Hacker News Reader

I was thinking about the number of technologies that came out (essentially) fully baked and haven't seemingly improved since.

I had the thought while grumping at my Alexa. I say the same phrase every day, and every day there's a 50% failure rate. I keep looking for phrases that it can understand better than the ones I've cycled through, but they all suck.

But Alexa, as shipped, was pretty great. The fact that it hasn't seemed to improve at all since confounds me. The reports of the high body counts in the division surely attest to the lack of focus or interest in core Alexa skills. All those product and integration priorities were on the wrong product stack layer.

Same with: iPhone typing, Siri, Tesla software (and beyond FSD -- the only thing that seems have happened in the 2 years I've owned my vehicle is the addition of Apple Music), Apple Watch, etc.

It was fully-baked in the sense that it was novel and could recognize fairly basic voice commands including simple questions of fact. Once the novelty wore off, a lot of people concluded that wasn't actually that useful given they had a computer in their pocket anyway.

Voice is still just an implementation detail. I'd argue that we still aren't at the point where general purpose computer assistants are all that useful relative to humans. It's the same 10% (or whatever) problem as self-driving has.

I don't think it's true that the Echo was only a novelty, it was and still is an extremely effective way to do a couple of tasks that people find very useful - mainly playing music of course. The hardware and processing required to recognize its wake word very reliably and arbitrary voice commands somewhat reliably in an open space was an incredible leap forward for consumer products at the time. But it has managed to add essentially zero value to the initial offering in the decade since then.

YMMV of course. I maybe use it to play music in my bedroom a couple times a month. Usually I play it from either my Apple TV or my computer.

It's useful for a few other things. A weather forecast if I haven't gotten out of bed yet, setting an alarm, turning my bedroom lights on. The occasional quantity conversion in the kitchen.

But all those add up to something closer to "novelty" than whatever investment Amazon put into the technology.

> Tesla software (and beyond FSD -- the only thing that seems have happened in the 2 years I've owned my vehicle is the addition of Apple Music)

If you want changelogs you can look at this fan-run site. A lot of small iteration and a few big splashes (depending on if you like the feature or not)

https://www.notateslaapp.com/software-updates/

Fwiw, I've had good luck creating routines to cover wordings that should work but don't. One of the available actions is a (single) custom command, so you can use a one-action routine as a map from this wording to that wording.

So, for example, on my system “start my car” actually maps to “tell blue link to start my Ioniq and set the temperature to 72” because the skill is godawful picky about getting all of that.

I'm convinced that the feedback I send using the Alexa app goes nowhere. It definitely isn't used to improve Alexa's ability to understand the simple phrase: "play NPR".

Edit: I almost forgot. There was also the time it stopped understanding "stop". It properly parsed the word but didn't know what to do with it. IIRC I had to use the mobile app to stop playback that day.

This is true, now. It is the case that Video, Music, and SmartHome teams took over much of Alexa support in middle of 2022/2023 (many other teams were fired), and the result is basic functionality such as "stop" doesn't work (music, timers, alarms), "play" goes to some random incorrect cache of Music or Video, not News. What happened to the data (including your feedback)? It disappears, of course. Nothing but SMILES (a program designed to only report positive trends and feedback) for Amazon leadership.

Alexa devices are spyware; they listen in on you, and if your device has a camera the camera can be accessed without your knowledge too.

Unplug this stuff already...

> I was thinking about the number of technologies that came out (essentially) fully baked and haven't seemingly improved since.

Google Maps (mobile). Horrible UX in so many ways.

I think it's hard for a company to sponsor a product like Alexa if the monetization is near impossible. It's basically a vanity, money draining product.

I deeply resent it when Alexa tries to upsell me something and I certainly would if it fed me an ad every time I asked it to do something.

And what's the alternative? A subscription?

Alexa certainly isn't useful enough to me to pay anything meaningful.

Agreed. I get angry. How dare you try and sell me garbage when I ask you about the weather?

Make good speakers, improve language detection, improve music playback, simplify that stupid-complicated-always-rearranging Alexa app, decrease my curse-at-Alexa rate, make Alexa part of Amazon Prime or charge me an additional $20-30/year for simple audio processing services.

This might not be popular here, but in my opinion, the H-1B visa program has been a disaster for tech companies. It oversaturates the tech labor market with career-seeking individuals, drowning out the hackers who immerse themselves in technology (and may be on paper "less qualified"), and recruiting/HR teams have virtually no idea how distinguish between these types of candidates.

