MoreRSS

site iconA learning a dayModify

A learning a day, since May 12 2008, by Rohan.
Please copy the RSS to your reader, or quickly subscribe to:

Inoreader Feedly Follow Feedbin Local Reader

Rss preview of Blog of A learning a day

Charts – the Nat Bullard edition

2025-07-05 19:25:00

Analyst Nat Bullard shared an insightful presentation on the state of decarbonization. I picked out a few charts that stuck with me.

That chart makes clear the biggest challenge that we face as humans – a rapidly warming climate.

Fossil fuel emissions were unfortunately at a record high.

Wind and solar are growing faster than any generation source.

Their costs are still declining.

China accounts for 4 in 10 cars sold. And plug in hybrids and battery powered EVs have crossed internal combustion engines.

Plastics are a massive source of emissions.

Energy is actually a tiny part of AI cost.

And a fascinating look at history.

The fear of striking out

2025-07-04 19:46:00

“Don’t let the fear of striking out stop you from playing the game.” | Babe Ruth

Deciding to play is deciding to deal with all the fear, pain, and embarrassment that comes with it.

AI news, narratives, and uneven distribution

2025-07-03 19:12:00

The only thing faster than the pace of AI is the pace of AI news. There’s something in it for everybody.

If you’re worried about existential AI risk, look no further than Anthropic’s Claude Opus resorting to blackmail when engineers threatened to shut it down.

On the flip side, if you want to comfort yourself that this is all hype, you can. Early research in Denmark doesn’t yet show changes in productivity or pay.

The opposite side of this story is about how 25% of the most recent Y Combinator start-up batch has 95% of its code written by LLMs.

And if you’re thinking about how robots might take over human jobs, look no further than the dairy robots who are transforming bovian care or Waymo’s better-than-human safety record.

There’s some data to advance whatever narrative you’d like to believe should be advanced. We aren’t yet at the stage where we’re seeing meaningful consensus yet. It is early.

In this time, what’s most important is to treat these datapoints as soft inputs and instead to make sure we’re doing our own due diligence.

If there’s any part of your job that you might do differently with the help of AI, it is important to try doing so… immediately. If you work in technology and there is a system that is worth replacing with a large language model, it is worth doing so immediately as well.

My read is that the idea that “the future is here, it is just not evenly distributed” is truer than ever. And the more we experience this future, the more likely we’ll be able to shape it for ourselves and those around us.

Soccer choices

2025-07-02 19:53:00

I play a game of soccer every weekend. It’s a fun game that’s just the right amount of competitive – enough that we care about scoring/winning and not so much that it ruins anyone’s mood or threatens a dangerous tackle.

One of the things I’ve observed is that my results are correlated with what I optimize for.

If I optimize for just having fun – defined by running as much as I can and spending more time up front – there’s a higher chance we’ll let easy goals through.

If I optimize for winning, I’ll be doing less of both running or spending time up front, but there’s a higher chance our defense will be better. Being a team player comes with trade-offs.

It’s just a game of weekend soccer. Even so, there’s a powerful lesson to be learnt in strategy and trade-offs.

There are learning opportunities everywhere we look.

Geeking out on data on flights

2025-07-01 19:45:00

Writing here is both a prompt for reflection as it is a prompt to dig deeper into studies and analysis about topics I believe are important.

My process for learning more about these topics – especially around the impact of exponential technologies like renewable energy and artificial intelligence – is to compile links to interesting presentations and articles over time with the goal of going through them over the weekend.

But those weekends show up rarely – if ever.

I tried something different recently – attempting to catch up on all this reading during a couple of the long flights I take during the year. Thanks to inflight wifi becoming more commonplace and reliable, this is possible.

On one such long flight couple weeks ago, I ended up catching up on three months worth of reading. So, if you suddenly see a series of summaries of charts and analysis during these weeks, you know why. :-)

I still love catching the occasional movie. But I appreciate being able to use flight time to geek out on fascinating data.

Service costs aren’t exploding anymore

2025-06-30 19:34:00

Noah Smith shared a thought-provoking post on service costs in the US.

It started with a famous chart that many of us have likely seen about how service costs seemed to be exploding even while the cost of goods was coming down.

Interestingly, these service costs seem to be leveling off. This is spending on healthcare.

Also hospital services.

Medical costs have slowed down too.

College tuition has also leveled off.

While it is hard to pinpoint why with trends like this, it is likely that markets have reached some form of equilibrium. There’s only so much a student is willing to pay for college after a point.

It also helps that labor has also been more productive over time.

This doesn’t mean all services are in a good place – childcare has gotten expensive for example.

But these charts point to a good outcome for the quality of life for normal people.

And that’s a good thing.