TweetFeast

I've been updating the TweetFeast AI tweet generator code and I discovered some tricks you can use to get ChatGPT to write better tweets.

If you've ever used ChatGPT to generate tweets with a prompt like "Please write 10 tweets on the topic of startups" you'll know the results can be pretty hit and miss. Let's be honest "hit and miss" is an understatement. You get bland and over-excited tweets devoid of any value, brimming with emojis and zero-interest hash tags. Basically you get tweets nobody wants to read.

While working on the AI tweet generator I ran into this problem and I've found an approach that works well. You can do this yourself directly in ChatGPT, or you can try TweetFeast - judge the results for yourself. TweetFeast will also use your own tweets as a style reference, or tweets from anybody else you nominate.

The basic solution lies in getting ChatGPT to be more "thoughtful". You want it to come up with some meta context about what it means to write good tweets before it actually writes the tweets. Here's a prompt I give ChatGPT before I ask it to write the tweets:

Thinking like a smart and thoughtful person with a large following on Twitter and a strong social media presence, what are some tips for writing tweets that get a lot of likes, retweets, and are widely shared? Also what are the main categories of these types of tweets?

ChatGPT will respond with a well thought out answer about how to structure tweets and what kind of content to put in them. Then you can ask it to tweet on the topic of your choice. Here are some more things you can do to have it come up with even better tweets:

• Give it lots of content about the topic, not just "startups", but paste some text that says something interesting about startups or startup life.
• Ask it to expand upon and/or summarize the topic and come up with points from it that will make engaging tweets before it writes the tweets.
• Give it a concrete list of rules to follow. Things like "minimise the use of emojis" and "don't reference other accounts with the @ symbol".

Hopefully these tips will help you use ChatGPT for tweet authoring. As with anything in life there's no free lunch. You have to put in a bit of work on the prompts to get good results. Or you can let TweetFeast put the work in for you of course.

PS No AI was used to generate this post. I wrote it myself. :)

Dear fakemano13 thank you for trying out the new AI Generated Tweet feature on tweetfeast.com! I'm sorry that your request for tweets in the voice of "Levelsio " on the topic of "STartups" failed to generate any tweets. The space you left in the username and the incorrect capitalization reveals just how little testing I did before sliding this thin half-assed wrapper around the ChatGPT API onto the internet. I shipped that thing so hard and fast the wheels fell right off. Don't blame me. This is what the why combinators told me to do. Do I need to pull out the quotes to convince you? Fine.

If you aren't embarrassed by the first version of your product, you shipped too late.
-- Reid Hoffman

Well I did the right thing Reid because if throwing an amateur hour exception right in the face of the user when they capitalize in the wrong place isn't embarrassing I don't know what is.

Ship before you're ready, because you will never be ready. Ready implies you know it's going to work, and you can't know that.
-- Seth Godin

It's true Seth, there's no way I could have known it wasn't going to work. Well I guess I might if I had taken literally 5 seconds to test before pushing deploy and sprinting to the fridge to crack a self-congratulatory beer as I tell myself I am innovation king. Ah well, I am glad I have your support anyway Seth.

Good news though! I have fixed the bug with a bit of the old .toLowerCase() and a cheeky little .replaceAll(" ", "") and I didn't even have to ask Stackoverflow how to do it. This time I tested for all of three minutes before deploying so I am 100% certain there are no more bugs lurking in there. You will have a glorious error free experience now. It will be like Shakespeare himself is writing tweets for you. The new version is up on tweetfeast.com/ai-tweet-generator and you can try it out again. I am sure you will find your tweets on the topic of "STartups" in the voice of infamous indie hacker Pieter Levels most satisfactory this time around.

Best of luck with your shameless AI generated tweeting. I wish you the same luck that I have had with placing poorly crafted software online well before it's reasonably ready for public consumption ($120 MRR already from doing this hell yeah!!!). All the best!

I've been working on a new feature for TweetFeast. It's as simple as you can imagine. It uses a large language model to generate tweets for you on any topic or text of your choice. If you've had trouble in the past thinking of what to tweet your troubles are over. With TweetFeast AI Tweet Generator you can generate a ton of tweets and queue them up for later using Twitter's "schedule" function.

My goal with this is to build the simplest web facing implementation of an AI Tweet generator. I'd like you to be able to get to generating tweets with as little friction as possible. Once you sign in with Twitter you can immediately start generating tweets in your own voice on any topic, or in the voice of any other Twitter user. Enjoy!

https://tweetfeast.com/ai-tweet-generator

TweetFeast is launching today!

I wrote a launch post about this but I want to put it on the record as a milestone.

I have been slowly adding features week by week and implementing the payment gateway etc. I kind of wasn't expecting to be ready today but as I worked through all of the last issues I ran out of things to do at about 2pm. So I decided to go ahead and pull the trigger. \o/

Unfortunately it looks like Hacker News have shadow banned my Show HN launch post which is a bit unfortunate! Hopefully launching on Twitter, Product Hunt, and Indie Hackers is enough to give it a good start.

Signups are trickling in as people test out the tweet download functionality. Somebody reached out to ask about bulk educational pricing. Once I launch properly this will be something to consider.

I also dabbled with a couple of ideas for pivoting TweetFeast to a different Twitter app idea. After testing some stuff and then thinking better of it I came back to the core idea again. Got to stay focused.

Work continues on the features and organic search traffic is gradually starting to trend upwards.

I had a couple of slow weeks. struggling to motivate myself to work on TweetFeast. Eventually I managed to get back on the horse.

I built a simple blogging system and wrote the first article for the TweetFeast blog which is called Twitter Sentiment Analysis. It shows people how to do their own sentiment analysis of tweets using Python or JavaScript.

I started seeing a small trickle of traffic for these keywords soon afterwards.

This week I added table view and download of the tweets you searched for.

This is the core feature of the platform so it's great to get it in there. Now people can search for tweets and then download them to CSV or JSON format.

I did a bunch of research on keyword volume for the problem space, looking at existing alternatives, and seeing where I could differentiate.

I decided to build completely in public, posting updates on Twitter every week. This is the day I made the first post and you can find the full thread here:

In the weeks following I got basic stuff working like authentication and interfacing with the Twitter API.

Later I ended up skipping some weeks but having this thread really helped me come back to it and re-focus again. Building in public is a good motivator.

My music apps are doing ok making a small amount of money each month. I'm making sales, but each individual sale is only a few dollars and I can't help thinking "what if each of these sales was a monthly subscription for $49 instead?"

A good micro-SaaS idea should fit these criteria:

• Has business payers.
• Has existing demand.
• I have experience in the domain.
• Low on-going maintainance.
• I can reach the customers.

A Twitter based micro-SaaS seems to fit this mould and checks all o the boxes. I have personally had the need to download Twitter data for analysis myself. Time to start researching.