
├── main.tex # Master document
├── preamble.tex # All packages and macros
├── chapters/
│ ├── 01-introduction.tex
│ ├── 02-background.tex
│ ├── ...
│ └── 08-conclusion.tex
├── figures/
│ ├── chapter-03/
│ └── chapter-05/
├── references.bib
└── appendices/
└── derivations.texKey Optimizations
External TikZ figures — Complex Feynman diagrams were pre-compiled to PDF and included as images, cutting compile time for unchanged figures.
Selective compilation — Using \includeonly{chapters/04-methodology} to compile only the chapter being edited.
Cached bibliography — BibTeX only re-run when citations changed.
Collaboration with Advisor
"My advisor still preferred Overleaf for reviewing," James notes. "I'd push my changes to a GitHub repo, which synced with a shared Overleaf project. She could comment there, I'd pull the feedback and continue locally. Best of both worlds."
Lessons Learned
Looking back, James has advice for other thesis writers:
1. Don't Suffer in Silence
"I thought slow compiles were just part of thesis writing. Everyone complained about it. I didn't realize there were better options until I was desperate."
2. The Setup Time Is Worth It
"Switching editors took me about 2 hours—exporting, importing, verifying everything worked. That's nothing compared to the hundreds of hours I lost to slow compiles over my PhD."
3. Local-First Isn't Just About Speed
"The speed is what sold me, but the privacy and offline capability turned out to be equally valuable. My thesis had unpublished results. Knowing it never left my machine was reassuring."
4. Keep Your Options Open
"I still use Overleaf for smaller papers where the convenience outweighs the speed issues. The right tool depends on the project. But for anything substantial, local-first is now my default."
The Outcome
James submitted his thesis two days early—a first in his department's recent memory. His defense went smoothly. He's now a postdoc at a national lab, still using the same workflow for his papers.
"That 2 AM crisis was the best thing that happened to my PhD," he laughs. "It forced me to find a better way. I just wish I'd found it sooner."
Dr. James Martinez completed his PhD in theoretical physics at Stanford University in 2024. His thesis, "Quantum Field Theoretic Approaches to Strongly Correlated Systems," was 287 pages with 412 citations. He now researches quantum computing applications at a national laboratory.
Are compile times slowing down your research? Try Thetapad free — local compilation means your 200-page thesis compiles in under a minute, every time.