Problem Forge
Summary: LLMs are allowed. The competitive aspects of the site have been adjusted accordingly.
LLMs are permitted. That's not quite the same thing as being encouraged. Nor should you consider this an endorsement of using LLMs in any other context. We expect that most participants will use LLMs. The suggestion for exceptional participants not using AI assistance is to focus on just the Problem of the Month or only one of the ten repeated challenges. Those people are to be feared, but please consider that almost all of the insights that an LLM might offer came from humans originally. In fact, much useful material was generated long before LLMs became practical. The challenges here, as elsewhere, are likely beyond what a human can accomplish without an LLM, programming skills, or, at the very least significant spreadsheet skills; still, here are two studies of how humans fared at solving the travelling salesperson problem without any computational assistance at all:
- J. N. MacGregor and T. Ormerod (1996). Human performance on the traveling salesman problem. Perception & Psychophysics 58(4), 527-539. doi:10.3758/BF03213088, PubMed 8934685.
- M. Dry, M. D. Lee, D. Vickers and P. Hughes (2006). Human performance on visually presented traveling salesperson problems with varying numbers of nodes. The Journal of Problem Solving 1(1). doi:10.7771/1932-6246.1004.
Now, numerous challenge designers, problem setters, and puzzle creators share cetain ideological goals. There are many websites designed or engineered to provide multiple hands-on programming exercises. Most lean toward entertainments and education. To the extent that this site fits the common pattern of offering specific challenges, the individual problem instances offered here would be considered fiendishly difficult on any other scale. The staff here finds that entertaining. We're easily amused, and the fiendishness seems appropriate since our beta version launched on Halloween (to only a handful of invited site testers). Box ticked. As far as educational content, novices may find that the other sites offer easier ways to get started or may be more suited to their learning styles. Much of what is available here should encourage autodidacts to investigate Theoretical Computer Science and Problem Complexity. These aren't typically first-year subjects, but please note that useful resources abound.
Anyway, we can put a checkmark in the educational box too. At the same time. it should be very clear that we're not hiding the NP-related difficulties. Any programming competition allowing participants to use LLMs has to find a way to level the playing field. This is ours. We've done enough testing to know what to expect from most common LLMs. So, ... How hard can this be?
Formally, solving a single fixed and finite instance of a NP-difficult challlenge is a very different proposition than resolving P versus NP. (We may be cruel, but we're not monsters.) All the engineered challenges on this site do have solutions (often unique solutions), and these will be shared when the month ends. Just revealing the fact the most of the puzzles have unique solutions is an exploitable hint, verging on being a spoiler.
All that said, a general guideline for this site (and many like it) is that you should never spend more than two hours stuck on a problem without asking for assistance. If you need help (or want to offer help), consider these forums.
The audience for this site is encouraged to explore all of these resources to see what they have to offer.
Please follow their rules. Most of those sites have expressed some very clear usage restrictions. In particular, we ask that our users respect the licenses of any software (or related material) that they might post on any external website. (Respect both the forum and the material.)
As always, your mileage may vary.
Good luck!