10k Tokens is All You Need

In the paper, researchers introduced the idea that should be very familiar to anyone who has ever heard a joke with a punchline -  a joke only becomes funny once you've received the punch line. Everything that comes before the punchline is context.

This is basically what the above paper proposed - that it was possible to accurately and cost-effectively predict the next word in a string if we looked at all the preceding words and tried to make sense of them as a whole. 

Until then, text predictor AIs used to only look at the immediately preceding words to predict what would come next. 

The title of today's post is an homage to the title of that paper.

For a hiring manager, 10k tokens is enough to create a handful of job descriptions for open roles in your company. 

Based on feedback from our beta users, we are redoing how JDs are created by automating and consolidating in a single screen all the boilerplate stuff relevant to a job.

These include things like Job Title, Location, Remote/Hybrid/Onsite and a few other pertinent options.

Earlier versions of SquadGPT forced the user to interact with, and spend tokens to, input the same boilerplate info to the backend AI. 

Now, thanks to the above change, our users can save tokens and time by completing their JD much faster.

In our tests, we have found that 10k tokens is more than sufficient to adequately complete one screening interview with the AI backend.

Now is probably a good time to talk about our very fair pricing for candidates. Here's our pricing plans page for your review. 

Our $8.99 plan buys candidates 200k SquadGPT tokens which is about 20 screening interviews. 

There have been a number of memes and products created around the title of the original paper  and this blog post is just one more in that space.

Here, we have tried to give you a sense of what you can do with our free tier while staying true to our hacker ethos. Can't want to see how people use us.