Ask HN: Overwhelmed with Non-Actionable Email | Hacker News Reader

Ask HN: Overwhelmed with Non-Actionable Email

8 pointsineedasername13 comments

I have a zapier script setup to text me likely important messages and it helps but isn't perfect and things fall through. Special filters for to surface organizational email addresses help but the 1-2 times a week I receive an important non-organization message fall through that crack and look like the ever-improving auto generated spam content.

I declared "email bankruptcy" last week for > 1500 "new" messages, scouring them as best I could for things that needed a response, but these were very few. Most messages were nonsense. Organization-wise messages that can mostly be ignored and 80% spam (my workplace uses Google's ecosystem for mail etc so spam should be filtered well!)

After "bankruptcy" I buckled down and determined to deal with each batch of mail first thing and last think each day, ubsubscribing one by one to slowly stop the problem, but the shear volume of marketing crap & process of unsubscribing etc took so long that after two days of spending 45+ minutes dealing with it I had to move on to pressing projects. On top of that it seemed like some unsubscribes immediately generated a near identical spam message for something similar, albeit a different domain, as if the unsubscribe system is just feeding those unsubscribe emails to the next marketing link in the chain in a never ending chain of spam!

How are people managing this? Is there a tool, anything, any idea people use to overcome the shear avalanche of crap without missing the important things? My boss loves me, knows I work hard and am busy, but is sometimes having to remind me to respond to something I completely missed among all this crap. I need a solutiont!

Thank you!

So here I am a week later with 400+ messages in my inbox again!

I work with a lot of people that also seem to struggle with this.

Personally, I don't really get what's so hard about it. I have a few rules that filter out total automated garbage/noise. That leaves me with a few hundred emails each day - I spend 5 or 10 minutes scanning through those looking for important people/topics and click into them if they seem interesting, and after I've done that, I mark all as read... and do it all again the next day.

Often, I find something that is interesting/important, and I chat it to other people in the team, i.e., "RE: Question from important client about new feature not working <---- Did you see this email?" and they're like no... I didn't get that, who was it sent to? It was sent to <list@organization.com>, of which you are a member... if you did not receive it, please raise a ticket with the mail people... ooh, yes, found it!

The impression I have is that these people are not willing to even spend 5 or 10 minutes a day scanning, or they have attempted to set up an elaborate and brittle set of rules to do that for them and failed.

Dedicate a full day to the following:

1. Buy a domain that looks like a knock-off email provider (to avoid some scrutiny later). Something like emailco.biz, etc.

2. Setup simplelogin for that domain.

3. Go through all your existing accounts and change the email to [service]@emailco.biz.

4. Use that technique to sign up for all future services. If you start getting unwanted email to a simpelogin alias, simply turn it off in the app or browser extension.

5. Click unsubscribe on every email you don't want. The first day it will be a lot, the second day less, the third day even less, etc.

6. Ruthlessly click spam on anything you didn't ask for.

Soon, you'll have it under control.

I'd say that Gmail and unsubscribing are your problems. At $Job, I receive email for 8+ different addresses, all publicly known for 20+ years. Most days, I receive <40 spam messages. $Job's spam defenses are: (1) a ~1,200-entry email blacklist (built up over ~12 years), and (2) an IPv4 firewall blocking about 24K addresses from our mail servers (based on hitting honeypot email accounts, obvious password guessing attempts, etc.). Falsely blocking legit email has happened, but it's very rare.

I check the sender. Looks spammy, hit spam. This also sends the mail to Trash.

Then check the subject. Looks spammy, hit trash. This likewise sends the mail to trash.

If it looks like a newsletter or subscription I don't want, I typically don't want updates more than once a month as there's RSS for feeds and email is not your push feed to my eyes, I scroll to the bottom and click unsubscribe. But check the link first. If the link's to a 3rd party domain i.e. not the same as the email comes from I hit spam as there's no way of knowing if it isn't phishing. If it doesn't link to an unsubscribe page but some 'manage my email preferences' page it's in violation of my personal compliance so I hit spam.

If the email doesn't have some form if text below the fold on my not that low resolution screen I'll delete it.

