HN user

float4

1,219 karma

Love all, trust a few, do wrong to none.

Posts1
Comments214
View on HN

As someone who's in his late 20s and didn't (consciously) witness the dotcom crisis I want to ask the older people here: was this also part of the dotcom bubble era? Were people working in bookstores angry at Amazon, people working in retail fashion angry at fashion ecommerce stores, etc?

Wouldn't you get >50% of the usefulness and 0% of the risk if you add read+draft permissions for the email connection through a proxy or oauth permissions? Then your claw can draft replies and you have to manually review+send. It's not a perfect PA that way, but could still be better than doing everything yourself for the vast majority of people who don't have a PA anyway?

It feels like, just like SWEs do with AI, we should treat the claw as an enthusiastic junior: let it do stuff, but always review before you merge (or in this case: send).

For me it's a bit of both. I'm working on exciting energy software with people who have deep knowledge of the sector but only semi-decent software knowledge. Nearly every day I'm reviewing some shitty PR comprised of awful, ugly code that somehow mostly works.

The product itself is exciting and solves a very real problem, and we have many customers who want to use it and pay for it. But damn, it hurts my soul knowing what goes on under the hood.

I should have read this 12h ago! This afternoon, I tried to create my first simple agent using LangChain. My aim was to repeatedly run a specific python analysis function and perform a binary search to find the optimal result, then compile the results into a markdown report and export it as a PDF.

However, I now realize that most of these steps don't require AI at all, let alone agents. I wrote the full algorithm (including the binary search!) in natural language for the LLM. And although it sometimes worked, the model often misunderstood and produced random errors out of the blue.

I now realize that this is not what agents are for. This problem didn't require any agentic behavior. It was just a fixed workflow, with one single AI step (generating a markdown report text).

Oh well, nothing wrong with learning the hard way.

You call it silly because you could be doing useful stuff on your phone. I'd go one step further and say that even if you're slacking off that's not necessarily a bad thing. Everybody, including politicians, slacks off from time to time. Be it due to stress, awful sleep because the neighbor's dog barked all night, illness, or something else. It's just human and there's little wrong with it as long as you do your job well most of the time.

Bezos nailed it on this topic: “[...] [I]n our retail business, we know that customers want low prices, and I know that's going to be true 10 years from now. They want fast delivery; they want vast selection. It's impossible to imagine a future 10 years from now where a customer comes up and says, 'Jeff I love Amazon; I just wish the prices were a little higher,' [or] 'I love Amazon; I just wish you'd deliver a little more slowly.' Impossible. And so the effort we put into those things [...] will still be paying off dividends for our customers 10 years from now. [...]”

You should consider what won’t change, and the following is a (non-exhaustive) list of things that I think won’t change: I believe AI is and will continue to gain intelligence

Okay, but that way you can frame every ongoing change as a constant. "Change X will continue, and because it's already ongoing and will simply continue, I consider it a constant and therefore add it to my list of 'things that won't change'". But that's clearly not what Bezos meant.

Except Dairy Queen, Wendy’s, and McDonald’s outside of the U.S. don’t have this problem.

Heck, even Ikea has successfully been selling ice cream (from self serve machines!) here in the Netherlands for like 20 years now. €0.50 back in the day, €1 now. Can't remember the last time all machines (yes, they do have at least 2 usually) required maintenance.

I was under the impression that Netflix was not deemed video sharing platform because they don't actively dictate what is and is not allowed on the platform the way YouTube does. But maybe it was the 10,000 rule, who knows. To my knowledge the EU never made explicit why companies were not deemed gatekeeper.

Multiple reasons:

1. Size. DMA requirement is >=7.5b turnover in the EU, or worldwide market cap >=75b. Not the case for Spotify.

2. Non-provision of core platform services. It's not just "important gateway between business users and consumers", it's "important gateway between businesses and consumers *in relation to core platform services*". Core platform services are e.g. search engines, operating systems, browsers, app stores and so on. This is why Spotify isn't a gatekeeper.

And the exact same reasoning applies to American companies. Video streaming platforms like Netflix are also excluded for example, even though Netflix has a market cap >=75b.

he made his students work use pens while taking his tests

This is very common in the Netherlands, I think that's why it was a rule of his.

In general, the Dutch education system seems to be against pencils (at least this was the case until recent; I'm Dutch and mid 20s). You're tought to write using a fountain pen, not a pencil. In high school, you're allowed to switch to ball point but absolutely not to pencil. In university, write with pretty much anything you want, but... not with a pencil. If you do take your test with a pencil, there's genuinely a chance your teacher will give you a 0, although most of the time they'll probably be forgiving.

I majored in CS in the Netherlands and every test was done with good old pen and paper. Students still make mistakes all the time, which is why everyone uses a scrap sheet.

look for the spouse sitting quietly on his phone, but please don't bother him

Might as well just be someone with social anxiety. Many people like that in tech.

Shakespeare had an exceptionally large vocabulary. He constantly used words his audience / readers didn't understand. And that's okay with you, because he was "expressing and transferring ideas". Except that it's not okay with you, because he didn't "use the words that [his] readers will understand and that express [his] ideas the clearest".

Good writing is about expressing and transferring ideas.

Not everything is a scientific paper. Good writing can also be art. There's a reason why Shakespeare wrote his 18th sonnet and not just "I think you're very beautiful".

What happens when you place your phone in the microwave (don't turn the oven on, obviously) and walk away with your speaker? I'm curious what kind of range you're getting.

For reference: I just tried this with iPhone 13 mini + WH-1000XM3 and the connection dropped after ~5 meters.

There is no one way of doing this. What you write is correct, but most people use the (IMO simpler) 3rd digit rule: the decade is decided by the third digit of the year. So 2020, 2021, ..., 2028, 2029 belong to the same decade, as do 2030, 2031, ..., 2038, 2039.

This indeed causes problems for the 1 - 9 AD decade, which only has 9 years this way, but most people won't care because we live in 2023.

The S3 was the real problem.

VanMoof released the VanMoof S3 in 2020, which was €2000 (the predecessor was close to €3000). The S3 sold very well because of its low price and great design, but its build quality was really subpar. In fact, insiders say around 10% of all S3 bikes had defects that required repairs (note that 1 - 2% is standard for a typical bike model).

So in the end, VanMoof sold many cheap bikes which was great for their market share but not so great for their margins. Then, on top of this, they had to repair many of the bikes they sold which ate into their margins even more. Lastly, because of supply chain issues, they weren't able to offset the repair cost by fixing the S3 issues and quickly selling many proper units to new customers.

I'd also be surprised if they used custom cores for smaller chips like for airpods/airtags.

AirTags have an nRF52xxx SoC by Nordic. The only Apple silicon in them is the U1 UWB chip (which is also present in iPhone, HomePod, Watch, ...) which runs on a 16nm Cortex M4.

Recent AirPods run on H1 / H2 SoCs though, which are custom Apple chips. Wouldn't be surprised if they are built around the E-cores from the iPhone SoC.

Bet $100 that all classical music

ALL classical music, in just 3 years. I'm perfectly willing to put my entire net worth on the line for this one.

PayPal has restricted our business account because we have invoiced a license key containing the random letter sequence “ALEP”.

This makes me wonder: what's the best way to generate "safe" license keys? Binary feels like an obvious solution (binary keys surely get through virtually all blacklists?) but at the same time: binary license keys would be very long and very atypical, so maybe fraud detection systems mark them as suspicious anyway.

Maybe just generate random alphanum license keys and run them through some open source blacklists yourself? I doubt "ALEP" is in those lists though.