Pensieve: An intelligent repository for the million things in your head(pensieve.ai)
pensieve.ai
Pensieve: An intelligent repository for the million things in your head
http://www.pensieve.ai/
17 comments
Hi Colin, I'm sorry to hear you feel that way. We're working hard to make Pensieve a reality, but we need to know that we're building the right thing. If you really are excited by the potential of an idea like this, we'd love to hear from you about how you would use it and what features it would need to be most valuable for you.
Where will the data live? If it's outside my servers I can't use it.
How will it decide what's important? We have so many things that automated systems have declared unimportant and yet which turn out to be critical.
Currently commercial spam systems are running at around a 10% false positive rate. Quite simply, we can't rely on them. Why would your system be any different about deciding what's important and what isn't? Critical information is often not recognised as such until significantly after the fact, and very well may be in a throw-away comment.
So we have a system that crawls our email archives, chat logs, and other data sources, and uses an automated word2vec-like and tf–idf system to find similarities. Then we have a browser that makes the database look like a wiki - you just click on links till you find what you want, usually just one or two clicks away.
But I have no idea what you're building. The description is of something magical that will solve all your problems, but it doesn't say anything about what it actually does. The web site is slick, but it makes it look like you have something ready to go. Then the rug is pulled out from under me and it's clear that while you might have something in development, you don't have a product.
Why not be honest? Say up front you're working on this and offer decent inducements to get quality feedback on people's concerns, what they use, what their current system gets right, their pain-points, and what they'd like to be able to do but currently can't.
This may be unjust, but the site feels dishonest. I know it's the current trend popularised by some book or other to pretend to have a product and see how many bites you get in order to gauge your potential market, but it puts me completely off. I now don't feel like I can trust you.
To address your actual questions:
Would I use it? No idea - you haven't told me why I would. What features would be valuable? No idea, I don't know what it might be able to do.
What do I want? A machine into which I type a vague query, it asks for clarifications, then gives me a definitive answer, or a definitive "That's not in my records."
So there you are, 400 words, 10 minutes of my time. Even though I feel cheated by your web site I hope you find that useful, and wish you the best of luck.
How will it decide what's important? We have so many things that automated systems have declared unimportant and yet which turn out to be critical.
Currently commercial spam systems are running at around a 10% false positive rate. Quite simply, we can't rely on them. Why would your system be any different about deciding what's important and what isn't? Critical information is often not recognised as such until significantly after the fact, and very well may be in a throw-away comment.
So we have a system that crawls our email archives, chat logs, and other data sources, and uses an automated word2vec-like and tf–idf system to find similarities. Then we have a browser that makes the database look like a wiki - you just click on links till you find what you want, usually just one or two clicks away.
But I have no idea what you're building. The description is of something magical that will solve all your problems, but it doesn't say anything about what it actually does. The web site is slick, but it makes it look like you have something ready to go. Then the rug is pulled out from under me and it's clear that while you might have something in development, you don't have a product.
Why not be honest? Say up front you're working on this and offer decent inducements to get quality feedback on people's concerns, what they use, what their current system gets right, their pain-points, and what they'd like to be able to do but currently can't.
This may be unjust, but the site feels dishonest. I know it's the current trend popularised by some book or other to pretend to have a product and see how many bites you get in order to gauge your potential market, but it puts me completely off. I now don't feel like I can trust you.
To address your actual questions:
Would I use it? No idea - you haven't told me why I would. What features would be valuable? No idea, I don't know what it might be able to do.
What do I want? A machine into which I type a vague query, it asks for clarifications, then gives me a definitive answer, or a definitive "That's not in my records."
So there you are, 400 words, 10 minutes of my time. Even though I feel cheated by your web site I hope you find that useful, and wish you the best of luck.
Really appreciate you taking the time to write this down. To be fair, we found someone who thinks about this problem as much as we do. So maybe the books weren't wrong after all ;) It sounds like you've built something inhouse and are using it. Would you be open to us emailing you to learn more about it?
Feel free to email me and I'll tell you what I can, but my time is severely limited, and my colleagues are totally paranoid about talking about anything that's in-house.
But you can always ask.
But you can always ask.
This is a neat idea but unless I can operate my own private version then it's not really useful.
I've got a private wiki that provides essentially the same function. It's not as convenient to add new notes, but that convenience comes at the cost of organizational sanity.
I tried using Evernote for this purpose and what I discovered was unless you can control it yourself then it sucks. Eventually they're going to have money problems or they'll get DDoSed or whatever and suddenly this vital resource you've learned to rely on is not available when you need it.
You might as well use notepad & nextcloud.
There are only 2 things an app like this can offer to make it 'special' compared to the thousands of other available methods. Searchable tags, and a convenient UI.
I've got a private wiki that provides essentially the same function. It's not as convenient to add new notes, but that convenience comes at the cost of organizational sanity.
I tried using Evernote for this purpose and what I discovered was unless you can control it yourself then it sucks. Eventually they're going to have money problems or they'll get DDoSed or whatever and suddenly this vital resource you've learned to rely on is not available when you need it.
You might as well use notepad & nextcloud.
There are only 2 things an app like this can offer to make it 'special' compared to the thousands of other available methods. Searchable tags, and a convenient UI.
Thanks for the feedback. We're definitely thinking about tags - even beyond just making them searchable. A lot of people run into issues like forgetting what tag they used and having a user-feedback and a UserFeedback. Also issues when a project name changes and so tags need to be migrated. Do you have that problem?
What do you mean by private version? Do you mean running it on-premise?
Yeah I would rather have a php based webapp that I can download and install on my own server, than any sort of SaaS thing. I hate SaaS, I think it's a kind of gamble, I am gambling that the people providing that service will be successful and operational until I quit using it.
But in the real world that isn't always the case.
However with a localized application that I can download, install, and run without any reliance on "the mothership", if the developers should go belly up (obviously nobody wants that), I can at least continue to use the software for a length of time while searching for a suitable replacement.
That's just one example. There's other things, such as data security, short term outages, and so on. In general SaaS has always been a terrible idea. It's good for corporations, but it sucks for users, and computing is about what's good for the users.
But in the real world that isn't always the case.
However with a localized application that I can download, install, and run without any reliance on "the mothership", if the developers should go belly up (obviously nobody wants that), I can at least continue to use the software for a length of time while searching for a suitable replacement.
That's just one example. There's other things, such as data security, short term outages, and so on. In general SaaS has always been a terrible idea. It's good for corporations, but it sucks for users, and computing is about what's good for the users.
Your signup form is behaving strangely. Seems like a state management issue. Text input gets narrow onmousedown.
I currently use a git repository for this purpose. It would be nice to have a better interface for it.
Interesting. How do you use it? Like a key-value store?
I love that line 'Conversations that begin with “Got a minute?” are almost never that short.' Use it so often! :D
Oh ... vapourware ...
How I hate these posts that pretend that this thing exists and is wonderful and tells you everything it can do and how it will solve all your problems, and only then tells you to sign up so you can get told when the miracle occurs.
Closed. Good luck, hope it works out for you.