danpalmer

24,169 karmajoined 14 tahun yang lalu

Software Engineer at Google – Android SRE.

Formerly Google Play App bundling and delivery Formerly at Thread (thread.com).

Opinions here do not reflect the opinion of my employer.

https://danpalmer.me/

comments

kemarin duludiscuss

Yeah this was a generalisation across the US/UK/AU. I come from the UK but now live in AU, and know the US a bit.

I think the looking down on trades is not quite as simple as I summarised it as, and I think AU does a lot better at all of the above than the US/UK, but I still see aspects of it. Tradies make good money in all 3 countries due to lack of supply, and yet there are still stereotypes of jobs like lawyers, doctors, (software engineers?) being better in some way.

It's a nuanced problem, but I don't get the impression that trades here are culturally valued higher than a lot of "white collar" work. Compared to ambiguous "desk jockeys" yes, but that's due to negative stereotypes about bullshit jobs, if you actually named a specific job I think you'd find different attitudes. Lawyers, accountants, sales/marketing, various engineering disciplines, IT, I think these are widely considered "better" jobs than trades, even though in most ways that's far from true.

kemarin duludiscuss

College is a box to check off because we have lost the ability to support a workforce without college degrees (at least in the US, UK, AU, where I have some experience).

Trades are critical but looked down on. Manufacturing is gone (which isn't in itself a bad thing). The service industry doesn't pay a living wage (thankfully it's reasonable here in AU). Apprenticeships don't pay enough. And pretty much all knowledge work jobs expect a degree from the beginning at a junior level.

We should be planning for a system where <20% of people go to university, instead of expecting >60% to go. Robust minimum wages, good trade schools, apprenticeships that pay enough for a wide range of roles, and changing the culture to not look down on folks that take these paths.

3 hari yang laludiscuss

Coding yes, copywriting, design, identity, no. Using AI doesn't mean giving up on quality, unless you don't care about quality. Most of these issues come from folks who don't really care about quality and ship the first slop that comes out.

4 hari yang laludiscuss

> It's more art than engineering

I'd say games can be either art or hype. Call of Duty is not art, really, it's hype. In the same way that no one thinks Marvel films are moving film forward, but they are hugely popular. GTA is somewhere in the middle, being mostly hype driven, but based on solid "art" in good gameplay. Indie games tend to be art over hype.

Microsoft can't do the art because it's too big, too safe, and it can't do the hype because it's not cool.

5 hari yang laludiscuss

I've not seen anyone yet implement a true cost to productivity assessment or guardrails for AI usage yet. Sure this is hard to do with people, but performance management is a well understood field with a hundred years of practice for knowledge workers.

We don't get unlimited hiring budget, so we also won't get unlimited token budgets, and we as the operators will be responsible for the productivity of our agents.

What does performance management for engineers look like when dollar token cost is included in reviews? I think it's going to change a lot of assumptions and a lot of strategy around AI use.

5 hari yang laludiscuss

It's also about competition.

Consoles are a significant investment, to only be able to buy content from one source creates monopolies. An equivalent would be if TVs were sold by Netflix or Disney, and only played content purchased from those entities. The second hand game disc market prevented those monopolies from taking prices to stratospheric highs.

I believe this is why digital distribution has seen little push-back on PC, because we have the convenience without the monopolies. Steam is huge, but Epic, Ubisoft, EA, gog.com, and others provide plenty of competition.

If consoles go digital only, from a single source, I think we'll see >$100 games quickly, we'll see subscriptions required to play be much more prevalent, and we'll see the death of consoles within a generation.

8 hari yang laludiscuss

I think there's a measure of success in thought leadership and/or product.

Apple is wildly successful at both, arguably more so the leadership than the actual product. Amazon, despite its faults, has a ton of businesses many of which do well, and it continues to innovate. I'm biased but I think Google is also in that category, with many new products that are widely well regarded (yes some were acquisitions, but typically smaller ones).

Meta on the other hand... Facebook was huge, no doubt. Instagram too, but that was already semi locked in on acquisition, they already had product/market fit at least. WhatsApp has languished under Zuckerberg, having had their explosive growth independently.

Oculus? Nope. Metaverse? Nope. Crypto? Nope. AI? Nope, or at least not yet.

By business metrics, very successful. By innovation in ads, very successful. But building new consumer businesses? Not really.

8 hari yang laludiscuss

Rails takes this approach[1] to some extent, but it's still possible to plug in different things, it's still all flexible enough to build any type of web app pretty much, and it still requires a ton of developer (or LLM) involvement.