Having done this pruning, which for 100 mails only takes a couple of minutes if that, I go back and deal with what seemed more important on first pass.

That's my way. YMMV.

Get at least 1 more email address and keep it only for services you pay for. The biggest problem with subscribing is the on selling of email addresses.

Get a proton paid mail account and your own domain. Setup a "catchall" address. Now when you subscribe to "something" then use the name of the "something" with your domain name as the email address. Example: to logon to Hacker News use the address hackernews@yourdomaim.com. You do not need to setup an email address the catchall receives your email. If you get spam addressed to hackernews@yourdomain.com you know Hacker News on sold your address. If you don't after a month or so change your hacker news address to your legitimate address. You cannot send from these catchall addresses, but if you truly need to you setup a legitimate email with that particular address and send it then delete it. Proton Mail have excellent spam filtering anyway. There are other providers that do the same.

The biggest problem with "unsubscribe" is that it effectively verifies that email address. You may unsubscribe from one list only to be added to lots of other lists because it is a verified email address.

I found a great way to reduce spam is to have email address that looks like a big company ie google@yourdomain.com or microsoft@yourdomain.com

Your email client needs attention. Make sure you haven't set it to save all incoming addresses and the spam filter to "ignore spam setting for emails in my address book". I am not familiar with google mail but make sure the same sort of thing is not setup for gmail. (...we don't read your emails we just extract anonymous information for market research purposes..so spam is never an affiliate?)

To solve your present problem set aside at least 1 hour a day and setup block lists. You need to ruthlessly organize your contacts lists. The problem will get worse before it gets better. I found unsubscribing just breeds more. It is going to take weeks or months, but the short term pain will be worth it.

By the way I think gmail may have limits on its spam filtering. (...we don't filter out affiliates....really?)

You could get your corporate IT to setup all lists with separate sender addresses to allow filtering, but they will probably hate you for asking...

A few things I do:

- have a separate email to sign up to various websites with so I can separate my inboxes. - religiously use email filters to archive or categorize emails based on sender or keyword. - I have a custom script that uses GPT-4 to read new emails, categorize them as important, and summarize the long ones - then it sends me a message. Trying to find ways to further improve this as well.

Wish I had access to GPT-4 API… it’s been months, since the day it went live. Not sure I want to trust an importance filter to 3.5.

Would you be willing to share the code you use to implement this against the GPT-4 API? Rolling my own code is something I’d barely have time for without sacrificing a weekend

I 'll try to share this when ready but unfortunately it's a disaster of a Python notebook right now. The GPT API is pretty easy to use with the openai package, the harder part is parsing emails for their content and getting rid of HTML crap.

GPT-4 waitlist can be quite a lottery, but you should get decent results with GPT-3.5 with enough tweaking and allowing it some space to reason. Try something like this for the system prompt:

``` Analyze the importance and urgency of the given email chain. Respond in JSON Format:

{ "analysis": "This email is about a sale at Best Buy, which is not important", "is_important": false }

{ "analysis": "This is a personal email from Steve asking me to fix the copier. Sounds urgent.", "is_important": true }

```

Thanks, I may give it a shot with 3.5 and migrate when I get access. I'm not sure how 2FA will work with API access to my organization-managed g-suite gmail. Zapier doesn't seem to stumble over the need for 2FA that I have to go through IRL (which seems a potential security hole?) So maybe it won't be an issue if I write against the API (gmail) myself too. I think that's what I would have to do, if I use a simple python client then 2FA would almost definitely be triggered.

Yeah, just follow the Gmail API Python quickstart tutorial. You'll authorize your app to your account and then it can re-use the token without 2FA.

I used to use POPFile as a trainable Bayesian classifier to mark spam and ham. Worked well.

I used to use a system that decreased the amount of email I read by 100x.

I recently built another system that decreased it to zero.

The last two aren't organization friendly. The Bayesian classifier might be your best bet.

You can create a gmail filter that filters out all emails with the word 'unsubscribe' in them.

I get emails from outside temp/consultants that have unsubscribe as part of their boilerplate footer. That’s an edge case though… many e I could figure something out

Reply on news.ycombinator.com