diatone

Agree, there are pointer-centric patterns of computer use. Usually visual things: art, design, some CAD, marketing, and of course the GUI patterns of personal computer use that most people have been taught

Oh and on one more thing I hadn’t addressed. I’ve seen teams try to “swarm” on a problem, with varying levels of success. It’s easy to be cynical and draw allegories to storming a trench. If we just all run together, maybe one of us will make it to the other side…

Sometimes this comes from a place of wanting to improve the bus factor, or minimise siloing.

The truth is that it’s hard to bake in this level of redundancy, and still be effective, without rigorous and brittle process, or without genuinely motivated team members who trust and appreciate each other, who are invested in each other’s success. Either way it takes a lot of time, and a lot of training and personal growth for the team members involved.

It really depends on the person. And people change too!

I’ve seen scenarios where someone will refuse to collaborate tightly (eg: pair programming), but also be surprisingly slow at getting around the tools and getting code out the door. Or they rabbit hole on something that ends up being a waste of time.

I’ve also seen scenarios where someone who is highly proficient just needs context to do a fantastic job - for example, a designer pairing with a front end engineer to iterate on some UI changes in an hour.

I’ve also seen scenarios where people expect a huge amount of direction and task definition upfront, and refuse to work in ambiguous circumstances. And the opposite.

And finally I’ve seen scenarios where you think you’re about to sit down with someone to work together on something, and you realise your time is better spent understanding the issues they’re facing in their life, and their concerns, and dignifying them before considering the task at hand.

If you suspect that siloing is preventing you or your team from accomplishing more, faster, more sustainably, make a case for it. But don’t cargo cult pair programming or backlog grooming or any other process for the sake of it, then you’ll just look silly. Good luck!

It’s not enough for the method to be possible, it has to be economical. If it’s not prevalent yet it means one of two things:

(1) cheaper methods work too well for now to even bother with sophisticated approaches

(2) it’s too expensive to be worth the effort

It would be interesting to plot out the cost of an integrated AI stack for this over time.

People do it because they like to, for reasons entirely up to them. Maybe they find it interesting. Maybe it’s therapeutic. Maybe it gives them a social opportunity. Maybe it’s fun to push your own limits, for its own sake. None of that is stupid or pointless.

If someone started looking at me do my hobby with my friends and decide I was a failure, that’s their perspective. But I think the response the kids give to that these days is “touch grass.”

A high rate of customer acquisition isn’t the same thing as a network effect.

OpenAI has some network effects baked into its product suite, but it’s not even in the same league as Microsoft’s bundling strategy (setting aside Google for a moment). Microsoft’s history is littered with competitors who had a better product, a faster adopted product, who were snuffed because of Microsoft’s superior distribution capabilities.

I’m on the train at the moment, but the big exception is the mobile market, and in hindsight that makes perfect sense - most of the phone market consumer grade, where bundling productivity software isn’t much of a value proposition.

It’s far more likely that OpenAI remain where they are on the value chain, because it’s easier to capture and integrate AI startups into their products. If they were to compete on productivity software, they need to differentiate with entirely new modalities of AI-informed user interfaces, or they will be yet another slightly better program that got blown out of the water by MSFT enterprise distribution

edit: to be clear my point is, why would you become a bit player in an established market when you can dominate a new market you created yourself?

This is a great insight, the idea that management has to look at leading indicators for the org itself. Makes me think that beyond line management, a manger’s job is to behave as an investor and steward of the org, instead of getting involved in the details.

Not a management book per se, but Working Backwards popularised an Amazon approach of thinking about the organisation in terms of mechanisms. You might find that useful.

I’ve always wanted to dig into Scarlet Ink, Dave Anderson’s blog, but never have, and I suspect it’ll be quite useful for you too.

I’m not sure you can characterise all software with such a sweeping generalisation. There are plenty of use cases outside commerce - outside optimising distribution and trade of goods and services - which software contributes to immensely.

Also, maybe there’s an ironic counterfactual in payment and financial reporting networks, like Visa and Peppol. Peppol is not for profit, and Visa infamously facilitates five-way transactions with a huge degree of backwards compatibility.

As an aside, I wonder what code quality means to the author of TFA, and what craftsmanship looks like. Would be fantastic to get clear examples of this from the history of the industry.

If you haven’t already, and I know this doesn’t hold up for GUI emacs or vim, but consider running them through https://mosh.org/

Check out TigerBeetle!

It’s intentional, to distinguish between known unknowns and unknown unknowns.

I'm not sure it's fair to say that the industry has added overhead versus the 90s. Unless you're talking about gamedev, which I can't comment about. But it's hard to say that tech in general is slower moving as an industry than thirty years ago; I think there's more nuance to it than that. What exactly that nuance is I'm not completely sure, but here are some counterexamples:

Intel shifted the entire direction of its company away from memory and into microprocessors as early as the 1980s, in response to smaller, allegedly faster moving Japanese competitors.

IBM has been a slow-moving behemoth for decades; the story of how DOS began is an idiomatic David vs Goliath tale of the industry.

