The first official newsletter for jaypeters.dev!

Every week I’ll be highlighting interesting links as well as an update on what I’ve been up to. Thanks for following along!

Topics for this week:

• beehiiv

• ChatGPT’s Image Input

• The Metaverse

• Blog Efficiency

• LLM Pipeline

beehiiv

First up beehiiv. I’ve heard good things about beehiiv, and I started following Ben’s Bites, an AI newsletter that he operates through beehiiv, so I decided to give it a try. There’s plenty of similar blog/newsletter platform options out there, but this one has a good set of features to make things conveniently located in one place. I’ll be operating my newsletter and main blog through beehiiv so let me know what you think in the meantime, and I’ll be reporting back on my experience with the platform after some time using it.

ChatGPT’s Image Input

I ran across this on twitter, from Mckay Wrigley, who’s been hacking away at ChatGPT for a while now, but this particular video was pretty impressive:

I’ve been in this situation many times in my career, gathered in a conference room for a white board session to understand a new system or idea. And usually once everyone left, it would be a few days minimum before anything demonstratable was built. The fact that this took a picture and spit out working code for it is quite mind-blowing. The power for prototype rapidly will be invaluable when this feature is available to the public.

*This feature is rolling out slowly to ChatGPT Plus and Enterprise users right now.

The Metaverse

I’m not sold on the “metaverse” or at least the way Facebook and crypto have tried to sell it. Unfortunately I think we already have the metaverse, it’s just much more difficult to describe how we all interact with devices connected to the internet. Not nearly as marketable that you’re already using the metaverse.

But the technology Lex Fridman and Mark Zuckerberg use to have this conversation in the “metaverse” is really wild. They both joke about being the two most monotone people to listen to have a conversation, but it’s interesting never the less. You don’t have to watch the full hour long video, but it’s worth watching a few minutes:

Blog Efficiency

Since I’m planning to blog more regularly, I’ve been looking into ways of making the process more efficient. It will still be a human-written blog, but I’m happy to leverage AI and automation to repost the blog content to multiple platforms.

I’ve been using Zapier to connect apps quickly and set up simple Trigger and Action templates to make things easier and automatic, but as of now I only have one solid workflow.

If I create a Google Word Doc in a specific Google Drive folder, it will automatically copy the content of the document and use it to create a post on Medium, which is quite handy! This same flow can be used to post to WordPress as well, but you have to have a Business account with WordPress to enable the plugin to do that. That may be an investment in the future.

I was led astray, though, with beehiiv. Zapier supports connectivity to beehiiv, but some how doesn’t have the action to create a new Post yet. I might have missed something in the options, but hopefully that functionality will be added in the future.

Also, if you’re looking for help with similar automations with Zapier like I mentioned above, I’ll be offering them through my business LastHalf.tech. Let me know what apps you use regularly and we can figure a way to streamline some of your workflows!

LLM Pipeline

This is a deep one. Most large scale software businesses will have a development and environment pipeline. There’s a number of code checks and testing that needs to happen before an app can safely (or sometimes even after that lengthy process still unsafely) make it to production where real world people use the app. It’s a process I spent countless hours learning, fixing, researching, discussing and dreading while worked at Salesforce, and I’m sure hundreds of other tech companies feel the same.

I saw an ad for this company, Autoblocks.ai, and skimmed one of their blog posts which had a really fascinating idea: implement multiple LLMs (large language models like ChatGPT) through the pipeline process to debug faster at each step. An LLM is much quicker at digesting large amounts of data and answer questions about it in human understandable natural language. I could see something like that saving countless hours of digging through semi-understandable debug logs, trying to figure out what caused the pipeline to stop working, or why something expectedly became an issue the night of a major release.

Granted this is coming from a blog post on the company’s website so they’re going to have some bias toward their own process and product, and I haven’t tried it myself, but the concept is a great one assuming it’s implemented correctly.

An army of LLMs that easily read and interpret complicated tasks that would otherwise take humans potentially hours and days to do the same. It’s a more real world example of a concept discussed in one of my own blog posts. Leveraging AI will enable technology companies of all sizes to accomplish more than ever before. It’s a good train to be on.

And that’s a wrap! The first newsletter done. Those are some of the tech topics that stuck with me throughout the week or I personally worked on. Let me know what you think at @jaypetersdotdev or email [email protected] and thanks for reading!