The Entrepreneur Forum | Financial Freedom | Starting a Business | Motivation | Money | Success

Welcome to the only entrepreneur forum dedicated to building life-changing wealth.

Build a Fastlane business. Earn real financial freedom. Join free.

Join over 80,000 entrepreneurs who have rejected the paradigm of mediocrity and said "NO!" to underpaid jobs, ascetic frugality, and suffocating savings rituals— learn how to build a Fastlane business that pays both freedom and lifestyle affluence.

Free registration at the forum removes this block.

[Progress Thread] Launching my AI web app

A topic related to SAAS or APPs

Jeannen

Silver Contributor
FASTLANE INSIDER
Read Fastlane!
Speedway Pass
User Power
Value/Post Ratio
217%
Apr 9, 2020
354
769
France
Hey there!
So, I started to code a few months ago (here) and I recently started my first "real" project one week and a half ago. Based on the feedback I received, I'm thinking it has potential, so, I'll focus on developing it and see how it goes. This thread will be my "update".

It's an AI writing assistant that helps you write content. It has two parts: One part with specialized modules that helps you write short content (product descriptions, Facebook ads, etc...) and a long-form editor to write stuff like blog posts, newsletters, etc... (article 100% generated by it here)
It doesn't replace a copywriter, but it's a great way to get ideas and get rid of the blank page syndrome.

Yes, I know, similar apps exist but that's a good sign. Also, I have a specific target + value proposition in mind that not many people can focus on (thanks to my experience)

Where I am now:
I started building the app two weeks ago. The specialized modules part is partially done, it can generate good product descriptions and Facebook ads already. The "long-form editor" is WIP but already working and producing good content, I just need to make some tweaks to make the user experience more enjoyable.

Current goal:
Find people around me that would test the app and watch them doing so, to gather feedback on the user experience. That would allow me to improve it before releasing anything. Then, if I find enough people interested in the app and if they are ready to pay for it, I'll focus on making a "public" version.

No idea how it will go, but I guess it's experience in any case

P.S. I used the app to generate part of this post
P.S.S Feel free to DM me if you'd be interested in testing the app
 
Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

heyben

New Contributor
User Power
Value/Post Ratio
113%
Aug 7, 2022
8
9
The internet
What API are you using for the AI part?

I recently got access to DALL-E 2 and have been creating some really cool things with their AI graphic program :)
 

Jeannen

Silver Contributor
FASTLANE INSIDER
Read Fastlane!
Speedway Pass
User Power
Value/Post Ratio
217%
Apr 9, 2020
354
769
France
What API are you using for the AI part?

I recently got access to DALL-E 2 and have been creating some really cool things with their AI graphic program :)
Oh damn, you're a lucky one haha, I'm waiting to get approved

For the model, I'm mainly using GTP-3 with a mix of fine-tuned models, prompt engineering and external APIs. Wasn't expecting it to be that hard but I'm getting really nice content now
 

Jeannen

Silver Contributor
FASTLANE INSIDER
Read Fastlane!
Speedway Pass
User Power
Value/Post Ratio
217%
Apr 9, 2020
354
769
France
10/08/2022

I'm making progress. I managed to fix the issues that were left in terms of content creation, now it will likely be customer experience optimization. I released the first version of the app so one of my friends can try it and give me feedback already.
I think I'll focus on improving the UX for a few days and then start giving limited access to people.

Not gonna lie, I'm pretty proud of my work. Those past 2 weeks have been intense, from finding the idea to now having a functional prototype.
(To give some context, I work from 7-8AM to 8-9PM every day, with just a few breaks to eat and cool off my brain)
 
Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

Kirk84

Bronze Contributor
FASTLANE INSIDER
Read Fastlane!
Speedway Pass
User Power
Value/Post Ratio
226%
Jun 9, 2020
113
255
Munich, Germany
test the app and watch them doing so

I think this is smart. Additionally, you should install logging and offer a feedback button for your users in your app. Two tips I got from SaaS Masterclass.

with a mix of fine-tuned models, prompt engineering

I tried fine-tuning GPT-3 a couple of months ago but I could not find the training examples needed. I guess this is very task-specific, or is there a smart way to easily find training examples for GPT-3 fine-tuning?

Is there a good forum where developers exchange information on how to make GPT-3 projects work very well? I feel like a lot of people are playing around with it, but not very many are actually building products and usable tools that are able to monetize.

Can you hint me to a good resource on prompt engineering?

Thank you!
 
Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

Jeannen

Silver Contributor
FASTLANE INSIDER
Read Fastlane!
Speedway Pass
User Power
Value/Post Ratio
217%
Apr 9, 2020
354
769
France
I think this is smart. Additionally, you should install logging and offer a feedback button for your users in your app. Two tips I got from SaaS Masterclass.
Yess, the feedback button is on my priorities list, it's essential for anything related to AI!
Btw, If you got good SaaS resources, I'd gladly take a look!

