I remember moaning about twitter reducing the number of characters people sent, and then yo! came out, and it was down to a single bit of data. That made me think 'what would even lower information content messages look like in a social network, perhaps half a bit or even less?'. I figured that a half bit of data would be where a social network sent some sort of hello message automatically with the same frequency the human user sent one manually, so given a message, it'd be 50/50 whether it was sent automatically or manually. Although I got to it theoretically, in the end, I thought it could be pretty useful practically, as a way of sparking conversations again when you haven't spoken to someone in a while, etc.

Maybe combined with ChatGPT we could even make those automatic getting-back-in-touch messages indistinguishable from the manual ones.

> combined with ChatGPT we could even make those automatic getting-back-in-touch messages indistinguishable from the manual ones

extending this bit further - we might as well reply with chatGPT. then it's chatGPT all the way down, who needs human interaction when we tick the boxes of "I called them" and "I replied"?.. :)

This is exactly what happened in the show Silicon Valley with Dinesh and Gilfoyle's AI

In this situation, neither party need be aware that the other party is 'antisocial' in this regard. The way I see it, the following situations apply for a tool that produces communication indistinguishable from the user's own writing (AI)...

"AI" for the following cases would be beneficial:

Social -> Antisocial

Social <- Antisocial

"AI" for the following cases would be net neutral:

Antisocial -> Antisocial

Antisocial <- Antisocial

And finally "AI" would be inapplicable (since neither party would use it) in the following cases:

Social -> Social

Social <- Social

In this drastically simplified model, there aren't any cases where the existence of a sufficiently competent AI would be detrimental to any party involved, while still providing value for those who choose to use it.

Thought-provoking and somehow ingenious, even if something about this makes me deeply uncomfortable.

To try to be a little constructive, is a relationship really worth anything if a half-bit is all one can spare for it?

The half-bit you actually send means more to the recipient than the 160 bytes you don't.

I'm more unsettled by the automation part. At least the half-bit received came from someone who consciously thought about you enough to send it. Once you emulate that part, all you have is a MITM initiating conversations between strangers.

> Half a bit of data

I’ve never heard that idea before and I love it.

Half a bit... yeah. I read a good description of a bit (I think in The Information by James Gleick) as "a yes or no answer to a single unambiguous question. I'm misremembering the quote but that's close enough.

So half a bit would be like hearing "Umm..." in response to that same question?

I think "Umm..." would be zero bits, and half a bit would be more like "Probably"/"Probably not"

"Probably" is more than one bit, not less, no? It conveys both uncertainty and an opinion on the answer.

"Is this tree deciduous?" "Yes" - one bit

"Is this tree evergreen?" "Probably" - one bit (they don't know), another bit (their guess)

But I'm way off any formal understanding of this, and it can be rigorously defined.

Meta comment: "half a bit" was clearly a joke, and so was my response; now I'm taking your reply at face value and debating it seriously, while admitting that I haven't got anything close to firm enough ground under me to actually debate it :)

I meant based on definition of a bit as "a yes or no answer to a single unambiguous question".

"Umm..." definitely carries some kind of information, but it doesn't actually help answer the question.

Based on the information theoretic definition of entropy we'd need to go from 50:50 to 89% certainty to get half a bit of information, and I'd probably qualify 89% certainty as "probably"

Yo: https://en.wikipedia.org/wiki/Yo_(app)

To be pedantic, this is not a single bit of data. This is zero bits attached to an event. The event is not discrete however, so I'm not sure what form of information theory would be appropriate to describe it.

Maybe you could also "reduce" the bit in a different way, where it would be 50/50 if a message would actually be sent upon clicking the submit/reply button.

Like if you'd walk on a noisy street, see someone you know, and greet them half-heartedly and wouldn't really mind if they didn't hear you.

Oh man, the thrill of not knowing if this message will actually be sent when I hit the "reply" button!

There have been several iterations of this sort of idea — a personal CRM — and the main issue I have is they should automatically scan my emails, texts and chats to figure out my friends and suggest that I reach out.

Also, want it to be 100% local and privacy first.

> Also, want it to be 100% local and privacy first.

But I also want an AI personal assistant to know what to write to everyone, what we last did, what would they be interested in, and so on.

Something like monica may interest you to scratch the personal CRM itch.

https://github.com/monicahq/monica

I loved the idea of this until I realized the amount of PII I was collecting of friends and family in one place that could be hacked.

Yeah, I entered exactly one contact before I realized this.

There’s also the problem of entering unflattering details, and subsequently leaking them out.

The search continues.

> should automatically scan my emails, texts and chats

Great, easy to integrate one service with the APIs of a bunch of others!

> Also, want it to be 100% local and privacy first.

...but in combination? Never going to happen.