I'm thinking of going far further, to the point that perhaps we should use a new language designed only for web-app development. I'm thinking about removing almost all options so that the LLM only gets to write custom business logic and data modelling and doesn't need to do much else. Again this is all at the cost of being more generally applicable, but I see a lot of software that is fundamentally CRUDL and it's still hard to build well, and I also see a lot of LLMs reinventing the wheel but implementing too many sides on that wheel. They need guardrails.

[1]: https://dhh.dk/2012/rails-is-omakase.html

8 hari yang laludiscuss

Something I'm thinking about and doing a bit of experimentation with is using LLMs to write specialist higher level code.

Rather than ask them to write web-apps in webby languages with open source frameworks etc, providing a very fixed, on-rails development process where everything is abstracted away. Accept that it'll be less powerful, but take the trade-off that it'll hopefully be faster and produce much more controllable software.

Concrete example, why do we let the LLM choose a database, schema, migration procedure, library, etc. We could decide to only support one database, enforce schema design (such as every table containing access control), enforce a migration process, enforce a library, even do schema design in a fixed config file rather than arbitrary DDL. Same for auth, deployments, even UI.

8 hari yang laludiscuss

At some point these things become cheap enough that you might as well. The price difference between 500MB and 1GB for me is very little, and so for the peak usage time improvements and few times a month that I download a steam game or movie, it's worth it. I pay significantly less than 70 USD a month for my 1GB connection.

8 hari yang laludiscuss

The Australian system is much better than the US system, even with 1/10th the country-level density. The internet here is generally faster and cheaper than the suburban US, with a similar system to Switzerland.

8 hari yang laludiscuss

But the article does point out that 1G plans are very widely available at low prices. US consumers would love that but it's not available anywhere near as widely as in Switzerland.

8 hari yang laludiscuss

I can also get 400Gbit in my office, that doesn't mean it's a useful benchmark for the country. The article seems to represent the state of the 3 countries compared pretty well.

Where exactly is Ziply available? Their website is vague, but it seems to be at most a small corner of the North West, and it seems like their 50G plan is not as widely available as their 2G plan.

8 hari yang laludiscuss

I ran speed tests a lot more when I had internet that varied from 5-20MB depending on the day/weather/etc. Now I'm on >1GB it's so rarely a concern that I don't bother. I suspect this skews the data significantly.

8 hari yang laludiscuss

It's fair to critique this article not covering this, but I also think this is largely a red herring. The vast majority of the issue in the US is suburban, where density isn't really a problem. The US has a lot of rural areas, but they represent a tiny fraction of the population.

As a comparison, Australia has roughly the same land-mass as the contiguous states, but with less than a tenth of the population. It has its fair share of ISP and telecoms issues, but not as the US for the most part. Most people live in cities with good internet infra, most of the rest live in towns with at least some choice. Not perfect, a long way to go, but better than the collection of monopolies the US has.

8 hari yang laludiscuss

I like the idea at a social level/product level, but at this point I don't think I'd use any social network that wasn't open at a protocol level in some way. ATProto/etc would be a great tech backbone for this, and would then let users self host their content, it would let orgs run their own views, it would let archivers archive content for posterity, it would open the door to possible federation in ways that suited the network in the future. Re-inventing all of that from scratch would be a terrible idea, and would lose focus.

8 hari yang laludiscuss

The competition here is hiring a cleaner. Cleaners are insured. And it's even more than that, it only does a very few jobs, so we're talking a total of <5m of human time per day that it currently automates.

8 hari yang laludiscuss

But it can't lean and balance. Making a bed requires reaching potentially 1m across a surface that might also be 0.5m high. That requires balance, humans can bend and balance like that very naturally, but that's hard for a robot, and this doesn't look like it has the manoeuvrability to achieve it.

9 hari yang laludiscuss

> Handles loaded hampers

What does it mean "handles"? It doesn't say it puts on a wash, but that's what I'd want. I can only assume they're vague because it doesn't do anything useful.

> Makes beds

A robot the height of a child makes a double bed with sheet, duvet, and pillows? I highly doubt it could reach.

> Isaac 1 is autonomous for Laundry Flow and Daily Reset by default, with teleoperation assistance when needed to guarantee we complete tasks.

That's a lot of words to say "a person will drive it around your home". What sort of insurance do they have for that person breaking something in your home? What audit trail do they have for the operators?

9 hari yang laludiscuss

Why didn't they just make it "Staff SWE-Bench", would be much better smh. /s

But seriously, as an industry we're terrible at assessing engineering levels, I've worked with "senior engineers" who can't code and I've worked with "junior engineers" who could run rings around them.

Benchmarks like this should be much more precise about what they're actually testing, and what axes they're hard on. We also need to rise above prompts like "you are a senior engineer", it's woo, and it's far better to ask for precise outcomes.