I tried fine-tuning GPT-3 a couple of months ago but I could not find the training examples needed. I guess this is very task-specific, or is there a smart way to easily find training examples for GPT-3 fine-tuning?

Is there a good forum where developers exchange information on how to make GPT-3 projects work very well? I feel like a lot of people are playing around with it, but not very many are actually building products and usable tools that are able to monetize.

Can you hint me to a good resource on prompt engineering?

Thank you!

Yeah, for fine-tuning, you'll need hundreds/thousands of quality content, which is one of the main difficulties in the AI field from what I saw (anything related to training AIs -> MASSIVE opportunities in this field).
One thing you can do is generate lots of content using the base model and then filter to keep only the best (synthetic training). But for very specific fields it's going to be harder since you won't be able to generate datas.

If you want to go into this field, I highly suggest you take a look at how AI works underneath, it will make everything easier for training. Feel free to DMs me if you want good resources (might be hard if you don't have math background, but do-able)

I don't know if there is a forum other than GPT-3's official forums (could be a biz opportunity?), I just spend days and days doing trial and error, so far that's the best thing I've found for prompts.

I agree with the "not many people build products", the thing is that now it's publicly available, if you only have basic operations on your app, you're basically just "re-selling" GTP-3
 

Kirk84

Bronze Contributor
FASTLANE INSIDER
Read Fastlane!
Speedway Pass
User Power
Value/Post Ratio
226%
Jun 9, 2020
113
255
Munich, Germany
Btw, If you got good SaaS resources, I'd gladly take a look!

Take a look at these two posts I made:

Yeah, for fine-tuning, you'll need hundreds/thousands of quality content, which is one of the main difficulties in the AI field from what I saw (anything related to training AIs -> MASSIVE opportunities in this field).
One thing you can do is generate lots of content using the base model and then filter to keep only the best (synthetic training). But for very specific fields it's going to be harder since you won't be able to generate datas.

That is why my first GPT-3 project failed. I could not find the training examples needed. Maybe it makes sense to partner up with a company that does provide the resources needed to train and share profits...

If you want to go into this field, I highly suggest you take a look at how AI works underneath, it will make everything easier for training. Feel free to DMs me if you want good resources (might be hard if you don't have math background, but do-able)

Thank you, I don't think I have a deep understanding but my foundation is solid. I took the Deep Learning Specialization by Andrew Ng last year, which took me forever (roughly 8 months) to finish but I feel quite comfortable with the maths background.

I don't know if there is a forum other than GPT-3's official forums (could be a biz opportunity?), I just spend days and days doing trial and error, so far that's the best thing I've found for prompts.

That's a great idea. Will take a look at this!

I agree with the "not many people build products", the thing is that now it's publicly available, if you only have basic operations on your app, you're basically just "re-selling" GTP-3

Are you sure, though? Maybe it's enough to offer basic functionality with GPT-3, like having good prompt engineering for the task at hand, to be able to make users roll in. And if it works out, improve the algorithm by fine-tuning... hahah, I could even ask the users "do you think this is a good answer?" and if they answer "yes", use the answer as synthetical training data...
 

Jeannen

Silver Contributor
FASTLANE INSIDER
Read Fastlane!
Speedway Pass
User Power
Value/Post Ratio
217%
Apr 9, 2020
354
769
France
01/09/2022
After getting some feedback, I decided to start from scratch again. The new version of the app can now generate whole articles just based on other articles links. For example, you find 3-5 similar articles, you paste them in the app, and the app will make a new article from it.

It was a headache to come up with it but I'm pretty satisfied with the results. The first beta testers are now using the app, while it's not 100% perfect, results are usually fairly decent.

The code is not the best, app is probably full of bug, but I'm not sure it matter rn, I'm trying to get a proof of concept, and as soon I have it, I'll be able to hire people.

Main main focus is going to switch from coding & building the app to marketing and user experience. At least until I get enough feedback to make a new iteration.
 
Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

Jeannen

Silver Contributor
FASTLANE INSIDER
Read Fastlane!
Speedway Pass
User Power
Value/Post Ratio
217%
Apr 9, 2020
354
769
France
02/09/2022

Spend the last few days reading stuff about marketing and GTM (Going To Market), it gave me a totally different perspective compared to eCommerce. I'm going to have to find my beta testers one by one, first in the people I know, but then probably cold emailing or something.

I have a pretty good incentive program (Lifetime "tester" plan for free) for those who test the app and give me feedback.

Goals:
- Crash-proof the app (dev)
- Get people to use the app (user experience + product/market fit)
- Get feedback (user experience)
 

Post New Topic

Please SEARCH before posting.
Please select the BEST category.

Post new topic

Guest post submissions offered HERE.

New Topics

Fastlane Insiders

View the forum AD FREE.
Private, unindexed content
Detailed process/execution threads
Ideas needing execution, more!

Join Fastlane Insiders.

Top