For most of history if a ruler wanted to either avoid a violent conflict or cement an alliance there was a standard protocol. Marry off the children. Alas times have changed. There is precisely zero chance that US government leaders are discussing the possibility of uniting Hunter Biden and Xi Jinping’s daughter – Xi Mingze – in an effort to improve US-China relations.

I posit that the closest modern-day equivalent to marriages of state (diplomatic marriages) is economic integration. Trade deals. Foreign direct investment. Apple building their iPhones in China. It might surprise you to hear that US imports from our arch geopolitical rival will likely hit a new record in 2023. See below chart showing import data through 2022:

This is a good thing!

The chart above – while impressive – is in one sense quite misleading. China’s role in the global supply chain has evolved dramatically over the past decades. They don’t just make cheap shoes and steel anymore…

Apple had nearly $400 billion in revenue last year:

According to 9to5mac – an Apple focused blog – more than 95% of iPhones are assembled in China. The figures for Mac, iPad and Wearables are probably similar. And, without the products assembled in China there would be no services business. Hence it’s not much of a leap to say that virtually all of Apple’s revenue is enabled by China.

Going back to our chart above US imports were nearly $540B in 2022. The right way to think about this figure though is that it is the visible tip of the iceberg.

I’m going to introduce two new terms:

China-dependent revenue (CDR) and China-dependent value (CDV)

I think it’s reasonable to consider all of Apple’s revenue as being China dependent. 95% of hardware is made in China. Probably 99%+ of services revenue is generated by users of their hardware. The remaining 5% of hardware that isn’t made in China is still reliant on the strength/R&D made possible by the rest of the business.

Net-net we have ~$400B of China dependent revenue from just one company.

The next obvious low-hanging fruits to make my point are Amazon and Walmart. It was hard to find estimates of the percent of products sold on Amazon and Walmart that were made in China, but the figure is at least 50% and could be much higher. For reference, electronics are the largest category on Amazon and 90% of computers are made in China.

Amazon’s gross merchandise volume was $600B last year, and Walmart’s non-grocery sales in 2022 were $268.84B. If we add the figures and then attribute 50% of it to China then we come up with another $434B in China-dependent revenue.

To summarize, just these three businesses have $834B of CDR - starting to make the import figures less impressive eh?

Now let's look at CDV.

The reason I'm thinking about this at all is because of the ubiquitous discussions around the potential for a China-Taiwan conflict. In the event of such a conflict it's likely (obvious even?) that

China would be able to cut off Taiwan's exports to the rest of the world by setting up no fly zones and encircling the Island with their Navy. Hence, I think it's fair to lump China and Taiwan together when considering CDV and CDR.

Without China there is no Apple. That much is obvious. So, we can attribute 100% of Apple's $2.38T valuation to our CDV category.

Amazon probably has between 60 and 75% of their market value attributed to AWS. AWS would not exist without chips from Taiwan, so we can add between 60 and 75% of Amazon's market cap to our CDV category right off the bat.

If we use the percent of retail revenue derived from China-based products to calculate the CDV for Walmart and Amazon's retail business then we come up with $92.5B for Walmart and a range of $119B to $190B for Amazon (depending on percent of their valuation attributed to AWS).

Net-net, between the three companies we have the following range of CDV: $2.59T to $2.66T

Just from three companies...

There are some caveats. Amazon could buy Intel chips after Intel had ramped production. Apple would over time be able to move manufacturing out of the country. Vendors on Amazon would find new sources for their wares. But this would not take months - it would take a decade +, and the end result would be more expensive products and less profitable businesses.

It all leads me to believe that if there is a conflict - and even if it is violent - it will be short lived because the global value destruction is too immense and would benefit neither China nor anyone else. Beyond that how things play out is anyone's guess.

Is to buy a business. It's also the most common way (after starting a business) for people to hit $10mm+ net worths.

I'm going to weave in commentary that is relevant to stocks in some of these posts. The first thing I want to mention related to buying a business is that in my personal experience business owners are less flustered than other people by the gyrations of the stock market. This is not universally the case, but it is certainly the case with a solid majority of folks I've observed (also FWIW this is NOT true for people in the real estate business).

I think a lot of this comes from knowing/feeling intuitively how ridiculously insane it is to think that the value of a business - especially one as large as Google - fluctuates by 20% in any given year (40% in the past 12 months) and can drawdown by far more. Business owners have all had bad years where their profits took a dump - literally all of them. I don't think I've ever met a single business owner who didn't have horror stories. None of them were thinking at the time: "Gee, now that my profits have dropped by half I sure as hell better sell my business before it goes to zero". They instinctively know that you don't sell when times are tough in a way that is super difficult to internalize for people who haven't owned a business.

