This post is part of a series where I talk about technologies and make guesses about how they’ll deliver value in the future.

One of my goals in life is to be a great technologist – an expert at identifying technology that can deliver value and improve the human condition. I hope this blog series will help me with that.

Updates Since Last Time

Summer 2024 was my most recent post in this series. I don’t expect everything I write about in this series to change in so short a period. Things that matter tend to take decades to be proven right or wrong. That said, I’d like to get in the habit of checking in to see if my previous statements are revealing themselves correct or if there are things to learn from.

• Brain interfaces – since I last discussed this we’ve gone from one human with a confirmed installed inner-brain device to three – 200% growth! Remember, there was VR in the 1980s and we still don’t have mainstream adoption 40 years later, so it is possible patient 3 and patient 3 Million have a lot of space between them, but I think this is a trend with reason to tip quickly.
• Autonomous Driving – I was running over some numbers recently at a dinner with a few bankers to see what the breakeven on a Waymo vehicle would be. The napkin math says the unit economics aren’t there yet. The main factors in the equation are the cost of paying a human driver, the cost of the vehicle and the carrying cost of capital. All three of these figures can change over time though;  human labor is getting more expensive in the US, cost trends on the cars might follow Moore’s law and carrying cost of capital could change again – in either direction. It will be interesting to watch these three trends.
• AGI – we’ve seen some very visible signs of people backing off their expectations of AGI or trying to move the goalpost. I suspect as we get closer to having computers that are ‘intelligent’ we’ll learn more about what human intelligence actually is. The more I learn about the human body and brain, the more I think we perhaps have more FLOPS than we’ve classically assumed.
• 3D printing – The children’s museum near me had a 3D printing exhibit that I had a chance to get hands on with. There are a few cool displays of progress there like flexible filament and different printing patterns, but largely what I saw was interesting trinkets and a reject pile of mistakes from a process that still has a super high error rate. I haven’t seen the killer app yet, the thing that makes it worth it for the average person, or even some specific 5% of people to 3D print at home. It remains a hobby for technologists that like to tinker as a way to spend time. One of my engineering counterparts printed me a very cool trinket for my office recently though and I suspect they are closer to the 5% than the 1% in terms of desire to tinker – so perhaps there is slow progress there. What does the usability need to be at to have a huge impact? 5% 10% 50% 90% I suspect it varies based on a number of factors, but I’m not sure what those are – clearly the network effects of social media and smart phones were helped by hitting >50%. Does 3D printing need that or is there a critical impact mass at 5%?

Tech I think is undervalued

• AI for solving simple work behind common problems.
  • Most AI use in businesses so far has been proof of concepts. The three use cases I’ve heard about having real tractions are: reducing customer support costs, increasing programming efficiency increasing design options for non-critical design tasks.
  • I suspect there is a wave on the verge of being unlocked though. There are likely some companies that have a defensible position and a cost profile where AI can soon be good enough to lower their costs without allowing too many new people to enter the space.
  • What comes to mind to me is something where creativity is important but not critical.
    • I’m struck by the Godaddy website builder. Their revenue is tied to hosting but some level of creative work is a bottleneck. Automating the creative work would increase their SAM, but automating the creative work might not be enough to let another competitor disrupt them as they have a lot of value in other factors
  • I wonder if cell companies will see a boost as well. They are going to be hit on both ends – increased volume of data transfer from more network use as data from models flows across the wires. Meanwhile one of their largest cost areas – customer support via call centers – will be reduced. Is that enough to make those companies attractive places to find return or is it a race to zero?