Ask yourself: who's signing up for the API keys to enable the client-side service to talk to all these services? Is it the end-user, or is it this software's developer?

If it's the software's developer, then they're effectively leaking all these API keys by embedding them into the software itself — where not just the end users, but anyone else could come along and reuse these keys for anything they like. The service providers will find this out, and block these keys. (No, you can't avoid this by proxying requests to some gateway, operated by the service-provider, that holds the API keys. Then you lose the "local/private" aspect.)

If it's each end-user, then the aggregate traffic from all the instances of this app running at once, will look exactly like a bot that's trying to evade API rate-limits using a "residential proxy cluster" like https://www.zyte.com/smart-proxy-manager/... and so the services will block these keys.

---

Mind you, in theory, you could do this on the OS level, using OS accessibility APIs to effectively "read" the messages off the screen. But 1. is there any third-party ISV — who isn't a certified accessibility-software provider — who you'd trust enough to allow their software the ability to constantly "read" everything on your screen? That includes your passwords, you know! And also, 2., the messages need to be on the screen for accessibility software to read them. An accessibility-API-driven CRM can't load your chat history unless you also grant it the ability to literally take your mouse and scroll through it for you.

Or, alternately, coming at this from the perspective of the Operating System vendor themselves, you could do this "in" the OS, by forcing emails/text/chat message handling to go through system APIs that can see these as special document types, and so do things with them. IIRC there was at least one pre-iPhone mobile OS that did this (BlackBerry OS, maybe?), enabling all of these types of messaging-app traffic to be muxed together into a single first-party app that did indeed manage all conversations with your contacts in a multi-channel way.

I recall Google has an API for at least part of the Gmail data (tasks) and that API is used on mobile devices.

Look into it, you might be surprised.

Can’t it just scan the client? For email at least, I feel it can just read your inbox or whatever and store the results of the analysis locally.

You were very active with this friend for n years and it dropped off. Drop them a line.

That presumes the client stores anything locally + in plaintext. Clients like Facebook Messenger, Slack, and Discord have no local storage other than (opaque, fragmentary, useless) cache databases, because they're just "webapps in a can"; while many of the more modern native chat clients like WhatsApp, Signal, Telegram, etc keep all their synced state encrypted-at-rest on disk.

Also, that presumes you're on desktop. On mobile, you just can't poke into another app's sandbox/container that way (unless the target app hasn't explicitly granted made some part of its container-filesystem externally accessible; and there's no reason these apps would.) And many people only have a mobile device.

If you're okay with a partial solution, a far easier one would be scraping the data out of the few clients that have web-app versions through browser extensions that read the state out of the page they're running in.

Yes, good point and I like your final partial solution. I think it’s a good compromise.

Lots of attempts over the years at this personal crm problem space but no success to date. I did one on Facebook platform in 2008-09 called Socialfly, our tagline was ‘be twice the friend in half the time’.

Problem is there’s a lot of input and upkeep which limits appeal, the most important social data sources (text & phone) are not accessible via api, its a personal tool that you don’t necessarily want to tell others you use which limits distribution & scale, and even after all that in general it’s hard to scale personal authenticity.

I think the problem needs to be approached from a different angle along the lines of a personal assistant rather that an explicit data management tool. And it likely has to come from those with access to privileged social data sources like Apple or Google or Facebook.

I remember using Socialfly! I wasn't a heavy user, but I actually quite liked it and was a bit disappointed when it went away.

Oh cool, so great to hear from a user after all these years! Some of our users were suuper invested, like keeping up with hundreds of people - far more than we anticipated. I think our top user was a pastor in Australia who kept up with like 700+ people.

We would have continued a while longer but the 08 recession really burnt us. We were raising $$ in the weeks where banks were literally failing. RIP Good Times era. Our iPhone app was just about ready to ship and we were going to start building integrations like email next, but no runway. We had an angel lined up who’d have gotten us about 4-6mo of runway but I didn’t want to burn that relationship in such an uncertain environment, we were all pretty down about prospects then.

Fun fact tho: the app continued to work for years after. FB took a long time to deprecate API and our code was pretty solid.

Would accompany count? https://www.accompany.com/ not really personal, I believe it's business oriented.

They were the closest but yeah, biz is where the $ is if you’re anywhere near CRM functionality

Don’t TikTok and Instagram access the phone book and hence numbers?

Numbers yes, when you last contacted no. Core value prop of personal CRM is ease of keeping up w/ people, which means smarts of reprioritizing people based on several factors sort of like a personal assistant would, which means system should know when you last spoke/texted/emailed/etc. Ideally it'd even have ease of access to transcripts (chat/email logs) so you could jog your memory.

The goal is to avoid just a dumb timer saying 'its been 4 weeks, why don't you chat with X'. That's how Socialfly worked, and it was pretty high friction in practical use.