There are also huge tax benefits to being a business owner that are relevant to investing in stocks as well. For example, as a business owner there is tons of flexibility in when you realize accounting profits - and in many cases you can engineer accounting losses outright. This would enable you to do things - for example - like rollover a traditional 401k to a roth and offset or remove outright the tax bill for the conversion.

Below I will run through what is not only plausible, but what I'd call a reasonable base-case outcome of purchasing a small business using seller financing and an SBA guaranteed loan. The net result is a high teens CAGR over 10 years - something only a handful of professional asset managers will achieve over the same period (out of literally thousands). Now on to the post.

Buying a small business.

The SBA has a program that guarantees a certain percentage of qualified loans banks make to small businesses. For loans above $150,000 they’ll guarantee up to 75%, and if my memory serves this is available for loans up to $5m. The bank and SBA like to see 20% buyer equity in the deal and do permit seller financing so long as the seller’s debt is subordinated. Seller financing is incredibly common, and my recommendation to anyone wanting to buy a business (with some exceptions) is to use it even if you don’t need it - it’s a good thing for the seller to keep skin in the game while you are learning/taking over the business (I’ll discuss this in further detail below).

Here’s the numbers we’re going to work with:

Purchase price: $5,000,000

Down payment: $1,000,000

Seller financing: $1,000,000

SBA Guaranteed loan: $3,000,000

Closing costs: ~2% of purchase price or $100,000

Diligence costs: at least $25,000 (you can do diligence on your own but I think it’s worth paying for)

Interest rate: ~9% today but I’m going to use 7% because that’s probably where they’ll be in the not distant future

Loan terms are usually 10 years, and I’m going to assume seller financing is the same rate as the bank financing (common). This gives us a monthly debt payment of $46,443.39 or $557,321 per year. I just googled it and saw that most lenders require business profits to be at least 20% higher than debt service - though in my personal experience most acquisitions have the debt covered by 30% or more unless they’re a stable real-estate related deal.

Businesses in the $5m valuation range usually sell for around 5-7X EBITDA (though there can be material adjustments depending on assets). You’ll see higher valuations if the business is growing/stable (margins, sales, etc), doesn’t have insane levels of customer concentration, and if it has management in place that could theoretically run the business without the owner (businesses that can operate without the owner can be owned passively, which dramatically increases the amount of demand). You’ll see lower valuations if the inverse is true or if you’re talking about less stable businesses like restaurants.

I’m going to use 6X because I want to keep the example simple and so will assume this is a company that can be owned passively.

6X EBITDA means $833,333 of EBITDA for a $5,000,000 purchase price. EBITDA is a good proxy for pre-debt/tax cash flow when it comes to small businesses - so in our example above we’d have the following:

EBITDA: $833,333

Debt service: $557,321

After-debt pre-tax cash flow: $276,012

After-debt post-tax cash flow year 1: $263,531 (I’ll explain how I got this below)

Now I’m going to run through the ten year return using two different scenarios. However, rather than using EBITDA I’m just going to use cash flow because it’s a more useful way to think about things when capex/investments/new hires are being made (very common after an acquisition).

In the first scenario I’m going to assume no change to cash flow over the entire ten year period and no change to valuation at the end.

Then I’m going to run through a more likely scenario where cash flow takes a dump in the year after acquisition as investments are made (I’ll explain below), but then starts climbing and hits a new high by year 5.

Scenario 1:

Cash out: $1,000,000 for the down payment + $100,000 of closing costs + $25,000 of diligence = $1,125,000. In practice we should probably add another $50,000 because the acquirer will likely have to quit their job in advance of taking over the business and there is an opportunity cost to the time spent before the acquisition closes, but I’ll just stick to the $1.125m figure.

Cash-in = after tax cash flow + exit value (assuming a sale at the end of year 10). In order to calculate cash flow we have to calculate taxes.

A simplified formula for calculating taxes is the following:

[EBITDA - Asset Write Offs - Interest] X 20%

When you buy a business through an asset acquisition you get to write off the entire purchase price against your operating income - usually this is amortized over 10 years - so you effectively get to reduce your operating income for tax purposes by 10% of the purchase price every single year. You heard that right - it’s insane and I’ll go into more detail below, but here’s the calculation:

EBITDA ($833,333) - Asset Write Offs ($500,000) - Interest ($270,928) = $62,405 - this is the figure you pay taxes on.