This is a very touchy topic because it's very hard to talk about honestly without accusations of race / xenophobia / prejudice / envy / whatever. I'll say: I've worked with tons of H-1Bs who were "immersed" in tech. I've worked with tons that weren't. Individuals are individuals.

The problem is that collections of people behave wildly different from individuals. The culture of an org that's mostly H-1Bs is wildly and starkly different than ones with different ratios. This problem, at least when I initially joined Amazon many, many years ago, was actively warned against. The onboarding had a little speech about cultural differences and how it's OK to push back here. We aren't here to be task takers.

Fast forward to today and the culture is wildly different. Many of the practices of offshore teams have leaked into local ones, because the same culture that bred them over there are now in leadership positions here. There's no shortage of teams operating like unquestioning ticket jockeys these days. Your job isn't to produce value, it's to close tickets as fast as possible. Do what management says. Don't ask questions.

This comment is not reflective of ground reality. Most of current workers (90% easily) at big tech, both US Citizens and Visa holders are far too removed from the hacker ethic and can't do more than using the company RPC framework to write boilerplate stuff. People act shocked when told about basic insights of computer science. Mind you I am talking about both low and high level engineers and management.

Lol is this the "best of the best" that FAANG supposedly hires? Here on the east coast, everyone is expected to have a sound CS foundation or equivalent skills and experience, our second tier state college has a fantastic CS degree, and the half of my team that is in India has no problems with everyday usage of CS concepts and the worst I can say about them is a lack of initiative or creativity but like, same.

Is FAANG just sniffing it's own farts about their "ability"?

When you are hiring at those high rates quality goes out of the window. The companies were so big and growing fast that all external leaders would hire their loyalists. The hiring quality was dropping steadily and then dropped sharply. There were some well intentioned special hiring initiatives which were in theory supposed to preserve the same quality of hires, but in private the HR leaders responsible for the initiative would send everyone a private message saying "Wink wink, this hiring is for special initiative, please please keep that in mind". The thing is if you don't have immediate customer feedback and the company is not in survival mode, empire building just takes over. I have worked at multiple of big tech comapnies and all just daily remind of "Yes Minister"

> drowning out the hackers who immerse themselves in technology (and may be on paper "less qualified"), and recruiting/HR teams have virtually no idea how distinguish between these types of candidates

Interesting thought ... could be a kernel of truth there.

I believe this is true. But FAANGs are really software factories, not places for innovation. The product requirements rain down from high up above, and they expect a shoddy implementation as fast as possible. The model largely works

Why do you think people from other countries aren't "hackers"?

The people from other countries who are genuine hackers aren't qualified enough to get approved for H-1B visas.

Immigrants are now responsible for the lost hacker spirit? Seriously?

I think you might be confusing correlation with causation. Software engineering becoming a well paid job for the masses is the reason for both decline of exclusive hacker culture and amount of career-seekers (both domestic and international) willing to join the industry.

"Many of the people running Alexa had been there since 2012. Its was tyranny by those who started in that organization first. Backstabbing, politics, bad engineering, nepotistic promotions. The scientists were so far behind, almost all models were off the shelf implementations pulled from Github. Huge amounts of capital flushed down the drain."

Wholeheartedly agree. Any time a team or individual had a good implementation that deviated from that core Alexa group, they stole it and fired the people who designed and implemented it. Nevermind the amount of personal customer data they leaked in the process. Embarrassing.

People who left Alexa to go to places like Meta actually showed return on investment. The people who ran them off? They're still at Alexa, they probably aren't the ones being fired, and you can easily recognize their work in projects like Amazon Music and Prime Video. Some of them even have the nerve to boast working on "AGI," as if the echo of their Titan flop was not loud enough.

This is extremely correct and captures all the problems quite well. I will add that the Amazon 2 pizza team model is inherently unsuitable to building a whole class of products. I remember all kinds of small 2 pizza teams having multiple "scientists", even if they did nothing other than running models from Github.

Almost all big tech companies face the same problem in slightly different ways, and further increased due to the sloppy hiring in recent years. In Google they hired tons of extremely incompetent folks, and now those folks band up together and try to closely guard information, as that is the only way they can ensure their performance is not very bad compared to others. People who are both technically and politically incompetent get pushed out, while the politically savvy and technically incompetent ones manage to create moats for their survival.

Maybe all the back-stabbing and nepotism was data-driven[1]?

[1] https://aws.amazon.com/executive-insights/content/how-do-you...