Not 100% what you describe, but functionally the same: https://github.com/monicahq/monica

Monica is an excellent "Personal CRM" as people are describing. It strives to achieve exactly the goals GP describes.

Personally, I would just like to see a contacts app that doesn't suck, and actually supports having both "companies" AND "people" as contact without treating them exactly the same. If I want to call some hotel or something, why does it have to be "First name: Holiday Inn" in order to even show up correctly...

MacOS/iOS Contacts supports this, for what it's worth. You can enter a 'Company Name' and leave fname/lname blank, and it'll give it a different default icon, as well as show it as the company name in lists/sorts.

The same in Android Contacts and GMail Contacts.

There's been a few of these since the topic has come up regularly in the last few years -

* https://keepmyfriends.com/ * https://getdex.com/

I haven't used any personally!

> I was thinking about something similar, but less cringy.

Well, if you are worried about it being cringy, you probably shouldn't watch the 2022 Salesforce Dreamforce (their annual convention). It's cultish and shows the unbelievably arrogant confidence they have in their own importance - along with the most inexplicably childish moments for a professional conference I've ever seen. (Let's put foam rabbit ears on our "co-CEOs" to entertain the "trailblazers" as we gather around a fake wooden stage imitating the outdoors with faux trees, won't that be funny?)

A personal CRM

This combined with the only useful feature FB had for years (people's birthday).

Integrate a scraping and keyword search for public facing social media postings to alert to people in distress, perhaps. Or just a general "you should check on X" feature. Not with automated messaging, that's too much.

sounds like henlo to me: https://henlo.app/

> hey you haven't messaged your friend in 6 months, maybe you should ask them how they are doing. A priority queue but to help keep friendships alive.

This is what social media and messaging apps should innovate on.

But, without VC money, without ads, without feeds drive by algorithms that optimize for engagement. Without dark patterns and antifeatures that make our lives worse. Just a non-profit that churns out open source code and is supported by monthly recurring donations from a % of the userbase.

I want one for remembering stuff about people for personal and business networking.

For a while, I was going to a weekly event where I'd meet people, but I'd never remember their name next time because I'm terrible with names.

So I started jotting down notes after it was over. Like, today I met person X who is really into this one hobby and just moved here from such and such city. And person Y who is looking for a new apartment. And Z who came with Y.

Then before going next time, I'd review the list, and if I saw X, I knew their name and could ask how they're adjusting after their move. Or I could ask Y how the apartment search is going.

At first I thought if people saw this list they might think it was a little weird. And maybe, but I'm OK with it since it's a way of making an effort. As long as it's genuine and your motivations are good, people like that you remembered stuff about them. (I'm not doing it to impress people, etc.)

Anyway, I didn't have a good way to organize it. I stuck it all in a document, which didn't work great.

I've been building FriendApp the past year. The core feature on iPhone right now is to just group people into lists, and preserve messaging integration with whatsapp, text etc. Trying to make it super simple. Also has ability to sync for recent contacts list. Would love feedback, and thoughts on the next features to iterate on.

https://www.friendapp.com/

I've found just setting a recurring TODO in the calendar to "Call XXX" is sufficient.

I signed up for https://infrequent.app/ when it was a Show HN just to try it out and it actually worked (probably about as well as a "non-busy" TODO in my calendar app with an alert). Enough to remind me to call 1 or 2 people that I like to talk to but often leave for too long, anyway.

I've thought about similar problems like rewatching your favorite movie, and concluded that treating it as a priority queue or trying to watch it every _n_ days is the wrong approach; you want to instead be reminded when you've forgotten about your friend Alex or what _The Leopard_ is about, so you can then ping them & catch up or rewatch that movie with a near-fresh mind. (If you ping Alex while you remember them, then maybe there's another actually-forgotten friend you could've pinged instead who would be more valuable to get back in touch with.)

It is, in other words, the exact opposite of 'spaced repetition', where you want to review right before forgetting to strengthen the memory the most; in this use-case, 'anti-spaced repetition', you want to review only after you've forgotten it. (https://www.gwern.net/Statistical-notes#program-for-non-spac...)

This is where I'd like to go with FriendApp. Would love if you wanted to check out where we are right now with our iPhone app. https://www.friendapp.com/

I've seen a couple tools pop up around this idea. Haven't explored them much, but:

- MonicaHQ (open-source)

- Dex (free version seems good enough for most people, i'm trying this one out rn) getdex.com/

- Custom system in Airtable - https://news.ycombinator.com/item?id=30329475

- Infrequent app - blog post talks more about it, linked in comments - https://news.ycombinator.com/item?id=33308084

A person on reddit actually made this app and it worked pretty well. I can't remember the name of it for the life of me.

you might be thinking of Monica? I seem to remember that coming out of reddit a while back, could be wrong