The last few months have been rather quiet in this forum and thus I am wondering whether people have moved to other venues to get assistance with their LAMMPS problems or if there is something else that needs to be done.
I cannot imagine that suddenly people, especially those that are new to LAMMPS, don’t run into the same problems that they have for the last 20+ years, I also don’t believe that we suddenly have improved the LAMMPS documentation or that the LAMMPS code itself has in a short period become so much improved that people can resolve their problems on their own. Or that suddenly people research old discussions thoroughly before posting their questions. Thus I have two questions:
Is there some other venue that you use to get help with your problems? If yes, what is it and what makes it a better source of assistance than the forum?
Is there something that the LAMMPS developers or the forum member frequently responding here could do to provide better help?
Please don’t worry about political correctness or being critical of somebody or unsupportive of LAMMPS. The more honest feedback we get on this, the better. Please recall that feedback is just a summary of your personal assessment.
It may also be that this time period coincides with teaching duties. That probably isn’t the case in America but for me personally (Australia) I’ve been teaching the last two months, and while it hasn’t stopped me too much from being active here, it does mean I haven’t done much research.
I wonder if activity similarly dropped over this period last year.
According to the metrics provided by the Discourse website in the admin interface it has not.
Last year in the period May/June/July/August 2024 we had a steady average of about 200 topics created and about 900 posts every month that dipped only in December 2024. This has declined to about 150 topics and 650 posts in May 2025 and further to 85 topics and 450 posts in August 2025.
Documentation is improving all the time, the recent improvements to Lammps tutorials with @simongravelle are very good, and the comprehensive testing/bug fixing is reaching a level of software engineering excellence that proactively resolves issues before a user has a problem that needs assistance on this forum.
Also if you search this forum it’s rare you don’t find someone else had the same issue before.
@alphataubio This sounds great, but these are changes that happened (and still happen) gradually over the last 5+ years and thus I am not convinced they can explain a drop in activity to less than 50% over the course of a year with most of it happening in the last three months.
@jtclemm I believe that Stack Overflow is much more susceptible for being replaced by AI due to the way it is organized and how responses are ranked. So far, I have not seen anything where AI generated input decks were sufficiently well set up for anything that is not straightforward to do and for which sufficient public tutorials exist. The publicly available material, like this forum, is full of examples that do not work and very few that explain how the issue was resolved.
The presentation at the last workshop is an example. There are regularly input decks posted, where I suspect they are AI generated, and the fact that people don’t respond to questions about them make it quite likely that this is true since the people cannot answer the questions in that case.
@srtee activity is still lower than usual. We need about 50% more. Specifically we have a decline in first time posters (which require approval as an anti-spam measure) which is what made me think that people go elsewhere.
I don’t know if this information is available (or could be shared), but I wonder if the number of posters (and first time posters) categorized by country (determined by e.g. IP address) reveals anything.
I don’t have access to that information, but with many years of responding to mailing list and forum messages, you can somewhat guess based on the time when the message is posted and the name/username. Of course, one has to factor in that science is international and - especially in the US - there are many foreign nationals (like me). On the other hand, I have had exchanges with researchers from all around the world through my participation in events at the International Centre for Theoretical Physics for the last 20 years, so you get some “conditioning” what country or region people could be from.
That said, my overall impression is that the distribution has not changed much, and it feels like it somewhat matches the distribution of remote participants during the recent LAMMPS tutorials during the LAMMPS workshops. Outside of researchers in the US, the dominant groups seem to be from India and neighboring countries. China not as much since there are active LAMMPS forums in Chinese (they sometimes contain references to the LAMMPS online docs, so I see those our the web server logs). There seems to be a bit of a decline of posts from the middle east, but that may be related to political troubles and blocked internet access. For example, countries like Iran and Turkey traditional have very strong research and that includes simulations with software like LAMMPS.
On the other hand, there seems to be an uptick in LAMMPS users from Africa (both North and Sub-Saharan). These trends are quite gradual and the decline I see is across the board, as far as I can tell.
Since AI has been ruled out as a potential parallel source of information, could it be that there are less people posting because there is less research going on in the US since the funding has been decreased?
I guess the US is/was a big source of research (%, compared to other countries), so it would justify a big decrease in the public here. And I guess most people that post are beginners who have not yet learnt to look up for previous posts and etc, so it helps support the theory.
LLM tools help to narrow search spectrum in Lammps documentation, the documentation is generally too extensive to be studied fully, so I can find my answer way faster than before in official documentation.
second, LLM gives (false/correct) summary of already discussed threads in Lammps forum, so I can find who and when has asked a similar question, if user is active with a real name, I directly contact, so user gives advice to other user.
Third, Lammps developers require more detail than LLM to answer the same question, particularly if it is on source code error.
So personally, I am interested to learn what will be the new functionalities of Lammps in near future, and this type of question can’t be answered by LLM, the expert answer is reliable.
While there is no question that responses from LLMs are more accessible (because this is what they are programmed to do), they are also often littered by subtle mistakes or sometimes just plain wrong. That is because LLMs are programmed to synthesize a text that looks like it was a correct answer, but there is currently no way known to me to have that response properly checked to be correct. Documentation is sometimes complex and convoluted because the problem is complex and convoluted.
One of the main motivations of having a forum is that discussions remain “in the open” and thus people can later learn from the discussion of others. Your approach is bypassing that and you are making the forum less useful by discussing offline. We switched from a mailing list to the forum exactly because it was too simple to just reply to an email and not reply to the list as well.
How can a response be reliable, if it is based incomplete information? You will get an answer from the LLM with less input, but it can be completely misleading. The only information a LLM has available is what was asked and answered previously. What if the code in question has changed? What if the problem is different but just looks similar?
What is particularly annoying in this case are situations where an expert first has to argue and explain why the LLM generated answer is incorrect before one can begin to resolve the actual problem. This is very exhausting and irritating and ultimately wasting resources.
This is a new topic and should this be posted as such and not appended to an ongoing discussion.
That could be a compelling explanation, if you can also explain how this would explain a decrease in traffic across the board and not only from within the US.
Also, based on personal experience, I expect the impact of the funding cuts and stop-work-orders to become more visible in the coming months. People have contracts and not all of them can be easily canceled and thus it will take time until the full effect materializes.
