--- title: "Newsletter Issue 9" description: "New features for the logseq openai plugin and my thoughts on why Large Language Models are interesting" url: "https://briansunter.com/newsletter/issue-9" --- NOV 26, 2022 · 5 MIN READ # Newsletter Issue 9 New features for the logseq openai plugin and my thoughts on why Large Language Models are interesting  ## Summary I’m continuing to add features to the [Logseq GPT-3 OpenAI plugin](https://github.com/briansunter/logseq-plugin-gpt3-openai), my plugin that allows you to perform AI text generation tasks in the Logseq note-taking application. I’m really interested in this new AI technology called “Large Language Models” which allows AI to understand the meaning behind text, as well as generate humanlike text. ## Logseq GPT-3 OpenAI plugin I recently added some new features to the logseq OpenAI GPT-3 plugin. #### Logseq OpenAI Plugin Video The [productivity Youtuber Shu Omi](https://www.youtube.com/@ShuOmi_Official) made a great video about my [Logseq OpenAI plugin](https://github.com/briansunter/logseq-plugin-gpt3-openai). I recommend checking it out as an introduction to the plugin and AI-assisted note-taking. I’ve been a fan of Shu’s videos for a long time while learning about notetaking, and I am delighted he highlighted my plugin in one of his videos! [YouTube video](https://www.youtube.com/embed/1R1Pq_41wFw) #### GPT-3 Page The new `gpt-page` feature sends the entire current page to GPT-3, which should be more convenient for certain use cases where additional context is needed to make the suggestions more accurate. This feature makes it behave more like the OpenAI playground.  [Brian Sunter](https://twitter.com/Bsunter/status/1587513598896525312) @Bsunter Released a new feature for the [@logseq](https://twitter.com/logseq) [@OpenAI](https://twitter.com/OpenAI) GPT-3 plugin! The \`gpt-page\` command sends the entire page contents to [#gpt3](https://twitter.com/hashtag/gpt3) for completions It works across multiple lines, unlike the current \`gpt-block\` which sends one line Let me know what you think! [github.com/briansunter/lo…](https://github.com/briansunter/logseq-plugin-gpt3-openai) https://t.co/IUCWoPWCUt Show more [](https://video.twimg.com/tweet_video/Fgf7MPsVQAAg1EH.mp4) GIF [Nov 1, 2022, 6:35 PM](https://twitter.com/Bsunter/status/1587513598896525312) *** 62 6 #### DALL-E OpenAI released their [DALL-E](https://openai.com/dall-e-2/) image generation API. I recently incorporated this into the [Logseq GPT-3 plugin](https://github.com/briansunter/logseq-plugin-gpt3-openai), so you can generate images inside Logseq, which are inserted into the editor and saved locally. It’s nice to use GPT-3 to generate the prompt, then send it to [DALL-E](https://openai.com/dall-e-2/) within the same tool.  [Brian Sunter](https://twitter.com/Bsunter/status/1590174757651378177) @Bsunter Just released a feature to generate images with [@openai](https://twitter.com/openai) DALL-E inside [@logseq](https://twitter.com/logseq)! Write a prompt, then type \`/dalle\` to generate an image using AI, and it inserts the picture underneath the prompt. Let me know what you think! [github.com/briansunter/lo…](https://github.com/briansunter/logseq-plugin-gpt3-openai) https://t.co/uQ2KCkxz7p [](https://video.twimg.com/ext_tw_video/1590173762434977792/pu/vid/984x1058/bEtLIxt2Rs505T0k.mp4?tag=14) [Click to watch on X](https://twitter.com/Bsunter/status/1590174757651378177) [Nov 9, 2022, 2:49 AM](https://twitter.com/Bsunter/status/1590174757651378177) *** 97 8 #### GPT-3 Prompt Templates I’m working on a future feature to add “prompt templates” to the [Logseq GPT-3 OpenAI plugin](https://github.com/briansunter/logseq-plugin-gpt3-openai). GPT-3 is a powerful machine-learning tool that can generate text. You give it natural language text, called a prompt, and it predicts what text should come after it. The prompt can be anything from a simple question to a complex command. You must write the prompt in specific ways to get good results from GPT-3. Prompt templates will be a way of creating shortcuts for common tasks such as summarization or translation. They will behave similarly to Logseq templates, where you can define them in plain text anywhere in your notes. Since they’re just plain texts, users can share them with each other. I’ll eventually include these in a modal popup, like other AI tools. [See more thoughts on writing prompts and prompt templates here](/logseq-openai-prompt-template) ## Large Language Models Large language models (LLMs) are AI tools that can perform a wide variety of tasks. LLMs can be used for both search and classification based on the meaning of the text and generating human-like text. They were trained by analyzing massive amounts of text collected from the internet I think they’re interesting for several reasons: 1. You give them tasks by writing commands in English instead of code. To translate, you can ask the LLM, “Translate this text into French.” 