Apple debatably rose to prominence on the back of tech lifted from Xerox, which for one reason or another (maybe they were too confident in their fax machines) didn't ship their own tech.

Google was at one point a fast-moving startup, up against slow-moving competitors like AltaVista, arguably Yahoo, MSN.

Amazon's original web stack was written in C, iirc. I personally would have a hard time arguing that things done today in web development could be done faster in C.

Yeah it’s bizarre; if someone’s going to get an LLM to ghostwrite their reply (not saying this is the case here but my internal ChatGPT alarm also went off reading this), will they use it for replying in the thread? Will they fact check their work?

Frequency distribution I think - normal, not uniform.

Worth pointing out, we live better lives not just from the technological breakthroughs, but also from the social movements that have brought us things like the 40-hour work week and keeping children from labour for a reasonable time. Which were direct responses to the Industrial Revolution.

This is an interesting position to take. I’ve been back and forth with my N+1 about the utility of sync vs async comms to lead, and I can never nail down a result I’m happy with. Like, sync comms is effective for ladder climbing as you say. But it has failure modes in remote-first teams of ICs across time zones that async comms don’t have. Async comms inherently scale further, too. Maybe their reach is is greater but the impact is lower? I wonder how far recorded video gets. Then you have mentoring juniors - much easier to do synchronously. What’s the appropriate mix?

> Pictures are worth a thousand words but if you don't know which words you want to communicate your picture is worthless. Something many artists eventually learn, what is it you actually want to say?

If I had a dollar for every time someone drew a diagram that expressed their thought, but didn’t explain all the hidden assumptions and omissions that clear writing would flush out, I would be a poor man since they wouldn’t need me to review their work.

As someone who grew up in the area (mid 20s atm), there is a well known phrase amongst locals (at least my mates), the "Brighton Bubble." Also yes you're spot on about keeping ties through all levels of education, at least in my case. It was funny living in Sydney for a year, and returning to Melbourne, and when dating one of the first questions you'd be asked is what school you went to. At least, that's my experience, with a couple of people.

But it's not all gloom and doom, the closer you get to the CBD the more opportunities for connection open up. More activities to do in general, more people who are new to the city and are less likely to have ossified social circles, more public funding for that kind of thing. Melbourne has a rich variety of cultures and perspectives to immerse yourself in if you can align your life along the same axes that nurture those cultures and perspectives. Out in the burbs, particularly in the Bubble, less so. Doesn't help that Brighton and its surrounds are skewed to preserve the lifestyles of the people who grew up in the area (this is the politest possible way I can describe it). Speaking from experience there are plenty of people who left the insular communities of their youth around here, because they never felt like they belonged, and never looked back.

Also worth remembering the lockdowns didn't help. I'm not sure exactly what you're looking for in terms of social connection but my point is, please keep plugging away.

Can anyone weigh in on the state of the JS / TS ecosystem for linear algebra and matrix operations, and how that stacks up against the Python ecosystem?

Is this just a difference of perspective? Seems like GP's point was that software has latent value, and differentiation, and the act of selling is commodifiable.

Your point, correct me if I'm wrong, is that software is more the commodity and sales is differentiable?

I wonder how this stacks up against other software that yielded massive public utility, like UNIX, or the internet.

> Running an LLM query through a GPU is very high latency: it may take, say, 5 seconds, with a throughput of 0.2 queries per second.

Why?

It would be great for someone with experience from the dot com bubble to weigh in here, to describe if similar dynamics occurred back then

Everything in your world. You decide on the system boundaries. If you haven’t already I’d recommend watching the accompanying TED talk

If the business doesn’t have product-market fit, the single most valuable skill you can exercise wrt the company itself is zeroing in on that. Speed, business understanding, and speed again are the top 3 components of that

Whether or not that aligns with your own benefit is a different question entirely. If your comp is set up well and the company is supportive, it should be.

I feel the same way from time to time. But - please take care of yourself, and if you’re not already, please stop to smell the roses when you can

Not a great answer for the GP though.

Last ten years? Maybe something in deep learning.

> One simple reason to post: For every potential competitor reading about your project, there are thousands of potential customers.

If you’re happy with your existing cash flow this point seems less relevant; the better choice would then be to protect your competitive advantage and stay hidden, right?

Re your point on OSS being architected first, then the org shapes itself around it, a decent counterexample is the development of git.

https://www.linux.com/news/10-years-git-interview-git-creato...

Written specifically to match the politics of Linux kernel development, since there wasn’t an open source alternative that could do so already (paraphrasing the interview)

Maybe what you’ll see is a bigger distance between modes of the salary distribution for software engineers.

Top end salaries for language model fluent high performers… and the rest.

I know I’m simplifying and that there are many other factors at play in salaries, but my point is that language models are changing the cost structure of knowledge work. That is also the point of TFA.

My advice: critically assess the cost structure of your job, and which parts neural networks will commodify. Then reposition yourself accordingly.

Here is an Australian site expressing a similar view about the Australian economy. One prominent poster is ex Australian treasury and has been featured on news programs on TV.

https://www.macrobusiness.com.au/