HN user

Molomby

35 karma

[ my public key: https://keybase.io/molomby; my proof: https://keybase.io/molomby/sigs/WCJ-VABcIGls24jSa-iZAGrPzLMp68mudUu-jQbCkrE ]

Posts2
Comments21
View on HN

Side note but, man, I hate it when write ups like this refer to events in relative terms without including a published date...

Last month [...]

At the same time [...]

That was over 6 months ago [...]

From when?!? What good is a relative timeline of events and your "current" thinking without a timestamp?

Digging though the source, it looks like this is from a few days ago.. "created_at":"2024-02-14T01:45:45.254Z"

/rant

I pushed it on the subject and, to paraphrase, it claims to be similar to an LLM (in that it's trained on a large corpus and uses statistical models to determine which tokens to respond with) but that, unlike an LLM, it uses machine learning to continually learn from its interactions with people. According to Pi, that's the key differentiator between it and an LLM.

It's bad. I have a ASUS MB16AHP (claiming a brightness of 220cd/㎡, contrast ratio 700:1) and yeah, in a brightly lit room viewing can be difficult. It's ok under normal office lighting but you'll be out of luck if you have sunlight flooding in. I've been searching for a brighter option but most (all?) the current offerings seem to be based on the same or similar internals.

Still, it's a useful device while traveling – nice to be able to plop a decent multi-monitor setup out of a backpack. My model also has a micro HDMI port (and an internal battery) which can be handy if you need a temporary screen for a Raspberry Pi or a machine that usually runs headless.

Select multiple files with different extensions > hold [Control] and right click > Get Summary Info > Open with (select all) > click [Change All...].

The files don't need to be valid, just have the extension you want to associate, so you can do something like `touch _.mp3 _.mp4 _.m4a _.mkv` then select those files and set them all at once.

I used "+" aliasing with a Google mailbox for years but never really loved it. It leaks your actual address and, presumably, isn't even that good at tracking who you gave which address too. Either spammers or even the company you supplied the address to initially can drop the + part to strip the "alias" away so, if your data is going to be leaked, it's not clear how much a "+" alias helps. Plus, some sites have broken/overly strict address validation so you're forced to handing out your real address to the crappiest sites.

Last December I got serious about the problem and went looking for solutions. Specifically, I wanted to be able to route email to my existing mailbox, have unlimited aliases, operate the whole thing via and API and not pay much – this is for personal use after all.

There are stacks of businesses that will give you email forwarding on a domain but I found most of them either only relay messages using a webhook, rather than forwarding the actual email (SparkPost, SendGrid, Postmark, Mailchimp) or are expensive, often with the email forwarding service being a feature point of some premium plan or having arbitrary limits (Mailgun $35/mth, MailSlurp $21/mth, ImprovMX $9/mth for 100 domains x 100 aliases, MailerSend $25/mth).

The standout at the time was Forward Email (https://forwardemail.net) who offer unlimited domains/aliases/messages for $3/mth. They're privacy focused, open source and have some nice features like regex aliases with substitution in the forward address which is cool. (So you can have rules like `/^(domains|webmaster|admin|root)$/@example.com` forwarding to `$1@example.net` and `me+$1@example.com.au`)

I now have a bash script that wraps the Forward Email API. It takes the url of the page that's asking for an email (to be logged against the new alias) and an abbreviation for the service, and it spits out a new alias, with the abbreviation and a random component, that looks nothing like my real address. I registered a domain specifically for email which is short, easy to type, and a .com (so even the harshest/most broken form validation will accept my aliases). New aliases are immediately forward to my Google mailbox and still include the service abbreviation as a "+" alias, so I can still see at a glance who sent the message. (For example an alias for Hacker News might be `hn.u2la1@example.com` which would forward to my mailbox as `me+hn@example.net`). Any mail sent to a non-alias is silently black-holed.

I've been using this for all new accounts for a few months and it's working well for me. Forward Email isn't perfect but they're cheap and seem trustworthy. Support (which I used once) was great though I'm pretty sure the entire service is a one man band so arguably the bus factor is high. The docs aren't fantastic but has been working flawlessly since I got it running.

After I had this all setup I stumbled on another service, Hanami (https://hanami.run), which would also probably work for me ($2/month for 4 domains, unlimited aliases, regex routing). And now of course Cloudflare's solution is publicly available (https://developers.cloudflare.com/email-routing). If I was starting from scratch I'd definitely consider those options.

I had a great experience recently on the scifi StackExchange, specifically with the `story-identification` tag. My description of an obscure short story was quickly identified as content posted on the (now-defunct) site Kuro5hin. The answer pointed me to the author's website and other stories which I've now purchased.

Apparently Google had delisted the authors site as duplicate content because the content was originally posted elsewhere, blocking me from finding the story when I'd searched previously.

https://scifi.stackexchange.com/questions/tagged/story-ident...

Adblock Stick 5 years ago

Ad blocking is fairly mainstream these days but I've never seen (or even really considered it could be done) it in this format. The marketing material is unsurprisingly light on technical details but, assuming this is real, possibly it installs itself as a network device and filters DNS queries similar to Pi-Hole..?

What confuses me is that, surely you could achieve the same outcome using software with very little overhead. Doing so saves all the manufacturing and distribution costs so you'd expect it would be a more profitable approach. So why the stick? Is the filtering overhead large enough to justify a dedicated co-processor or is the form factor used here really a placebo? In the way that a heavier wrist watch feels more valuable, does plugging a physical device into your computer to protect it/make it faster, _feel_ more effective than installing software that does the same job?

This sounds great but i’m surprised more people aren’t recommending AWS as a modern “upsell free” alternative? Their prices are good, can be secured with TFA, solid APIs, etc. That specific part of the management console isn’t amazing but it’s powerful and if you’re using it a lot you should probably automate, right?

Totally agree with your sentiment but in this case I don't see how rate limiting helps, or at least not much.

If you're building a botnet you don't need to crack any specific machine so you can distribute your attempts across more hosts. Eg. rather than hitting one machines with 1,000 pwords/sec you target 1,000 machines at 1 pword/sec each (or whatever rate you're limited too). There's no shortage of badly configured routers.