Agree with @Solvetheriddle and @mikebrisy on this, I’m a big fan of NotebookLM (NLM) and probably use it more than any other AI tool.

It’s a ‘grounded’ model so it only runs off the data that you give it – not whatever is on the internet like typical (open) LLM’s.

It’s great for consolidating and synthesising a lot of information.

Example – load a few years of call transcripts and ASX presos and annual reports, then ask it where management has changed their strategy, messaging or KPI’s, hit or missed targets they set for themselves, etc, etc.

It’s also multi modal in both directions.

So you can input PDF, audio recordings, URL’s, etc, etc.

You can then output video, audio, slides, infographics, reports, etc, etc.

You can just generate these as default outputs but I think you need to customise them for 2 reasons. 1) To set the tone, or it’s just to light, confirmatory, hokey, American for me. 2) to fit your purpose.

I tend to use standardised prompts like:

I am looking for a careful, critical, balanced analysis including pros and cons, and a weighted assessment of these, not a glowing sales pitch.

Keep it professional, to the point, objective, incisive, analytical and evidence-based, focusing on practical application of the main concepts, frameworks, models and insights.

I customise all audio and video outputs with this at least.

I don’t need to use the standard LLM prompts like “Don’t make anything up, take your time, think through this deeply and logically, etc, etc.”

I find the audio (podcast) output to be particularly useful. Comes in 3 sizes short (5 mins) default (15 mins), long (40mins).

You can even interact with it – when it says something interesting you can pause and ask it to say more about that topic or challenge it.

Or you can download the long version ask it to generate a transcript and then listen at 1.5-2x while reading along.

It’s like any AI tool, the more you use it and iterate how you interact, the more you get to understand it and improve the value you get from it.

I’m on the paid version (outstanding value) but I think you can use a free version too.

Enjoy!

USING NOTEBOOKLM AS A USEFUL INVESTMENT DECISION-MAKING SOURCE

The following may be old news for some, but it may be useful for others. The following is my journey. Must thank @Slomo for the heads up here. Over the past while, I have been building a research library on NotebookLM (NLM). NLM is part of the Google stable, and I think it is very well-suited for research.

When I moved from professional investor to retail punter, one of my concerns was the quality of input information. I was truly shocked at the apparent extent to which people relied on social media feeds from unknown and/or untried sources. Something I was quite reticent to do. Taking a Google search, or a Twitter comment, etc., as a source of truth for a large investment decision is quite a hurdle for me (still is).

So what's going on here?

What NLM allows you to do is set up your source of truth by allowing only the information you have vetted into the library. That allows you to handpick the stuff you want the commentary output based on. For instance, I have spent the last couple of weeks filling hundreds of documents into my library. They are mainly result call transcripts, company presentations, and research that I thought were particularly insightful. Of course, the produce is to a large extent dependent on the inputs. I also have my financial models as part of the library. One issue I have found is that some research is protected; maybe we need a workaround here.

NLM is arranged by individual notebooks, i.e., they don’t talk to each other unless you summarise from one and paste into another, which you can do. Below is part of my current notebooks. Most of these have enormous amounts of content. I have settled on industry notebooks, thinking that there is some utility in that without being too narrow or too widely focused. My “Models notebook” below is a general notebook where I can ask it to look across all my models spreadsheets for trends or screening. One goal I have is to coordinate my investment philosophy and portfolio with the models and other data to critique my portfolio for consistency. There is a wide range of possibilities to use this structure.

What has become apparent to me is the value and skill involved in questioning the data source or prompting. Various sites can help with improving your skills here, and how to ask the best questions, which matters. Asking simple questions can get simple answers. I have also found that AI is not that intuitive, and outlining the scope and nature of your query helps. thinking about my past interactions with various analysts over the years, maybe they weren’t too intuitive either, lol. Eg, telling NLM to be sceptical and search for changes in narrative over time from management is interesting. It is worthwhile spending a bit of time thinking about what you want and how to get it. I suspect you can improve a lot with practice.

The other big positive I see from this is the reduction in wasted time and effort. We only have so much energy and time to put into something, and I have found much of my time and effort gets consumed in collecting, collating and coordinating data, or going over reams of stuff looking for one gem. NLM has the potential to turn this inefficiency on its head; instead of 80% of the time being used searching, collecting and organising, I can spend 80% of the time analysing and thinking about responses, a big improvement.

Google has now combined NLM with Gemini, its LLM. I have just started to use this feature, and it brings the wider web into the database. That can be useful for specific issues, although im just starting to scratch the surface here. It brings much wider content, but at a lower compliance or quality check. You can easily think of possibilities of its use.

Properly used, NLM is like having a trained junior analyst. In fact, that’s what I named the Gemini “Gem” my trusted junior analyst. There is likely a long way to go here, and as Slomo says, better to be on the train early.

I can already see that I will spend more time talking to NLM and less time on other sources.

All the best.