$62,405 X 20% = $12,481.

What we now have for after tax cash flow is: $276,012 - $12,481 = $263,531.

Pause a moment to observe how beautiful this is. You invest $1,125,000 and put $263,531 into your pocket in year 1. However, it gets FAR better than that. You also paid back principle of $286,393 using profits from the business. So - you really pulled in: $263,531 + $286,393 = $549,924.

Crazy right? Like hard to fathom crazy…Our tax code is designed to reward business owners and risk takers - thank Uncle Sam.

The after tax cash flow actually drops every year assuming no change to EBITDA due to an increasing amount of principle being paid (and hence less interest to write off), but the effect isn’t huge. Over ten years (I did the calculation) you would end up taking out: $2,408,100.

Then, at the end - assuming the business is worth exactly what you paid for it, you also have $5,000,000 less the broker’s commission (some people try to sell themselves but I recommend against this, in practice you probably aren’t selling for exactly what you got it for, and good brokers will be able to more than make up for their fee by getting you a higher price). Still, let’s assume the sales commission was 7% - you end up with: $5,000,000 X .93 = $4,650,000

Now, remember that you amortized your cost basis, so this is going to mostly be profit taxed at a long term capital gains rate of 20% (you will have some basis from new capital investment, but I’m keeping this simple and conservative). $4,650,000 X .8 = $3,720,000.

So, total cash out over the period is: $3,720,000 + $2,408,100 = $6,128,100.

We can now calculate our compounded growth rate over the period:

[$6,128,100 / $1,125,000] ^ (1/10) - 1 = 18.47%

Remember - that is your AFTER TAX return. There are literally a handful of professional money managers that perform that well over 10 years - out of tens of thousands - and unless you’re ultra wealthy you have no chance getting into them anyway.

Scenario 2

I’m not going to run through the second scenario in the same amount of detail, but in practice there are probably going to be two big differences between what I outlined above and what happens in practice.

#1, It’s pretty common for businesses at these valuations to be under capitalized and/or have an owner/operator who still makes it rain. Hence, it’s also common to need to upgrade equipment/machinery and/or hire a new employee (or even two). It’s also pretty common for new owners to want to make sure they keep the employees there - especially during the first year of transition - losing a key employee can be a nightmare, especially if they’re involved in sales and have the potential to bring business with them if they leave. Hence, employees often get a nice pay bump after a new owner takes over.

#2, the business will probably grow.

The result of #1 is that the business will likely take a big hit to cash flow in year 1 and then gradually work its way back to where it was by say year 4-5 (being conservative), and then start growing afterwards.

You’d be amazed how many businesses of this size operate with little or no marketing/sales and have websites that don’t drive any business.

I assumed a $100k hit to cash flow in year 1, then assumed it got back to it’s starting point by year 5, and then grew at 5% per year thereafter.

Using these assumptions EBITDA grows to $1,063,568 by year 10 from $833,333. It’s entirely possible that this business would command a higher multiple of say, 6.5 (up from 6).

This gives us the following:

Cash out during 10 year period: $2,718,708

Cash out from sale: 6.5 X $1,063,568 = $6,913,192 X .93 (broker fee) = $6,429,268.56

Using conservative assumptions about cost basis again (leaving it at zero) - we have $6,429,268.56 X .8 = $5,143,414.85

Total cash out: $5,143,414.85 + $2,718,708 = $7,862,122.85

Our compounded growth rate is hence:

[$7,862,122.85/$1,125,000]^(1/10)-1 = 21.46%

Final hypothetical for this section

There are obviously many cases in which a new owner comes in and increases profit right from the beginning. This is particularly true when the business is being operated with a godawful web presence, has no marketing budget, and no sales force (again, a very common scenario because in most cases the owner is who drives business at companies with these valuations). If we assumed 10% EBITDA growth in year 1 and 2 as the sales engine comes online and then 5% thereafter once the easy pickings are gone, we end up with a business doing $1,418,826 in EBITDA by the end of year 10. Since I’m working on an optimistic case here, let’s also assume this business is running beautifully with a completely passive owner, and has some asset or IP that makes it attractive to a strategic buyer who wasn’t previously interested either because they didn’t know about the company (because of no web presence, hard to find, etc) or because it was too small. So, let’s say it now commands a 7X multiple.