Tech I think is appropriately valued

• EVs
  • I think the recoil we’ve seen lately is appropriate and important to us finding an approach that lasts. When a technology decision becomes a moral one, it is easy for the costs to separate from the value and I believe long term we need the two closer together. I’m optimistic about EVs long term.
  • That said, a two ton car is never going to be an efficient way to transport an O(200lb) human. Even seven humans. I’m encouraged by the progress on electric scooters and bikes though. I don’t think most people understand how important the bicycle is in terms of impactful technologies of the past thousand years. Getting a battery and some efficient solar on it could increase its impact eve more.
  • I’m super bullish on the idea of a solar-roofed tuk tuk-like vehicle with pedal backup as a way to transport most things with a more effective carbon footprint. The current way we do things just won’t scale to 100B humans, and so you need inventions like this to keep scaling. We will scale – we always have – and that is likely a part of it.

Tech I think is overvalued

• AI – specifically it is overvalued in the form of investment in most companies with AI in the title
  • There are real JTBD there but that doesn’t mean every company that attempts it will have a positive ROI
  • I’m reminded of the days of webvan, a company which focused on a CUJ that we still believe is really valuable, but 25 years later haven’t fully solved. Nonetheless, the owners of webvan saw no return from that investment. Technology is one thing, but effective management is another.
  • I’m personally very bullish on the middle layer of the stack for B2B AI (Databricks – my employer) and the consumer interface for B2C AI (Perplexity, Glean and Google). I think many parts of the stack will be commoditized
• Foundation models
  • I suspect the world is coming around to this viewpoint more in recent weeks with DeepSeek. I think there are still an order of magnitude or two of cost reduction that is possible though.
  • Eventually you won’t need a lab with millions of investment to achieve those results, we’ll have a kid in a garage do something really cool that shocks the world. Of course we will, that is the trope in technology. I wonder if that kid will be in Silicon Valley this time though – more on this below.

Tech I believe in, but think we’re far from

• Voting / investing in ideas via coins
  • Funny enough, I think the $TRUMP coin is actually one of the best things to come out of the crypto space.
    • To be clear, I think it is 100% unethical for a sitting president to have such a clear bribe vector. There is a reason there have been rules about presidents managing their wealth while in office and this very clearly seems to violate the spirit of that.
  • The idea behind $TRUMP it is really solid though. Most things in life of a certain scale are a collective action problem and historically these were solved by having some powerful entity act as an intermediary. Over time most people have grown to distrust powerful entities but we have even more and even larger collective action problems to solve. How can you reconcile those two things? You need some form of reward you can trust. Does crypto provide that? Not yet. Can it? Maybe
  • The $TRUMP coin is actually a great way of expressing numerically something that already existed in the human social brain and I think there is value to that. Databases and spreadsheets have proven to be two of the most valuable technologies of the last 50 years and I believe those are just putting structure to something we naturally do in our head. I think crypto as a representation of trust, confidence, demonstration of commitment, etc. could become a societal primitive with the impact order of databases. It isn’t $TRUMP, but I think $TRUMP is close.
  • I suspect 50 years from now there will be a common concept that can trace its roots back to $TRUMP the same way we talk about DARPAnet when we talk about the history of the internet (though we rarely do that because now it is just prevalent and invisible, like air)

Tech I think is close to disruption

• The United States
  • There will always be another wave of technology and I suspect the next wave it is just as likely to come from Zhengzhou as it is from a place in the United States.
  • There is something about constraints and human nature that makes it more likely for innovation to come out places in a goldilocks zone that isn’t the leader of the current wave, but that has enough investment in infrastructure to make certain things possible.
  • Silicon Valley in the 1980s came out of a generation that grew up the children of a generation funded by the US government defense funds. Enough brains congregated and just enough money to tinker, but not so much that they didn’t have constraints. A well-funded IBM competitor would not have invented the PC – like Apple did – it would have spent money on a bigger and better mainframe. Will the idea that finally disrupts Google search come from a startup with a billion dollar seed round or something that finds product-market-fit in a garage and then scales?
  • Zhengzhou is interesting as there is enough investment and expertise that something interesting can happen there, but enough constrains to force some bold attempts. Those don’t always work out, but when they do, they are the seeds that turn into a sequoia.
  • Perhaps the transformative technology of 2050 will come out of a generation of kids who were the children of parents funded by the development of the iPhone