2. They are very general and can perform a wide range of tasks, even things they weren’t originally trained to do. 3. They exhibit a property called “emergence,” where as you train them on more data, they gain new abilities that the creators didn’t anticipate 4. They can search and classify text based on meaning instead of just by keywords that appear in the text. I’m really interested in using this for note-taking to find “similar” notes, even if they don’t share any text in common, to uncover interesting connections and relationships. [See more of my thoughts on why large language models are interesting here](/why-large-language-models-are-interesting) ## Headphone Eyemask  image 1667858995655 0 The [headphone-eye-mask](/headphone-eye-mask) is my new favorite sleep hack. It’s an eye mask with built-in Bluetooth headphones, so you can listen to podcasts while falling asleep. I often stayed up too late on my phone or computer, overstimulating my brain and disrupting my circadian rhythm by beaming bright blue light into my eyes. The headphone eye mask helps slow down my brain with a nightly ritual I enjoy and helps me get to sleep earlier. I wrote about [how I use it and some podcast recommendations here.](/headphone-eye-mask) ## Alamack of Naval Ravikant I enjoyed this book called the [almanack-of-naval-ravikant](/almanack-of-naval-ravikant), which is a collection of philosophical\[tweets by Naval(https://twitter.com/naval) on wealth creation, learning, mindset, and health. It’s not a traditional book since it’s a collection of tweets, but there’s a good assortment of wisdom if you’re OK with the format. The book is completely free and you can [download it here](https://www.navalmanack.com/) [Read my review and notes on the book here](/almanack-of-naval-ravikant) ## Brain Food 🧠 ### Quote > Specific knowledge is knowledge you cannot be trained for. If society can train you, it can train someone else and replace you. [Naval Ravikant](https://twitter.com/naval) (from [almanack-of-naval-ravikant](/almanack-of-naval-ravikant) ) ### Link of the Week You need to try [ChatGPT](https://chat.openai.com/chat). It’s easy and free to get started. Just log in and ask it anything. It’s an AI chatbot that can do a huge range of tasks and is incredibly smart. Here are some examples of what it can do: It can write an A- college history essay  [Corry Wang](https://twitter.com/corry_wang/status/1598176074604507136) @corry\_wang I guess GPT-3 is old news, but playing with OpenAI’s new chatbot is mindblowing. [chat.openai.com/chat](https://chat.openai.com/chat) We’re witnessing the death of the college essay in realtime. Here’s the response to a prompt from one of my 200-level history classes at Amherst Solid A- work in 10 seconds https://t.co/z1KPxiAc1O  [Dec 1, 2022, 4:44 AM](https://twitter.com/corry_wang/status/1598176074604507136) *** 7.1K 153 It’s also good at coding  [Amjad Masad](https://twitter.com/amasad/status/1598042665375105024) @amasad ChatGPT could be a good debugging companion; it not only explains the bug but fixes it and explain the fix 🤯 https://t.co/5x9n66pVqj  [Nov 30, 2022, 7:53 PM](https://twitter.com/amasad/status/1598042665375105024) *** 6.2K 102 Some of my favorite examples are the weird, funny ones. Here’s one where it writes a Seinfeld script about Jerry learning programming  [Riley Goodside](https://twitter.com/goodside/status/1598077257498923010) @goodside OpenAI's new ChatGPT writes a Seinfeld scene in which Jerry needs to learn the bubble sort algorithm: https://t.co/yjPlrA3OmE    [Nov 30, 2022, 10:11 PM](https://twitter.com/goodside/status/1598077257498923010) *** 3.7K 70 ## Subscribe to newsletter I send occasional emails about new blog posts, side projects, and things I'm learning. By subscribing, you agree to our [Privacy Policy](/privacy). [Older Why Large Language Models are Interesting ](/why-large-language-models-are-interesting)[Newer Data Structures and Algorithms Guide](/data-structures-algorithms-guide) ## Related - [Central Pacific Update Sep 23, 2025](/central-pacific-update) - [ConvoCards Launch Retro Nov 22, 2024](/convocards-launch-retro) - [Cooking for Hackers Apr 8, 2024](/newsletter/issue-16) ## Share this article