We would then have the following:

Cash out during 10 year period: $4,844,380

Cash out from sale (not showing work): $7,389,244

Total cash out: $12,233,623.95

Total CAGR over the period: 28.46%

Two facts to close on.

Most millionaires in the US are tradespeople who own their own business. Painters, roofers, plumbers, etc.

The vast majority of people that achieve a $10m+ net worth (which usually excludes higher earners that work for a salary, e.g. lawyers) - did it by starting or buying a business. Starting a business is far harder than buying one...

My next post will also be on the same topic, but will instead use anecdotes of real businesses to explain possible return outcomes, downsides, considerations when choosing a business, what to look out for when doing diligence, and anything else I can think of that is interesting and related. I’ll also use that post/anecdotes to talk about the differences between buying a business for $1,000,000 vs. $5,000,000 vs. $10,000,000 (they are huge).

As a teaser, one of the anecdotes I'll explore is buying an asset management business - specifically a registered investment advisor (very very interesting w the highest returns and also highest risk - but definitely ways to mitigate the risk which I'll discuss).

I thought it would be fun to write some posts about all of the different types of leverage I’ve run across over the past fifteen years. I’m going to use real anecdotes - most of which come from my life as a consultant or from things I've seen friends do.

When I say leverage, what I mean is opportunities to exponentially multiply the value of time.

Einstein supposedly said “Compound interest is the 8th wonder of the world. He who understands it, earns it…he who doesn’t…pays it”. When people hear this quote my guess is that their minds immediately jump to stocks (especially on Commonstock/Fintwit!). And when people think about leverage their minds usually go to using margin or options to lever up a stock portfolio. Less often but still frequently real estate probably comes to mind.

But the world of compounding – which is insanely magnified by leverage (as I defined it above) - is so much vaster than public securities and real estate. And as I’ll show in my next series of posts the potential returns of other forms of leverage are in many cases far higher and less risky than Yoloing options.

There will be 6 topics in this series (though I may combine some into the same post), each looking at a different form of leverage.

• Buying a business. I’m going to run through some napkin math on return potential from purchasing a business for $5m with an SBA guaranteed loan and seller financing (most common structure for small business acquisitions in the US). As a rule of thumb you're going to need 20% equity in an SBA guaranteed deal - so that's $200k for a $1mm business and $1mm for a $5mm business. In my life as a consultant I’ve probably looked at over 50 different businesses in the $1mm-$10mm valuation range, and done enough work to get smart on maybe 15-20 (usually because a transaction/financing was taking place), so I’ll also use this as an opportunity to comment on some of the things to keep in mind if you’re wanting to explore buying a business yourself. I’ll also comment on some of the incredible tax benefits – if you aren’t already familiar with them this will really blow your mind.

• Owning a business. In this section I’m going to do two things. First I’m going to provide some interesting examples of operating leverage. Second I’m going to provide some examples (real anecdotes) of how owning a business gives you optionality, and will over time multiply by orders of magnitude your opportunities to “get lucky”.

• Having a following (social media, newsletter, etc). The reason I started my Substack was because I kept seeing examples of how effectively (and profitably) people monetized followings. I have no idea at this point how I will monetize my Substack, but it has already provided enormous value and introduced me to awesome people, and as I’ll explain in this section is surely giving me lots of optionality for the future. I’m going to discuss five monetization methods I’ve seen: 1) Selling products; 2) Selling subscriptions; 3) Selling advertisements; 4) Fundraising (e.g. to make one or a series of investments, start a fund, etc); 5) Selling services. One particularly interesting anecdote I'm going to run through is a Youtube creator who started an ecommerce business, grew it to multi-million $ revenue in the first year - and who has ridiculous profitability due to having effectively zero customer acquisition costs (CAC), and discuss a bit about how creators have far more pricing power than faceless products.

• Digital advertising expertise. The anecdote for this section is ecommerce rollups – which are a particularly hot space for HNW individuals right now. If you are an ad-guru you can monetize that expertise at a firm or as a consultant and get paid hourly/salary. You can start a marketing firm (more leverage than hourly, but it’s a competitive space and definitely doesn’t offer the most leverage possible). Finally, you can use those expertise to raise money for a rollup/aggregator. I’ve now run across multiple examples of companies who started out as small agencies but who have transitioned to having most of their business focused on buying ecom companies (essentially turning into small PE firms and making tons of $$ in the process)– I expect this trend to continue at an increasing rate over the coming years in no small part due to changes in the ad market (especially Amazon's). Becoming an ad-guru is not easy, but neither is it inaccessible to most people. It's something you can start working toward part-time and even monetize as a side hustle.

