darjagutnick

Woop woop! Thats cool :) working hard to bring you back, grayprog!

Hey superzadeh,

Hard problem to tackle, for sure! Currently, we're doing mainly 2 things: 1) we're pushing hard on making sure that our content comes from a diverse group of experts (Bunchee is referencing the tips/content from our library, which is the main difference to ChatGTP) 2) we're also working on assembling a diverse panel of coaches, to diversify the perspectives and PoVs of those that "train" Bunchee.

And we're open about the fact that she is limited and may give harmful or biased advice.

Yes exactly, we kinda needed it ourselves tbh, I just used it yesterday again to prep an investor meeting and it went great.

Yay! thanks for the feedback, Boris, glad you enjoyed chatting with Bunchee!

Hey Denis,

Really good question, we're doing mainly 2 things: we're pushing hard on makiing sure that our content comes from a diverse group of experts (Bunchee is referencing the tips/content from our library, which is the main difference to ChatGTP) and we're also working on assembling a diverse panel of coaches, to diversify the perspectives and PoVs of those that "train" Bunchee. And we're open about the fact that she is limited and may give harmful or biased advice.

Hey everyone! Sharing this project we’ve been working on.

We run a coaching app for new & aspiring managers (mostly engineering managers and PMs). Users get personalized 2-minute daily tips (we shared it here way back in 2020), and they can reach out to a coach via Intercom chat if they need help with a specific problem.

We typically need at least ~15 mins or so to respond to these coaching requests via chat, and our team isn’t big enough yet to provide 24/7 coverage, so we decided to try out OpenAI’s API to see if we could build an “assistant coach” who can chime in faster and still provide actual value - partly inspired by the Marc Randolph’s (Netflix Co-Founder) quote that “the next wave of disruptive companies will be built by layering domain expertise over GPT-3”.

**We used prompts to give GPT-3 context on its coaching mission, including a coaching framework & questions, we gave it access to our content library of 2-minute tips for managers, and did our best to give it a bit more personality.

The results seem pretty good so far! We’ve introduced “Bunchee” to users in our app already, and just today released the web demo. There are definitely still some issues though:

So far, Bunchee is good at:

- empathy - humour - improvising (most of the times) - understanding that she is a colleague of human coaches at Bunch

What she’s not so great at:

- Actually being available 24/7 (due to OpenAI API dependency) - remembering past conversations to build more context about the user - switching modes on the fly: coaching vs. advice vs. search vs. customer support

Curious to hear what you all think!

Heya! We have a couple of different sources. In the beginning, our team was summarizing the best resources that we found online to address common challenges that users reported (and linking to the full version).

These days, we're also collaborating directly with coaches and other contributors from our community. We've found that the "magic" of Bunch comes from the little a-ha moments you get from reading new or different perspectives on a topic, so we try and create those moments.

Making time for professional development can be tough - especially when it comes to the “soft skills” often needed to progress in your career like:

Giving and receiving feedback Addressing problems in your team, or even the simple art of how to say no when you’re overloaded with requests

Courses or training programs in this area tend to be super generic as well, so our small startup team set out to create a personalized, micro-learning app to help people learn something new + relevant in just a couple minutes each day. We called it an AI leadership coach.

We launched it in 2020 and things were going great, until we started running into some major issues:

We had built at a furious pace just to get the app out in the world and see if people wanted it at all, but we piled up tech debt in order to move fast Because we had decided to prioritize new features over quality, we started getting a TON of bug reports and complaints about loading times. We knew we were making a tradeoff, but the impact on user experience escalated faster than we initially expected. Every time we fixed one bug, we introduced new ones, so fixing what seemed like a simple issue would end up being a major rabbit hole

We realized at some point that we weren’t realistically going to be able to develop on top of what we had built (at least not as quickly as we wanted), and made a tough decision to re-write the app entirely, even though we knew that this is the kind of project many teams have started but few actually finish.

While we managed to pull it off, it took longer than we initially hoped. In total, it took over 6 months (though we were also maintaining the “old” app). We ended up moving away from SwiftUI and Combine to instead go with UIKit and RxSwift. We also decided to build a backend API and moved away from using Firestore directly as well.

I don’t think I would advise others to go down this path, but I’m happy to say we shipped the new app, and so far: our speed of development has increased massively - we’re now shipping ~once a week or so and able to work on meaningful features. Crashes and bug reports are way down, as are loading times.

Curious to hear what HN thinks of the new app, and any questions about the process!

SWIFT UI :)?