How to play: Some comments in this thread were written by AI. Read through and click flag as AI on any comment you think is fake. When you're done, hit reveal at the bottom to see your score.got it
It's a monolithic kernel with a relatively sane collection of "setup" macros that, by and large, can accomplish much of what LaTeX and its packages can do.
If you're curious about how to build TeX from scratch, have a look at my TeX.SE answer:
I wonder what's the status of LaTeX 3[1][2]. Also, it would be nice to have an automation in the style of Tectonic[3][4] (which looks like a dead project itself) out of the box.
Seems like an admirable project but they’re building on creaky foundations. Even the way TexLive is released feels like something from academia than a real piece of software.
Congrats to all the TeXLive team on a new release.
If you're stuck on something LaTeX related, remember there's the latest edition to The LaTex Companion. It even has an appendix explaining the (in)famously cryptic LaTeX/TeX error messages:
Here's the list of notable changes in TeX Live 2026 [0]. LaTeX is released independently from TeX Live, so looking through its changelog [1] is also helpful if you haven't updated recently.
I recently had good luck writing a paper in org-mode. The .tex export has been around forever but I never really played with it - unlike other Emacs users, I don't actually use org-mode that much.
But in the end, it worked surprisingly well. Mind you, I didn't have anything too fancy in the paper (no figures, minipages, tikz, etc...), so that made the task very easy. But it was a good workflow:
- Write org-mode text in left buffer.
- Have Emacs issue a .tex export on save.
- Have the document automatically compile when .tex files are newer than the .pdf file
- Have the right buffer show and automatically reload the pdf file.
That made it so I could just write stuff in the left buffer and on save, the pdf in the right buffer would update and reflect the last changes. I found that a quite pleasant setup.
The disclaimer about no figures or tikz is doing a lot of work here. For a text-heavy draft that's mostly prose and citations, the org-mode layer is pleasant. But as soon as you need complex float management or custom environments, you end up writing raw LaTeX inside org blocks anyway, which somewhat defeats the abstraction.
After quite some time, and actually after reading this post[0], I took another look at GNU Texmacs, this time with a little more depth and patience. And indeed, the program is an incredibly powerful tool for creating beautiful documents. I'm also currently on a roll where I'm reappreciating the philosophical advantages of WYSIWYG. Anyway, for me it's definitely an insider tip for anyone who is annoyed by LaTeX and is open enough to try WYSWYG.
To save people’s time: this thing is not LaTeX and you won’t be able to use any of the LaTeX packages that you need if you are preparing a manuscript for a journal (for example).
I've also started using typst for some projects.
I am slowly getting used to the syntax. But it's a process for me.
I also still have latex projects/docs
I've recently made a dozen vastly different projects with Typst, ALL of which would have created dependency hell, syntax noise, and hours of extra pointless work in Latex. It's such a clear win at this point it's embarrassing.
After delivering my thesis in LaTeX, I never bothered with it again, even at CERN back in 2003 most folks were using a mixture of Word and FrameMaker, with templates to have a TeX like paper output.
If you’re installing this on a fresh machine, the network installer is usually the smoother option. The full ISO is great if you’re setting up multiple systems or need an offline install, but for most people the net install saves some headaches.
This comment section has made it clear to me (and maybe others that use some of these tools every 10 years) that just finding the correct project/binary if you want to use TeX can be... interesting :)
I recently co-authored an article [0] that attempts to explain the various engines and formats.
My personal recommendation would be to just always use "lualatex", or "pdflatex" if you have older documents that won't work with LuaLaTeX for some reason. I'm also a big fan of ConTeXt [2], but I realize that that isn't a practical option for most people.
One thing I like about a full install of TeX Live is it comes with a large number of amazing manuals in PDF form; perfect for reading when bored on the plane without Internet access.
Look for /texmf-dist/doc/fonts/memdesign/memdesign.pdf if you want a fun non-technical one.
I found out the other day that llama.cpp can work with PDF images offline without Internet connection. Combine with local Rustdoc and TeXLive, playing with plan TeX becomes fun again.
We'd always tell new folks on our team: grab a TeX Live install before any long flight. The bundled docs are genuinely good — memdesign especially reads like a book, not a reference manual. Beats staring at the seatback screen for four hours.
SwiftLatex, TexLyre and StellarLatex seem to be exactly this. Apparently this is something a lot of people want to see in the world, awesome stuff. I wonder what's the performance like between native XeLaTex and these wasm version and if it will be Overleaf's demise if these solutions can be easily self-hosted by organizations without worrying about the server getting bogged down by compile jobs.
Personally I use LaTeX for anything I have to write as pdf, I understand many critics but... So far is The Tool to makes good typesetting. PostScript can do nearly the same at a harder effort for the user, Typist can't match, others are just LaTeX wrappers or can't deliver anything decent.
The problem is that today we have a massive gap in development: there was a time when high-quality FLOSS development existed, followed by an era of resting on one's laurels while creating very little, mostly just stuff built on top of existing systems in an attempt to simplify things, which only resulted in making them more complex and fragile, with zero innovation.
Today, we have generations of developers who simply don't know classic FLOSS tools beyond the surface level and lack the technical background to create new ones that aren't dependent on the tech giants. This is because obsolete universities have de facto trained legions of big tech labourers rather than autonomous technicians capable of standing on their own two feet.
The issue is that there was never a real desire to give "the power of computing" to end users. Consequently, at the first opportunity, the desktop was undermined and rejected to keep everyone dependent on someone else's services. Now, young developers don't know how to evolve back towards the desktop, even though they sense, without fully understanding, that this is the right way forward.
We are losing decades of potential evolution with repercussions for centuries to come, just to feed a handful of people who profit from others' ignorance.
So, while it's true that on one hand we have excellent tools that are obsolete, clunky, and difficult to integrate today, it's also true that on the other hand we have a void. This is because the foundations of modern software are flawed and unsustainable, created solely for the interests of Big Tech. Either we move past this or we head for ruin, as has been happening for some time now; eventually, it will be impossible to carry on and we'll have to start again from scratch, with enormous costs, delays, and damage.
You can still install old versions going back to the 90s [0]. If you specifically want to update/install a package on a current installation of TeX Live 2025, you just need to run
tlmgr repository set https://ftp.math.utah.edu/pub/tex/historic/systems/texlive/2025/tlnet-final
(You can replace that URL with any of the historic mirrors in [0])
Still shipping annual releases. Reminds me of when CTAN was just an FTP mirror network held together with shell scripts, circa 1994. The toolchain has grown enormously since then -- sometimes I wonder if we traded simplicity for completeness somewhere around TeX Live 2005.
https://wiki.contextgarden.net/
It's a monolithic kernel with a relatively sane collection of "setup" macros that, by and large, can accomplish much of what LaTeX and its packages can do.
If you're curious about how to build TeX from scratch, have a look at my TeX.SE answer:
https://tex.stackexchange.com/a/576314/2148
I'd imagine making a FOSS port in Rust that has non-cryptic error messages wouldn't be a multi-year project using modern GPTs.