• Real Estate. In this section I'm going to provide a brief overview of at least 3 different types of deals that are less well known. First, change of use – essentially buying a property that is priced as if it’s one thing (farmland or forest say) and then turning it into something else like home lots or commercial/mixed use. Second, sale-leaseback – specifically buying a vacant (and hence underpriced building) and then using your own business to occupy it – signing a long term lease with yourself and then selling the building at a market level cap-rate. Third, conservation easements - essentially you can contractually obligate yourself to never develop a property (though you can still use it for other purposes like hunting, recreation, etc), and in return receive enormous tax benefits - this is something developers and even individuals do when they buy some land but don't want to use the whole thing (for example you want to build 10 houses and have some nice green space around a pond - might as well monetize that green space if you're not going to build on it)- and is a way to immediately pull out capital in the form of tax credits.

• Being a financial advisor as a side-hustle. My view is that many people who are at least moderately financially savvy would be well served getting their series 65 (the license you need in the US) and becoming a financial advisor. As I’ll explain with some basic napkin math this is something that has incredible long term compounding potential, interesting optionality, and can be done with only a few hours of work per week. Even if you absolutely suck at “selling”, you are virtually guaranteed to build a 6-figure passive income stream if you do it long enough - this may be hard to believe but it's absolutely true.

Going to get started on the first post now - if anyone has other interesting examples of "leverage" I'd love to hear them! Also happy to answer any questions about any of the above if anyone has them and wants to learn more, feel free to ask in comments or DMs.

I have two different but close figures from Vanguard and Blackrock. yCharts has a completely different figure, and Yardeni has yet one more completely different figure. Macrotrends is the most different from the rest, with an almost 3 point difference from Vanguard/Blackrock.

I'm curious if anyone has ever looked into the methodology of the different providers or if anyone has any idea why the answers are so variable? At the end of the day what I'm most interested in is directional accruacy - so I don't care if one data set is off from the rest as long as it is consistently representing the movement of the multiple over time.

Thanks in advance if anyone has any tips!

I've loved Railroads ever since I noticed their margins and ability to increase EPS year after year with flat revenue growth. Just playing with a model and coming up with some pretty attractive returns relative to what I'd expect from the market, with probably less risk given the nature of their businesses.

Just put a spreadsheet together that runs through some potential exit scenarios, here's a snapshot of the $UNP version (if anyone wants I'm happy to send you spreadsheet, just DM me).

Also, please let me know if I've made any stupid errors or assumptions. Basically, I'm assuming the annual shareholder return is equal to analyst forecasted EPS adjusted to FCF by taking FCF as a percent of net income over the past 5 trailing year average. Then creating a matrix to compare exit multiples and adding the exit to the annualized return from FCF.

Other interesting things to note: UNP returned 7.26% to shareholders last year through buybacks and dividend, NSC was at 7% and CSX was at 6.32%. All companies have reduced FLOAT compounded at 4.3%, 3.2% and 3.9% respectively. Dividend growth past 10 years has been 14%, 10%, and 8% respectively.

Also, was just listening to last quarter's calls and management seems to be confident in ability to increase prices more than inflation (and judging by financials they have been).

I'd love to read anything anyone has on the railroads, they look super interesting w/ tech margins and steady businesses. Only long term major downside I can think of is reduction in materials transport (short term from less coal? long term from more energy being generated by solar/wind that doesn't get trasnported?)

This will be a bigger improvement on UI than the switch from a Blackberry keyboard to a touchscreen (updated)

I’ve been Tweeting a lot lately and writing on Commonstock about how LLMs are going to turn voice into a 4th component of the “GUI”. GUI - which stands for graphical user interface - is in quotes because voice doesn’t fit neatly under the acronym. Still, when people think of the GUI they think of the “mechanics with which humans interact with technology”, and that does fit.

The QWERTY keyboard was invented in the late 1800s. The mouse was invented in the 1960s. The computer screen was invented in the early 1970s. Essentially, humans have been interacting with technology in mostly the same way for the past fifty+ years. The only major change worth noting is that we now use touch (primarily on smartphones) in place of a mouse.

Have you ever wondered why you can’t ask Siri to do things like “Check my bank balance”, or “Open my Southwest app and look for a plane ticket to Atlanta”?

Or, have you ever wondered by you can’t just ask Microsoft Excel to sort the table by column F instead of column J?

