mathattack

Very interesting. I assume that we're at a point where can just take photos of the chips and have AI find the easter eggs, though it does take some of the fun out of it. I think the calling cards are a great signal of pride in work.

Does this lower your likelihood to use them for other things? It's painful to switch.

Or is Dropbox hunkering down? Their market cap is $7.6bln on a roughly flat $2.5bln in revenue (compared to $4.5bln and $1.1 for Box) which suggests they're still relevant. Perhaps they're accepting that growth is gone, and time to focus on the cost side?

My experience is every layer between the communicator and the recipient adds noise. It's like the game of telephone we played as kids. Some PMs are terrible and are 100% noise. Even the best are only 50% signal.

Give an engineer a clear understanding of the end need, and you have a tremendous gain in efficiency. I think there is a 10X benefit here that's similar to the 10X benefit from stronger engineering skills.

You still need PMs, though it's a different job that passing papers back and forth.

Yes - it's an enterprise versus SMB distinction that doesn't require asking questions like "What's your revenue?" in the qualification. Enterprises need it. SMBs don't.

On a recent Mindscape podcast Sean Carroll mentioned that rationalists are rational about everything except accusations that they're not being rational.

This!

In many large companies, there are non-aggression pacts. “I don’t air your dirty laundry so you don’t air mine.”

The best consultants comment on the Emperor’s wardrobe, or lack thereof. And they do it in a way that makes everyone pleased that the logjam is released. And they can only get away with it by being temporary.

And for all the complaining of consultant bill rates, independent consultants have a lot of overhead to cover. (Sales, taxes, insurance, downtime, legal…)

References another behavioral economics article on anchoring. https://therosen.substack.com/p/llms-anchor-too?r=2cmy

It's a problem everywhere in California. In Palo Alto they're getting rid of honors classes in STEM subjects like Biology.

Although this story was debunked, many Universities own Timberland in their portfolios. They’re a good inflation hedge for schools with long time horizon. (Real estate and paper investments were historically very correlated to university costs)

https://blog.realestate.cornell.edu/2018/04/20/harvards-natu...

The last paragraph should temper our enthusiasm. Allowing people to build more when there is a shortage is the best way to have markets solve problems.

—-

“ Meanwhile, the pipeline of new apartment buildings is drying up. The number of properties under construction is down by roughly one-third from the peak in 2023, the report found. That likely means fewer units coming available in the months ahead, potentially giving landlords room to start raising rents again.”

I have no idea if this story is true, but it should be.

Some added commentary on the concentration of programming languages: https://open.substack.com/pub/therosen/p/analyzing-the-stack...

Good question. A couple caveats on the data:

1 - These #s are just domestic.

2 - They don't contain near-CS degrees like math and .electrical engineering (which may not have grown as much)

3 - They leave out bootcamps and similar training programs.

4 - Domestic only.

My (only slightly educated) guess is that many US based CS grads who used to go to internal IT jobs are now at tech companies. CS grads from India and elsewhere now take up most of the internal IT positions due to outsourcing.

I asked Perplexity [0] and they said that from 2000 to today software engineering emnployment went from 680K to 1.53MM. That's closer to 2.25X. Maybe it rises to 3X if you back up to 1995.

[0] - https://www.perplexity.ai/search/how-many-software-engineeri...

It's supply and demand.

It could also be a 4X increase in grads for Computer and Information Sciences degrees since the 90s.

Source: https://nces.ed.gov/ipeds/Search?query=computer%20science&qu...

I've met several of these students. It's like an MBA, but less social networking and more math. (And can be done co-term or in a year)

So does it add some value to someone who is already getting a bachelors in EE, CS or similar? Sure.

Would I put a history major with an MS&E degree in charge of anything significant? Probably not.

I suspect that the admissions rate of 7% is independent of coterms.

I've found LLMs to be great in summarizing non-controversial non-technical bodies of knowledge. For example - the facts in the long swings of regional histories. You have to ask for nuance and countervailing viewpoints, though you'll get them if they're in there.

Good point. I can quickly intuitively tell if the suggestions for my writing is correct. Harder to tell on code.

Perhaps the analogy is better for "Writing code" versus "Writing Test Cases"?

Interesting. From a writing point of view this suggests that it's better to have the LLM "critique my draft" rather than "write the first draft." (Both for writing text and code) Also implies that we want to manually check all of the LLM's suggestions. This makes it sound more like a co-worker (agent) than all-powerful SuperIntelligence. I guess this is a symptom of the hallucinations.

https://open.substack.com/pub/therosen/p/should-llms-write-y...

Recall that this is a StackOverlow survey, and reflects their history and audience. [0] Even still, I'm also surprised.

[0] https://www.joelonsoftware.com/2006/09/01/language-wars/

Could the issue be that they went from technical founder to spreadsheet leadership without having a hyper-competitive visionary?

It’s seems like they had plenty of genius and adults in the room, but no hyper-competitive Musk/Zuck/Ellison/Jobs/Gates.

If you have a PhD in physics, and your option is to deal with this abuse and get Paod X, or join an AI firm and get paid 5X, which would you choose?

I would’ve thought that one was from the Onion.

Thanks for sharing. Folks who combine dual passions are a fountain of creativity.

Some universities with large endowments used to be referred to as hedge funds that happened to have professors. Now they happen to have pro sports teams too.

Perhaps it's getting repurposed?

Think about the demand and supply curves of calculations (or computation). For most of history, they moved in tandem, with supply moving slightly faster, so computers would always do more at slightly lower costs.

Now both curves are speeding up, but demand is moving faster, so the costs of hardware are going up. And when high end servers (with GPUs) are unavailable, people hold onto the older ones longer.