Socrates Reviews

I've really enjoyed being a Tier 4 subscriber to Socrates, it's powerful and works well for my document Q&A needs. It's simple to use and returns results quickly, and I like having a desktop app on Mac rather than just website access for the UI. Recommended.

Socrates is a true gem of a tool at an unbeatable price. I love that it has a native app (I use the Mac version) and that it may be used privately without internet connection. I basically always have the Deep Dive feature on as I'm typically looking for pretty in-depth information in academic/medical texts and so on. Nice that it can read ePubs among other formats, and that you can choose from various AI models. Another excellent aspect is the ability to reference multiple docs concurrently in a single chat. I haven't made a ton of use of the table output/comparison feature as of yet, but I'm glad it exists because I'll most certainly have a need in the future.

Is it a perfect app? No, but it's pretty darn great and the low cost makes it kind of a no-brainer despite a handful of drawbacks. Here's the main area I've identified where Socrates lacks and could benefit from improvement: the distinction between paragraphs, sections, and chapters tends to be a little sloppy, and although you can manually select specific paragraphs and sections in the Prompt Loops settings (not to mention pages/page ranges in the Chat settings), accurate identification of these can be really hit-or-miss. For example, the same document in ePub and PDF formats yielded wildly different results among the numbers of paragraphs, sections, and chapters. This shouldn't happen. One of those formats (can't recall which of the 2) identified only 1 section and 1 chapter for an entire ~400-pg book, whereas the other format was far more accurate, albeit far from perfect. Meanwhile, since paragraphs are apparently identified strictly by line breaks, Socrates identified a few hundred extraneous paragraphs by counting each short line of a poem, each index/table of contents item, each bibliography entry, etc. as individual prompt loop segments.

Sure, you can manually select/deselect segments, but that's a lot of work for a publication with a couple thousand segments identified. That made this feature fairly cumbersome and infeasible for use on this particular publication. For this same reason, I tend to hit the ceiling on prompt loop segments (where one has limited control over usage) well before hitting my Deep Dive Pages limit. The logic behind the way these segments are counted leaves a lot of room for improvement.

Although not insignificant, these are really my only major complaints. I find the overall analysis capabilities, precision/accuracy, and output quality to be quite excellent. I've run into a couple instances where the AI has missed certain information, and then corrected itself after further prompting, but these are fairly universal limits of modern AI in general. I've also seen the app get hung up when loading a large amount of output information in the Prompt Loops section, but it does still say that feature is in beta, and eventually is loads (I don't believe I've seen the app crash thus far). That's quite forgivable IMO. Hardly a dealbreaker -- more of a nuisance or pet peev -- I really wish the Mac desktop app remembered window size upon closing and reopening the app. I hate when apps don't, but I'm also a little dictatorial about app window management in general. In any case, should be an easy dev fix!!

One really neat feature for Tier 3 and 4 users is the ability to share chats externally with a link (visible to anyone), in essence sharing account access on a per-chat basis. Rather than share one's existing chat history (which can be exported to PDF), this feature gives other users the ability to query their own prompts in a blank new chat.

Re: the prompt loop segment limitations I discussed, it's feasible to stay within range and get plenty of analysis output if you're careful. Just takes some time to get to know the ins and outs. With that said, if there were a higher tier with higher usage limits here, I'd most definitely go for it. That is more a testament to how much I love this app than my frustration with the prompt loop segments. And who knows, maybe that'll be addressed if/when this feature comes out of beta development. Here's to hoping for that!

TLDR; buy it and try it. If you like document analysis, what's not to like about Socrates?

What i liked is the speed of data extraction from multiple pdf files with tables you can adjust the prompt immediately and table will change plus it will highlight the source of information so you can double check or modify your prompt, i updated to tier 4 i use the app for research and education. I wish the team all the best.

This is fantastic, leagues above the other offering with more VC funding. I use this for personal and professional purposes. I look fowards to the upcoming new features and find this to be one of my best purchases here. Well done dev.

This tools has been amazing for analyzing full books, the approach is unique, all the rest use vector db and come up short. Socrates with the prompt loops does an amazing job at reshaping content for different needs (like summarizing docs for study guides, or writing small lessons fro short online courses, etc...).

I can't recommend it enough! My only "wish" so far is that it would allow us to choose the model to use for output (and adjust credits accordingly, to make it sustainable). Different uses can benefit from different models