The answer is simple and it has nothing to do with the technological feasibility of connecting Siri to our apps (which is already possible and done in limited cases).

The answer is: because the user experience of using voice to navigate and interact with apps would have been horrendous. If you have ever used Siri, Alexa, or anything of the sort then you know how awful they are at understanding intent. While they can do simple things like tell you about the weather or set a timer, anything more complex results in them essentially passing keywords from whatever you said into a search box and saying “Here’s what I found”.

Again, let me stress this point because it’s super important: the technology already exists to use voice to navigate and interact with apps, what is missing is the ability of the digital assistant (e.g. Siri) to convert common language into understanding of what we want (which I call “intent”).

This problem has now been solved. ChatGPT already understands intent nearly as good as humans, and the new version coming out this year will be better still.

When you click with a mouse the computer knows exactly what you want. You want to open the app you clicked on, or edit a new line in a Word document, or move an image in a PowerPoint presentation.

The same is true with the keyboard. If you type in a web address and hit enter - you want to go to that webpage.

There is no ambiguity, the computer knows with certainty what it is you’re trying to do. This is what wasn’t true using voice commands - until recently.

Let me stress one more point: voice-to-text transcription technology exists today, people use it all the time. The error rate of transcriptions was already dropping fast, but ChatGPT has effectively dropped the error rate to near zero (I described how in a previous post). Essentially, even if a transcriber garbles up words the LLM is able to reverse engineer what the speaker was trying to say.

Digital assistants like Siri probably need to understand at least 90% of query intent in order for the user experience to be adequate enough to start connecting them to everything. After having used ChatGPT for many hours I’m confident that it and other LLMs to come will easily pass this threshold.

This connection will take place sooner than most people think. This is what Microsoft CEO Satya Nadella had to say yesterday at their Bing/ChatGPT integration launch event (I pulled the text from Ben Thompson’s most recent Stratechery post):

The first few lines in the top paragraph are the ones I want to call attention to, specifically: “All computer interaction is going to be mediated with an agent helping you”. Satya uses the term “co-pilot”. Not coincidentally Co-Pilot is also the name of the Microsoft product that helps programmers code.

Microsoft has already announced that LLM technology is getting integrated into the Office suite of products and Teams. They’ve also just announced that they will be releasing a new platform that lets any company build on top of ChatGPT’s underlying technology.

I think it won’t be long before people spend as much time using their voice to interact with technology as they do with the mouse or keyboard - and it won’t be long after that until voice becomes the primary medium of communication.

I’ll finish the post with three interesting takeaways.

As you can imagine, being able to navigate with voice will be an incredible user experience. It will be a bigger improvement than the switch from a Blackberry keyboard to a touch screen. As sticky as some products (especially Apple’s) may be, no company will risk losing users by not offering this new way to interact.

I ran across a company yesterday called Embra that is already working on building an app on top of ChatGPT for Macs that will let you query different apps on your computer. You will be able to ask it to do things like:

Search Google

Find text inside of a PDF (or summarize text from a PDF)

Search messages on Slack

Respond to emails

This app is text based, but as I explained above it’s only a matter of time before voice is connected.

That said, it is apps like this (in addition to competition from Google, Microsoft, etc) that will force Apple to release their own digital assistant (or upgrade Siri) sooner than later.

It’s one thing to have a chat interface with apps, it’s quite a different thing for voice to be the medium. It’s too personal and Apple won’t like it. They won’t ban apps like Embra, but they will release their own version that will effectively result in the same thing.

The big question is how long it will take before the upgrade is technologically feasible. Response times will have to be similar to what they are today even though the algorithm in the background will be having to work through more compute.

If anyone has any insights as to the technological details required to make this happen please reach out and let me know.

Or, at the least they are going to spend much more time sitting in comfortable positions.

Keyboards and mice require a desk to rest on. They require a body in a specific position.

Yes, I am aware of standing desks. I have a standing desk and in the past even had a desk with a treadmill underneath so I could walk while working. Still…

Standing desks don’t solve the problem of spending extended periods of time in the same position. Further, many people cannot - in fact - “walk and chew gum” at the same time. For example, I know that there are certain tasks I can do standing up - like data entry or having a conversation on Zoom. What I cannot do while standing up are activities that require a great deal of thought - like writing or doing complicated math. I’m sure this is not unique to me.

Voice is going to enable us to work in whatever position we want for an increasing portion of our day (as it improves).

