Automation

I hate overengineering and overcomplicating things.

And I used to think that trying to automate stuff was both, and not worth the effort.

But almost all automations I’ve done was worth it in my opinion.

Part 1: Sniping Restaurant Reservations

#

4 Charles is a popular NYC restaurant that’s very hard to book. This is due to:

1. It only has 9 tables
2. It reserves a lot of them for celebs

My then-financee-now-wife tried to book it at midnight (though you can book at 9AM now) and was pissed when she couldn’t.

We sleep early so had to set an alarm just for this. And we couldn’t book it!

The next morning at 6AM I spent 4 hours writing a bot that could automate booking stuff on Resy.

We got in.

I realized then everyone that gets in likely has connections or must have used a bot to book.

And if you can’t book a popular restaurant, you’re not slow. I have logs that show that all restaurant reservations are gone by the time .05 seconds after the reservations come out. This usually means the restaurant “releases” reservations, but in actuality they blocked off all tables for celebs already.

Or their system was broken.

Anyways, I thoroughly enjoyed this automation because for a one-time 4-hour effort I can book anything on Resy that’s bookable.

I did the same for Double Chicken which is also quite hard to get in, etc.

Part 2: Pickleball

#

There’s free pickleball every summer near where we live.

We played last year and it was fine. But this year, we had a week where we weren’t able to book any slots.

It’s possible we are slow, but it’s likely that there’s only 1-2 slots opened due to sponsorships, and bots booked those slots.

Using the techniques I learned in the Resy automation work, I was able to build a bot to frontrun pickleball reservations much faster (about 1 hour).

We were able to get any and all pickleball reservations thereafter.

If you think about it, the 4 hours doing Resy automation is doubly useful because the education I got from it helps me automate other tasks faster.

Part 3: AI Automation

#

There’s this service called Higgsfield. If you look at their API pricing, it is at least $0.13+ per 4 images.

If you use their UI, their pricing is much cheaper.

Again, using the same techniques as above, I was able to spin up my own backdoor API using the UI to save costs.

At least, in theory. I switched back to API because:

1. Hosting a chrome browser on a server requires a more expensive node, which offsets cost savings.
2. Their API used to cost $0.30+ per 4 images, so the price differential has gone down drastically.

Part 4: Lead Generation

#

I’m doing cold emails lately, and one of the problems with buying leads is that everyone uses the same lists.

Thus, you’ll get bad results using paid leads because they will be bombarded with offers.

Made a script that auto-pulls influencer emails from YouTube based on keyword search results.

Plugged those into Instantly to mass-email.

Bam, now I have ~1 appointment a week talking to influencers (can be more if I scale up but I don’t have too much time), ranging from the 250k to 5M subscriber range.

I didn’t automate the mass-emailing part because there’s no reason to write an automation for it if a tool exists already.

When To Automate?

#

As you can see, rewards can be plentiful if you automate. But you shouldn’t automate everything.

You should only automate if both of these are true:

1. It is a huge pain to do manually
2. The reward is outsized

Example: Reservation sniping

#

Booking restaurants/pickleball isn’t just a huge pain, it was impossible. And the reward was outsized (for me anyways) because I actually get to eat in places I normally don’t get to eat in, and play pickleball for free.

Example: API

#

Likewise, it would be a huge pain if not impossible for me to service SaaS customers by manually making AI images on Higgsfield, and then sending it manually to my customer, all to just to save a few bucks. The reward is outsized here because at scale, the small savings become huge.

Example: Cold emails

#

And finally, it is a huge pain to do cold email and lead generation manually. I used to look for emails on YouTube directly but this is a very slow and painful process. Takes me an hour to get 10 emails. Now, I can get 300 but just typing in a few keywords in my script.

The reward is outsized because I only get 20 responses sending to 600 emails. Of those, I might take 5 appointments.

Additionally, the automation does personalized cold emails. Each one, hand-written, might take an extra 10 minutes each. Or another 250 hours.

Caveat: Don’t Make This Mistake!

#

Don’t mistake overcomplication for automation.

At my work, we have these morons that create new flows every couple weeks, with mandates that everyone use them.

The flows, unfortunately, suck.

Each flow is more complicated than the next.

And at the end of the day, doing things the old way would be a drastic improvement than using new flows every week.

But of course, work is different. These are just morons trying to ‘do work just to show work being done.’ That is another story though.

Anyways, automations are supposed to make your life easier, not harder.

And it’s OK to pivot. For my API example, I realized automating was no longer worth it so I just switched back to a regular API instead of my backdoor one.

So if it’s making your life harder, stop. Get some help. You’re overcomplicating things, not automating.