Perplexity.ai has become an indispensable tool in my daily workflow. Its versatility allows me to tackle a wide range of tasks efficiently:
For blog posts and other written content, Perplexity.ai helps draft initial versions, which I then refine and personalize. My favorite refining use cases: condense text, bold the key points, sparingly bold key points, rephrase in layman's terms.
This blog post was drafted by Perplexity (using the Claude 3.5 Sonnet model) based on the outline of AI tools and how I use them. I also asked Perplexity to suggest human roles that are supported by the AI tool, based on my description of how I use them. I also asked Perplexity to sparingly bold key points in much of the text.
My takeaway from all this:
With AI making it so easy to produce polished content these days, choosing what to say has become more crucial than ever. As the barrier to content creation lowers, the importance of thoughtful, meaningful communication rises. It's about quality over quantity in this new landscape of effortless production.
Acting as a Research Assistant, it helps analyze and interpret complex grant calls, ensuring I don't miss crucial details or requirements.
Recently, I asked it to enumerate all the elements of a broad agency announcement required to be responsive to the call. This call had ~40 highly specific requirements, so using AI helped ensure I didn't miss one!
Perplexity uses RAG with retrieval of indexed web content including up-to-date scientific literature.
It's quite good at finding journal papers on many technical topics and citing them correctly. The citations aren't perfect though, so I do always check critical references.
Recently, I've started using it to interpret the logs from my time/goal tracker built using AppSheet, which talk about later. By analyzing how I've spent my time over the past week, it helps me assess if I'm maintaining a good balance between client work, business development, marketing, and operations.
I also ask Perplexity to summarize key insights and action items from meeting transcripts, which are automatically generated using the AI notetaker Fathom.video (mentioned later).
Gone are the days of googling an Error message while debugging my Python program, and perusing stack overflow for the accepted answer.
Now, I ask Perplexity and get code I can copy-paste into vim or Jupyter and execute.
It even generates full apps! (Claude 3.5 Sonnet has been great at this.)
In one project, we built a proof-of-concept microbial co-culture simulation platform to help identify growth optimizing strategies for their non-model organisms.
Our client had no dedicated computational biologists at the time, and really nobody to use the model directly. They relied entirely on us to simulate, interpret the simulations, and then help design experiments.
While this was nothing out of the ordinary, I felt my clients could unlock more with the model if they could access it and interact with it themselves.
I'd built web apps in Flutter before but wanted something in minutes, not days.
So, I asked Perplexity a series of questions that roughly went like this:
- Suggest some no code options to develop an interface for my simulations
- Compare the options, in this case open-webui vs. streamlit
- Write a streamlit app that shows a dashboard for my simulations
At this point, I already had a functioning interface that would load a metabolic network model, run a simulation, and display the results. Getting from the idea to a working app took ~20 minutes including installing packages.
Over the next five or so iterations, I asked Perplexity to update the code with more and more features, such as:
- Allow user to change model parameters: flux bounds, metabolites tracked in a dynamic simulation, the total bioprocess simulation time, the time step
- Plot the simulation results using an interactive library (plotly)
- Run increasingly complex simulations like dynamic flux balance analysis with protein constraints
Some screenshots of the resulting Streamlit App below.
While primarily a No-Code tool, AppSheets does incorporate AI features that assist in creating apps.
I use it extensively for time tracking, goal tracking, and as my CRM.
The app is built exactly how I want, so that I actually enjoy logging all my hours, and visualizing key metrics.
What's powerful is that as my company's specific situation evolves, I can easily update or add data sets and views.
Also powerful is how all the data collected can interface with the other AI tools.
For example, I recently started to do a weekly reflection to assess if my tracked time is consistent with moving the needle in my business goals. To do this, I upload a CSV file of my time tracker logs into Perplexity and ask for an assessment. The assessment is quite objective and includes some nice recommendations to reflect on.
Here are some screenshots of our internal time & goal tracking tool made with AppSheet:
As the founder and principal consultant of a bio-simulation consultancy, I've integrated AI tools into almost every facet of my business operations. This approach has significantly enhanced my productivity, streamlined processes, and allowed me to deliver more value to my clients. Here are some of my favorite AI tools.