There is another related takeaway here - I expect monitor sizes to increase. Monitor’s are the size they are because people are used to interacting with them up close. We’ll need bigger monitors if we’re going to be seeing what’s on the screen from further away.

I believe voice may be the catalyst for extended reality headsets to gain significant market share at the expense of monitors. Already the Meta Quest Pro offers a user experience far superior to having large monitors (at least for some uses to some people). I have a friend who spends hours a day programming with it, being able to sit in bed or in a La-Z-Boy while he works. Still, he must carry around a mouse and keyboard…

Apple’s releasing a headset this year which is rumored to be far more comfortable than anything on the market today, in large part because they detached the battery to lessen the weight. Apple’s headset will also use cameras that watch your hands so that you can use pinching motions to interact with apps. We’ll see how that goes, but one thing is certain - once voice is used as the primary input medium the user experience will be WILD.

Imagine walking around your home office (or even outside) with a mixed-reality headset. You can see everything around you so there is no danger of tripping over anything, but then right in front of you are giant monitors for you to work on. You might have one screen set to Teams so that you can see your workmates, another screen turned into a giant white board for collaboration, and a third screen connected to your newborn’s crib camera. No controller needed…

Today if I want to do research for a post I will spend 5 minutes Googling, open up a gazillion windows - and then get to reading - spending 1-2hours+

Maybe I see over the course of that process 15-20 paid links and click on 0-1 of them.

What if...instead of spending 5 minutes on Google and an hour on other websites, I spend a full hour on Google Search - just reading the information I was seeking out directly on search.

Publishers get screwed and Google loses the revenue from link clicks, but has more of my time available to monetize from display ads or idk?

It seems easy to imagine that time spent inside the Search window could increase by an order of magnitude - can Google monetize that big jump in time spent as effectively as the link clicks it will lose?

Poke holes!

Just spit ballin here, would love feedback/devil's advocates:

Impact of adding a 4th component

Excel monkeys like myself stand to save hours per week just on a single application. It’s not much of a stretch to imagine people using Word in new ways too. Rather than spend a bunch of time figuring out how to write the most concise blog post or memo, just ramble on as you would into a voice recorder, and then ask Word to convert your garbled mess into eloquent text (ChatGPT already does a GREAT job at this, it just hasn’t been connected to voice yet).

Programmers are likely to see the biggest change (and improvement) to their workflow/efficiency because they will be able to code without using a keyboard. ChatGPT has already transformed how people are programming, this will now remove the need for typing in prompts and copying/pasting results.

The less obvious

#1 - People are going to spend a lot less time sitting down in the future - or at the least they will spend much more time sitting in comfortable positions. Keyboards and mice require a desk to rest on, and they require a body in a specific position.

Yes, I am aware of standing desks. I have a standing desk and in the past had a desk with a treadmill track underneath so I could work while walking. But, there are a few problems with standing desks.

Standing desks still don’t solve the problem of being in the same position for extended periods of time. Sitting all day is worse than standing all day, but we weren’t meant to stand all day in the same position either.

People cannot “walk and chew gum” at the same time. For example, I know that there are certain tasks I can do standing up - like data entry or having a conversation on Zoom. What I cannot do while standing up are activities that require a great deal of thought - like writing or doing complicated math.

This will probably be intuitive to many of you - people actually think better when they are able to let their bodies move naturally. Sometimes we want to walk in a circle, stare into space, lean back in our chairs, or move our hands around. Using our voices as input will leave us free to move around in whatever way feels right.

#2 - Relatedly, using voice as an input will make VR and XR (VR + AR) headsets far more attractive. Apple may be the first to release a headset that you can comfortably wear all day, but it won’t be long until most headsets have solved the comfort issue. I have a friend who uses the Meta Quest Pro for hours a day - primarily as a replacement of his monitors. Today, he still has to lug the keyboard and mouse to his bed or La-Z-Boy - and even there using them is awkward. But that will change. Imagine being able to walk around using pass through vision so you can still see the real world, but having three giant monitors floating in front of you. You don’t need to carry around the mouse and keyboard anymore because you interact with the computer using voice.

#3 - Siri, Cortana and Alexa all get upgrades and start competing with Google search. Microsoft is integrating ChatGPT into Bing. Google announced Bard yesterday. As I’ve pointed out repeatedly on Twitter and in prior posts, Big-Tech companies do not want to let another company insert a product between them and how consumers use their products. Apple may be happy to let Google run the back-end of Safari, but they will not let the future “Voice” version of Bard or ChatGPT be the LLM based personal assistant that people use on their Macs or iPhones. My guess is that in the future when people use voice to query their digital personal assistants - the DPA will make an on-the-fly decision about whether to handle the query itself, re-direct to search, or do some combination thereof.

