I don't think this is controversial at all. Many of our users are the exact same way. For them, the value add can be a few different things: (1) searching across multiple accounts on the same platform, (2) better search than existing platforms (here’s an example of Needl compared to Gmail: https://imgur.com/a/3ZX8f8Q), and (3) general speed improvements.
Yes! We integrate with Local, though iCloud is much more locked down. We can index local files that are then offloaded to iCloud, but can't directly index iCloud.
iMessage is something we've explored, also locked down like iCloud with no API but there are some workarounds by using what's locally stored. No plans for a mobile app, though or web app works surprisingly well on mobile https://app.needl.tech/.
If more stuff comes up as you're using Needl, would love more of your thoughts and feedback. Can always reach me at [email protected]
Thanks for sharing -- makes a lot of sense. Indexing browser history is very high on our priority list, lets us cover the less-widely used apps much easier. (though is somewhat limited to things you've seen while the integration is set up, similarly to Rewind)
Biggest difference is search quality. Our entire product is focused on making information more accessible across web apps. We have generative Q&A search and semantic search, something Slapdash does not offer (to my understanding).
On the other side, we also don't offer the Command Bar-esque features that Slapdash and Raycast offer. For now, our #1 focus is search and search alone.
We've experimented with this a fair bit, very low occurrence of the model making up facts when the answer exists within a document, but users have reported outside information (general knowledge) being shown in their results.
We essentially just prompt GPT-3 to ignore everything that is outside of the chunks of information we provide it.
Are there plans to back up the "suggested answer", which I presume is LLM generated, by a definitive source? The first question in the demo returned the relevant document you were looking for, but I didn't see this in the search results for the second question.
I'm not sure I would trust a system like this unless I could click through and see the source of the answer I'm reading, and make sure that the LLM is referencing the correct email/document.
This seems to be a common growing pain in places where an AI model is expected to provide authoritative answers - I wonder if (at least in your case) it's possible to use a more traditional fuzzy search algorithm to attempt to find the source, based on the LLM's answer string.
Currently something we're working on with prompt engineering, but love your suggested approach. We'll definitely look into that more -- thanks for sharing.
For now, the search is always generated from the first 5 or so results. So you always have an idea of where it's coming from.
Good point. It's interesting because, in many ways, search is easier at the enterprise level than the individual level. The usage patterns & social graphs of large orgs is incredibly helpful for relevance ranking. For example, if someone on your team just accessed a doc, the odds you are looking for that doc are pretty high.
Our thesis is that Google Desktop + Cue and the dozens of others lacked was good enough search to be sticky. We're not entirely certain this is the only reason, but we're testing it out.
Understand where you're coming from a privacy perspective. Our encryption operates at the file system/storage device layer level so it doesn't affect our index at all other than a small I/O performance hit when storing and accessing data.
W.r.t. retention policies, information is immediately deleted upon account deletion. We also delete all data after account inactivity for 2 months.
Definitely exploring messaging platforms, interestingly enough iMessage (despite not being available via API) can be indexed locally on Macs. Just not something that's on the top of our backlog. Thanks for your thoughts!
Yes. We are SOC II certified and have strict limitations on who can access that data and under what circumstances (only if given explicit permission by the user). Definitely understand that could be a deal breaker for some, but with current methods, we believe that this approach will allow us to deliver more powerful search and greater value in the long run.
To the first point -- certainly some friction here. In most cases the employee is able to integrate the app and flags arise to IT after the fact. Then the employee connects us with IT and we resolve the issue by going through a security questionnaire / showing our SOC II. The majority of our current users are SMBs, where they don't have significant IT oversight, we're one of the few products in the space that's truly self-serve.
To the second -- very good point and something we continue to experiment with. Our basic thesis is -- if your search is really good, people will learn to default to your app. For many of our users, the unification is less so the value and more so the better search quality, especially compared to Gmail and GDrive. Another good point that writing things as a question is often a worse user experience than just using Google search esque terminology. For our Ask Needl feature, you can still get value without writing an entire question. Ex: website total budget --> $1000-1500.
Our YC journey was certainly formative. Much of the advice given was counter-intuitive in the bull market of Nov 21 when everyone was spending and growing as fast as possible. My biggest takeaway was the focus on solving hard problems with small teams and only caring about what your users think of you. Would highly recommend - the culture is unparalleled.
Certainly a very lively space, feels like just about everyone is trying to solve the problem at hand. Always happy to share thoughts / learnings in the space -- we should find a time to connect, feel free to email me at [email protected]