Ha! I was there for 6 years. The "data" at Amazon is really just "whatever the management chain wants" for their empire building purposes. If you help them in this goal, you go up, if you get in their way, you go out.

Exactly. The data is "cooked." We are literally told to make the numbers work, or else.

> The whole company is a meat grinder, poor technical implementations with an army of SDEs keeping it running.

> They need to flush out the legacy employees and make room for hungry young workers.

I think there might be some cognitive dissonance at play, here

Yeah the hungry young workers are the meat in the grinder! Do we think they don’t notice that role?

It sounds to me like they need to flush out incompetent management.

Arguably voice assistants are still ahead of their time. My Alexa is handy for a few things like setting an alarm--though I still want to set a backup for something like an early morning airport pickup.

But for an interaction in the car to see some available podcasts and play an episode? It's just not "smart" enough. And you go through a few rounds of frustration trying to get an assistant to do something and you just stop trying.

There are specific use cases. E.g. my dentist uses one as a timer especially when he's gloved up. But generally they're pretty optional.

Yes, its possible I am being too harsh, and the product was simply impossible for the capabilities available.

> They need to flush out the legacy employees and make room for hungry young workers.

Ageism demanded by the masses. Tech workers are a funny bunch in how they love digging their own graves.

> They need to flush out the legacy employees and make room for hungry young workers.

Good luck. Hungry young workers will leave and go elsewhere once they see how things are run. Speaking as such a worker who's now dealing with incompetent middle management, Amazon is where dreamers and hungry engineers become cynical and disenfranchised.

> trying to solve innovative business problems with incremental improvement.

This describes pretty much every AI startup I know, both good and bad.

I suspect the places where AI will really succeed will be places that entirely change existing workflows and user experiences. This requires a fair bit of research and experimentation, but most startups don't have the appetite for that, and VCs are putting on the pressure to do AI stuff fast.

A lot of AI potential is being squandered now due to a simple lack of patience and planning in the rush to market.

Amazon couldn't care less about research imo. They know they'll be able to quickly get into next bandwagon with scrappy engineering even if they aren't the ones inventing it

> flush out legacy employees and make room for hungry young workers

Super ageist attitude, my observation is many experienced engineers are the ones who would like to improve things. The "hungry" people you refer to are just as likely to be ladder climbing, job hopping, cutting corners for quick wins, and playing politics.

Does that happen also within the AI groups? For example, I've tried CodeWhisperer and it's hot garbage. But the same prompts I give it work reasonably well in Bedrock (though not as well as ChatGPT/CoPilot/etc).

Bedrock is an AWS product, and is relatively simple. Its AWS's strength, infrastructure provisioning and compute operations. I'm sure it works very well. I'm not saying all of Amazon is bad, the scale and reliability of AWS is an engineering marvel

Not to mention AWS products get strong customer feedback all the time and it is treated vitally important. If you are maintaining an internal product in Retail you will just treat your customers like captives.

I only need Alexa for like four things and it excels pretty well at that:

* What's the weather?

* What time is it?

* Set a timer for X minutes

* Where is my stuff?

Edit: Also, measurement conversions in the kitchen are handy.

> Where is my stuff?

It doesn't do this very well for me. It only knows things like "a package arrived", or "a package is coming tomorrow". It doesn't say what the package is, etc.

Even for measurements, it'll sometimes say a cup is 236ml and sometimes 250ml, because "cup" is ambiguous.

That's about all I need as well. Then they take those basic functions which they are very good at and started trying to push other usages. "By the way, did you know dogs are blah blah blah."

Fortunately, they seemed to have backed off on that a bit. It was a good reminder for me though - they control all the dials and can take something I bought and enjoyed using and enshittify it at the drop of a hat.

You can say "stop by the way" and it'll stop for some inderterminent amount of time then come back. I've heard that some people set up a Routine at 3am every day to turn the volume to zero, say "Alexa stop by the way", then turn the volume back up. I haven't tried it but I heard it works.

I yell at Alexa when this happens: "NO ALEXA STOP! ALEXA STOP! BAD ALEXA!" and etc.

I like to think someone at Amazon is reviewing these recordings.

> Backstabbing, politics, bad engineering, nepotistic promotions.

What job doesn't have this?

This is next level at big tech where at one point the money seemed near infinite so nobody cared what was done. I remember seeing presentations from VPs and Directors which were totally misleading and just plain lying if you had any clue about the subject matter.