Convenience vs. Openness

Open source is laden with barriers of inconvenience for everyday users, which is one of its greatest limitations. This is caused both by incentive structures (OSS developers are incentivized to be self-centered), and lack of funding (OSS developers are mostly self-motivated but often not compensated for their work). The effect of these incentive structures is that projects suffer from inadequate documentation, bloated dependencies, complicated/unfriendly interfaces, and lack of sustainability in the long-run.

• Open source software is often not as friendly as closed source software, because it is more likely to have been built by individual developers with their interests first-and-foremost, rather than the customer. It is more likely that labor in open source is put in for developers to address a need that they have, not paid to make something great for other people to use. (oss_not_friendly, 17)
• Open source software also suffers from containerization and dependency issues, as most often, it is not packaged in one executable. There are often installation steps provided in the README, but these can either be hard to follow without background knowledge, or result in a lot of unnecessary packages installed (typically with Python packages) such that it takes up unnecessary space in people’s computers. (dependency_hell_burden, 9)
• That the user of an open source software would have some technical background to use it is a a valid assumption for the developer to make. (oss_assumes_dev, 4)
• OSS lacks comprehensive documentation. From the developer’s perspective, it isn’t easy to prioritize as them knowing how to use their software may be sufficient, but from the perspective of wider adoption, or even just considering more developers being able to use and contribute to OSS, lack of documentation is an issue. (oss_lack_documentation, 4)
• Perhaps this is the most fundamental reason to explain the above, but the incentives for the unfunded (unpaid by a corporatation) open source developer is very different from someone hired to develop a product, and often times due to other commitments in life, any OSS project is subject to being sunsetted. (oss_lack_resources, 3)
• This is not only evident from the two comments under this subsubtheme but the variety of projects on GitHub that accomplish the same goal—the overwhelming number of tools that are potentially varying in contributor size, development history, and quality but serve the same purpose may be hard to navigate. (reinventing_the_wheel, 2)
• This one is advocating for people who find it difficult to (openness_learning_curve, 2)
• Open model features are hard to replicate. The implicit reason behind this may be related to oss_lack_resources, though generally speaking, it takes a lot of developer effort to make open implementations of features one sees on ChatGPT or Claude. (hard_to_replicate, 1)
• In relation to ^ossnotfriendly17, developers may prioritize the speed of getting something out the door before considering usability, whereas it is the other way around with commercial software. (prioritize_speed_usability, 1)
• Projects in OSS are prone to lacking sustainability. (long_term_sustainability, 1)

However, some believe that open alternatives can become industry standard enough to become a friendly tool to use even for everyday users.

• OSS like Linux have become industry standards in the past, and it isn’t hard to imagine open models becoming industry standard as well (open_industry_standard, 9)

Some just do not prioritize openness as highly as convenience, and even go as far as to say that openness for them means a model that is hosted online for free for everyone to use. To add onto this factor, models accessed online can be accessed on demand, whereas booting up a model locally takes longer. For them, accessibility is openness.

• There are only a select few who prioritize openness, and the vast majority of everyday users seek what is convenient. (openness_priority_minority, 3)
• Online models can be accessed on demand (online_on_demand, 2)
• “Open” means free and online. (open_llm_free, 1)

There are only two comments that discuss how one makes the tradeoff between convenience and cost, with one saying that their laziness is what prevents them from finding an alternative to ChatGPT, but could always be incentivized to explore if there is a financial incentive (i.e., ChatGPT becomes more expensive, or open models that can be run free become competitive). On the contrary, the other comment was actually pointing out that the unfriendliness of open source software is what draws people to invest money into their ChatGPT subscriptions. This may as well be in ^ossnotfriendly17, but since it’s mentioning more specifically the element of cost, I will be keeping it here. (convenience_cost_balance, 2)

Overall, the limitations of open source software seem to dominate the conversation in this subtheme, more specifically focused on tooling around LLMs and less on open models.