The biggest threat to Google’s search will probably end up being that the best search results end up mattering less than the user experience of searching. ChatGPT grew to 100 million users in two months not because it provides infallible results, but because it provides a unique user experience that is more natural by an order of magnitude than anything that came before it. This experience will be yet-further transformed by the introduction of voice.

On the flip side, it’s possible that Google - thanks to it having more data than anyone else - is able to create the best voice-based user experience. If it is sufficiently better then it may turn into their next billion + user product. Apple and Microsoft would both hate to have a Google product own the 4th “GUI” component, so they will invest heavily to make sure this doesn’t happen.

The only obvious thing here is that consumers will be the biggest winners.

The QWERTY keyboard was invented in the late 1800s. The Mouse was invented in the 1960s. The computer screen was invented in the the early 1970s. Essentially, humans have been interacting with technology in mostly the same way for the past fifty+ years.

Siri, Alexa, Cortana etc do not count because they offer terrible user experiences except for a limited range of tasks which usually involve very obvious keywords/instructions (e.g. “set a timer”).

Think about the answer to this question for just a moment: why can’t you simply ask your computer to do things?

As an experiment I asked Cortana (Microsoft’s Equivalent to Siri) to do five tasks:

I asked it to open Microsoft Excel (it succeeded, opening a new blank file)

I asked it to open a specific Excel file by name (it failed, and just opened a new Excel file again)

I asked Cortana to select a specific cell in an Excel file (it failed and said “I’m sorry, but I can’t help with that.”)

I asked Cortana to open a new Edge browser window (it failed, and provided an irrelevant list of instructions showing me how to turn on Cortana in Edge - see below)

I asked Cortana to open a new Chrome browser (it failed, and provided a list of instructions on “How to Make Cortana Use Chrome or Your Default Browser” - see below)

Back to our question: why can’t we talk to our computers today?

The most fundamental answer is: because they don’t understand what we want / they don’t understand common language to a degree that would enable a pleasant voice-based user experience.

Secondarily the answer is: because the software hasn’t been updated to connect voice commands to apps.

Crucially - the answer is NOT because the actual connection of voice commands to the operating system/software is difficult. The connection already exists. We can already ask computers to do things like set timers.

The second answer is only the case because companies hadn’t solved the first issue of teaching computers to understand our “intent”.

You can imagine how poor a user experience it would be to give commands to your computer and have it only be able to act as you had desired 25-50% of the time.

Contrast this with how a mouse and keyboard work. The mouse is able to accurately translate your “intent” into computer language 100% of the time.

You: Click on app

Computer knows: You want to open app

You: Click on new cell in Excel

Computer knows: You want to edit that cell

You: Click and drag a square on Powerpoint

Computer knows: You want to move the square

There is no ambiguity. Every time you use the mouse the computer knows exactly what it is you want to do.

The same is true with the keyboard.

You type: www.google.com

Computer knows: You want to go to google.com

You type (at least on my computer): Function + F9

Computer knows: You want to turn on/off Bluetooth

LLMs like ChatGPT are going to unlock voice as a new component of the “GUI”. I put GUI into quotes because it stands for graphical user interface, and voice doesn’t perfectly fit under that definition. Still, when people think of the GUI they think of the “mechanics through which humans interact with technology”. Using this broader definition voice is an obvious fit.

The vast majority of things we need to interact with on our computer are text based. There may be some icons without text, but image recognition software can already interpret common icons/images with near perfect accuracy.

There is nothing magical about the mouse or keyboard. They work because they can interpret signals with 100% accuracy and translate them to our computers. This makes for a pleasant user experience.

ChatGPT and other LLMs have largely solved the problem of understanding “intent”. They are just getting into their third month after launch, and subsequent versions will be able to understand intent as good or better than the vast majority of humans.

All that’s left is to connect them to all parts of the operating system and apps.

It won’t be long before you can just tell Excel: “Write a formula that shows a ‘1’ if the figure in a cell is positive, ‘0’ if it’s negative, and prints ‘n/a’ if the figure is text.”

You won’t have to know how to write a nested “if” statement - you will just be able to describe what you want and Excel will do the rest.

Based on the rate of progress I’m seeing - I expect we’ll have something like this by year’s end.