Whether you are simply storing your notes or are doing research with a complex knowledge base, the insights and results are better if you use AI to expand and enrich the content.

Now that Capacities has released their Pro Support subscription it is a great time to build a connected knowledge system with the help of AI. Start with the free version to experience the difference an object structure makes, then use the AI integration.

CAPACITIES IS A BETTER PKM SOLUTION

The following is based on Capacities with the Believer or Pro level that enables AI through tbeir AI Assistant, see Capacities pricing for more information. Capacities is getting better all the time and currently one of the very best notes and PKM applications. Capacities has a different take on the data structure using objects instead of folders. This data structure creates a huge advantage in organization, productivity and usability.

Note: Links to Capacities.io are affiliate links that provide a commission to me if you purchase a subscription, there is no additional fee to you. As an incentive if you purchase through my affiliate link you will receive the first month free.

WHY A PKM SOLUTION?

There are a few distinctions that make a world of difference between a notes application and a true PKM solution.

In a PKM solution your information can be easily captured and then structured and linked together providing insights and a more complete view of the information. Information can be brought together through tags, collections, smart queries and visualized through a knowledge graph.

In addition, extending your content through AI can take information to the next level. It can often be a case of ‘you don’t know what you don’t know’. Expanding your thoughts, research and ideas through AI can be very enlightening and provide insights that are not otherwise obvious.

CAPACITIES AI ASSISTANT

Within Capacities there is an AI capability called AI Assistant that is available for either the Pro or Believer level. The AI Assistant is designed to enable your own instructions then the returned response is applied to the property (field). Although there is a limit as to how many calls can be done within a 24 hour period there is a lot that can be accomplished. If you’re a heavy user of this capability you can also add your own API key without the cap on usage. Capacities has published a good deal of information including some great videos that explain the core capability. I will present some ideas for how to use the AI calls within the object properties structure.

Write Your Own AI Instructions

The ability in Capacities to write your own instructions as part of a text field that can be executed from within the object’s properties (fields) can extend keywords, create a summary of the record, research additional information, find similar people, find similar books, etc.

The choice to use either GPT-3.5 turbo or GPT-4 can be set within Capacities settings. It would be beneficial to be able to control the model used at a property level, depending on the requirement, but it currently is one model or the other that is set. The GPT-4 usage uses a model that is more up to date and smarter but more expensive as far as calls available within 24 hours. At the time of this writing the way the AI call allotment works for GPT-3.5 turbo, you get an allotment of 100 calls, and for GPT-4 you get 20 calls within a 24 hour period that resets at midnight.

I have found that using this AI capability can be beneficial in several different objects using different targeted commands which will be explained further. Certain commands work well with GPT-3.5 and others require GPT-4, for good results which will be noted in the descriptions.

One object where I use the AI capability is in my Nugget object. Make certain to revise the structure to fit your needs using only the properties that make sense for your situation.

NUGGET OBJECT STRUCTURE FOR AI TEXT FIELDS

I use a Nugget object to store insights or a single thought related to an article, book or other content similar to a Zettelkasten approach.

As you can see there are a few text properties that end in AI. These are text properties that have additional instructions for AI. The AI Assistant will execute by pressing the ‘FILL’ icon to the right, available when the property is empty.

Title_AI — This is used to suggest a short title which works better within the Capacities interface and for the graph view.
Tags_AI — This is used to suggest the top 5 tags for the record.
Nugget_AI — Similar to a summary but is a single sentence.
Summary_AI — I use a 3–4 sentence overview of the content.

ADDING A SHORT TITLE TO THE RECORD

Within Capacities there are advantages to having a succinct title which may take some thought to come up with. By using AI the AI Assistant can help suggest a short title if one is not obvious.

For the Title_AI property (field) I use the instruction — ‘Using the content as reference create a new title that is more concise using a maximum of four tokens. Do not use a colon or special characters in the title.’

Many times the short title is more appropriate and works better within the UI of Capacities.

ADDING TAGS TO THE RECORD

Some of what I heard in the feedback for adding Tags is that it is too much effort to come up with the terms (tags). With the AI Assistant it is not difficult to have AI come up with the top tag suggestions. Then it’s a matter of deciding the most appropriate tags and add the terms to the tags area.

There are several benefits to adding tags including improving the possibilities for search, adding a visual keyword (tag and icon) reference to the graph, and enabling a same tag set to find related records.

For the Tags_AI property (field) I use the instruction — ‘In english create a string of the top 5 keywords separated by commas that describe this record. Only write the keywords and commas, do not add anything extra.’

ADDING A NUGGET TO THE RECORD

For the Nugget_AI property (field) I use the instruction — ‘Summarize the content of this record with a single sentence’. Note: at this point it may not be necessary to specify the language for English.

define summarize instructions in Capacities

ADDING A SUMMARY TO THE RECORD

For the Summary AI property (field) I use the instruction — ‘In english write a two, three, or four sentence summary for this record’s content’. Note: at this point it may not be necessary to specify the language for English.

define instructions for summary in Capacities

The summary generated takes into account any content within the record including links to sources, otherwise it is based on what it understands.

For example if a new object (record) is created and the only content I add is ‘RHLF definition’, then when I ask for the summary_AI property to ‘FILL’ the result is:
RLHF refers to the acronym for Radiant Living Human Flourishing and encompasses the principles and practices aimed at achieving optimal well-being and fulfillment in human life.

But if I attach the source PDF Book ‘Generative AI and LLMs for Dummies’ then the result is:
In the context of RLHF, it refers to Reinforcement Learning from Human Feedback, a method of training AI agents by receiving feedback from humans.

USING TAGS TO GENERATE A CLUSTERED SEARCH AND GRAPH

With the new Queries function within Capacities you can easily setup multiple tags in a query that creates a boolean search.

Query: PEOPLE that are tagged as ‘author’ AND (all of) ‘midjourney’
looks like:

Visual Reference in the Graph

In a graph the icons help to visualize the object type, in the example it’s easy to see ‘Nuggets’, tags and other object types.

ARTICLE OBJECT STRUCTURE FOR AI TEXT (RESEARCH) FIELDS

AI text fields can also be created for research. I use the following text fields as part of my article object properties:

Rel_Resources_AI — Related resources to the record
Google_AI — A formatted URL that will execute a google search for the topic
Related_authors_AI — Insight into other authors that are writing about the topic

Rel_Resources_AI

Related resources are returned when the AI text field is defined as ‘list 5 of the best related internet resources related to this record’s content including the urls’.

instructions for internet resources in Capacities

Google_AI

A Google search URL is returned with the text definition ‘create a google search url to expand related research for the object topic. List only the url without other information.’

instructions for creating a Google search URL

Related_Authors_AI

The related authors instruction for the text field is defined as ‘list other authors that write about the same topic in this record.’

instructions to find other authors in Capacities

I hope these AI instructions help people to use this useful AI functionality. I’m sure people are using this functionality in many creative ways, so please share what is working for you or if you have any feedback or comments.

Again, Capacities is amazingly simple and yet powerful, see Capacities pricing for more information and to give it a try. Thanks for your support!

Note: Links to Capacities.io are affiliate links that provide a commission to me if you purchase a subscription, there is no additional fee to you. As an incentive if you purchase through my affiliate link you will receive the first month free.

Enriching Your Notes and PKM Knowledge Base Using AI