OpenAI is full of drama. Sam Altman was fired, seemingly a coup of sorts. Greg Brockman, the president, resigned. Other employees showed support for and solidarity with Sam, threatening to resign as well. Board considered bringing Sam back and resigning. People sent hearts on 𝕏. Sam might join Microsoft, might start a new company or might join another company. What is the fate of OpenAI?

It's an exciting saga, TV worthy. There is no doubt in my mind there will be a movie about OpenAI at some point.

But, I'm not really looking for drama with my AI.

I want repeatable trustworthy quality from my AI.

I want it to handle systems for me and my clients so we can focus on growing our businesses.

I don't want to wonder if my AI provider will exist next week. I don't want to be concerned about the stability of the models I use.

And if my AI Provider does decide to close up shop, raise prices, break models or change data usage terms, I want to be able to switch models without much disruption.

I am a huge OpenAI fan. They have set the agenda for what tech looks like this year and have pushed out innovative and exciting releases every single month.

But, I don't want to rely entirely on them. This drama confirms it.

Coincidentally, this drama seems to be resulting in a dispersion of OpenAI talent, which will undoubtedly slow down OpenAI and allow competitors, where some former OpenAI engineers may land, catch up.

I'm a firm believed in Open Source and want to leverage some Open Source models in PyroPrompts.

Introducing Llama 2

Workflows now support models that are outside of OpenAI. We use AWS Bedrock to get access to Llama 2 and have rebuilt some parts of the Workflows backend so we can easily integrate additional AI Providers as they become available.

All of the AI Providers are serviced by the same Credits payment. You buy Credits through PyroPrompts and they're good for any model!

When you're creating an "Execute Prompt" Step in a Workflow and selecting the Model, scroll past GPT-3.5 and GPT-4 and find Llama 2. It works well for some tasks. It doesn't handle the complex tasks as well as GPT-4, but it does well with simple summaries, categorization or lists, which can be sufficient in multi-step processes.

Enjoy!