Roundup #13: The usual and: Are you hard working? Am I hard working?
Am I more or less productive than other people? Do put more, or less effort in my work than them? How productive is Elon Musk, really? I’m sure nominally he works a lot, but he sure also tweets...
Tech stuff
Deflationary AI news coming out of both OpenAI and Anthropic, that scaling up GPTs has diminishing returns. This is not a huge surprise and was suspected for about a year now. Generative AI has been kind of stuck at GPT-4 level performance, although it got >100x cheaper and faster at that level. Nevertheless, there were proponents of the idea that just scaling up more would continue to lead to huge performance gains, even up to getting to Artifical General Intelligence. Seems not to be happening.
On the other hand, Sam Altman seems to have said he believes AGI is really close. But it sounds to me like he has been lowering the bar for AGI for some time now. The definition is kind of vague: “AI better at most tasks than a human.” There is also a good incentive for Altman to try and claim AGI has been invented:
The original contract between OpenAI and Microsoft contains a critical clause: If OpenAI achieves AGI, Microsoft's access to OpenAI's technology becomes void. Even more importantly, OpenAI's board gets to decide when AGI has been achieved.
I think I’ll just ignore anything Sam Altman or OpenAI says with regard to AGI.
I was at an event and met a real crypto zealot. It’s been a long time since I met a crypto zealot. While I always am very tempted to engage strongly with zealots I’ve learned to spot one and limit my engagement somewhat. This person was saying that crypto will change everything, including democracy itself ( he might be sort of right about that in a mechanistic sense). Another guest at the same event reminded me that I had brutally skewered his Web3 business 2 years ago at another event. That does sound like me. I was right in this case but that might just be lucky. In any case, I’m convinced that zealots are always wrong, including this otherwise very kind and engaging guy. It was clear that he was a true zealot because he rejected any debate about specific use cases, only talking about hyperbolic future stuff. Paraphrasing: “You’re just trying to find use cases in today’s world, but when the world is multi-planetary and countries and the monetary system have been abolished then crypto will solve every problem.” Uhm… Okayyyyyy.
The most worrying thing he said though, was: “Crypto is a system where the smartest people in the world are able to steal all the money from dumb people. Once all the money has gone to those that are the best at stealing it by scamming people, those smart people will solve all problems and make the world a better a place.” This is such a crazy and obviously stupid idea that I told him to his face that I thought it was insane. I’m still not sure if he was really not joking, but he insisted he wasn’t. Mr Beast for president I guess…
From Benedict Evan’s newsletter: The UK mobile operator O2 is testing an AI-powered tar-pit for phone scammers - a virtual granny who answers the phone and wastes their time. LINK
Coding with AI
When people like Jensen Huang say natural language will be the new coding language, I think they mean something like this:
Prompt:
In this component, we have made sure to automatically scroll down the text area to keep up with the stream. However, I would like to be able to override that by scrolling up. I am picturing something like:
current behavior, unless scroll up, then stop current behavior.
I still don't know a lot of React and JS, but I know the logic that I'm after. Claude wrote me this code:
const handleScroll = useCallback(() => {
if (!chatContainerRef.current) return;
const { scrollTop, scrollHeight, clientHeight } = chatContainerRef.current;
const isScrolledToBottom = Math.abs(scrollHeight - clientHeight - scrollTop) < 10;
setAutoScroll(isScrolledToBottom);
}, []);Funnily enough it’s actually better than what I had in mind. What I asked was that the auto-scrolling just stops whenever you ‘break’ it. But in this implementation, you can ‘reconnect’ with autoscrolling by scrolling down again to within 10 pixels from the bottom.
Original thoughts: Do you think you are hard-working? How would you know?
We don’t really quantify work output in the short-term in white collar jobs. There are large elements of our culture that glorify being busy and getting a lot done. Marissa Mayer infamously has said that she works 130 hours a week, and that she can do so by being strategic about when she sleeps, showers, and uses the bathroom. She was widely and I think justifiably ridiculed for it. It also does not seem to have helped her very much. After not being able to turn Yahoo around (admittedly nobody might have been able to do that), she seems to struggle very much with launching an AI startup. For the record, I simply just do not believe that she worked 130 hours per week in the first place. That’s almost 17 hours per day, 7 days per week.
Even during my 3 month internship in Mergers & Acquisitions at Deutsche Bank I only clocked more than 100 hours in one or two weeks. It was a very fun experience1, but not one I would voluntarily submit myself to for a long period of time. But most importantly, those were not even terribly productive weeks, on two levels:
Firstly, for most of the general public’s office hours (during the day) I didn’t have that much to do. My bosses were in meetings, and I was awaiting instructions. My work only really started at 7pm or so.
Secondly, much of the work was pointless. One night, the stock market had slightly gone up, and I was tasked with updating the valuation multiples in a super long presentation, so they would show a marginally higher valuation the next morning. Another time, a 2 hour analysis told us that the cost of capital (WACC) for a company was 8% (the WACC is usually ~8% for big, going concern companies2). An all-nighter deep dive calculation told us the WACC was 8.17% for that company at that time.
When I’m working, I always feel horribly unproductive when I have a lot of meetings, and always have to remind myself that meetings are also work. Lots of people probably feel the opposite way, like Marissa Mayer, preferring to pack their days with meetings to feel busy. I find myself worrying very often that my team mates are doing things in an ineffective way, or that they’re working on the wrong things. Or what if they are really not working at all? How would you know?3
Bringing that back to myself, how am I doing on this myself compared to others? Am I more or less productive than other people? Do put more, or less effort in my work than them? How productive is Elon Musk, really? I’m sure nominally he almost constantly works, but he sure tweets an awful lot too. Is he doom scrolling X.com most of the day? A big source of unhappiness and anxiety in my life has been the feeling that I’m not working hard enough. Even in my founder days at Ox Street I would settle in at around 50 hours per week working in meetings or behind my computer, and maybe 10 more from my phone. A far cry from the 100 hour weeks that startup founders are ‘supposed to’ pull. But on the other hand, my company made steady progress and exited (might have been just lucky of course). But I think also looking further back, there’s a pattern for me:
I always feel like I am a little lazy, putting in less effort and time than the ‘hardest working’ people around me.
I always feel like my output is at least as high as theirs and I often wonder how people can put so much time in with such low output.
So, bottom-line, my personal believe about myself is that I’m not the hardest worker, but I am a very effective one. The right kind of lazy, as Bill Gates has put it:
“I choose a lazy person to do a hard job. Because a lazy person will find an easy way to do it.”
―Bill Gates
But, I can’t know this. How about you? I would be very curious to hear from others how you look at yourself on these dimensions. Do you put in a lot of hours? Are you usually effective during those hours?
There were many very funny coping mechanisms within the team’s culture: When someone would leave the office earlier than you, at say 23:30 you would say something like: “Bye, enjoy your evening!” The person leaving would reply something like: “Thanks! Don’t stay too late.” One guy would always start playing The Good Life at some point during the night.
Just like a private equity deal will generate 2.5x MM in 5 years, the valuation is 10x Free Cash Flow (if company doesn’t grow too fast). It’s really remarkable how often the world is described quite well by very rough rules of thumb.
I’ve settled on a very clear expectation for my employees, that I also try to hold myself to when I’m getting paid as an employee: For the contracted amount of time (i.e. 40 hours per week), we shall put in as much effort as we can to make as much progress as possible toward our objectives.