scosman

43 ms·

661.
by scosman 3y ago
The is is amazing. I have a non speaking person in my life and the idea of them being able to create their own unique voice is so powerful. Stephen Hawkins kept his voice synth years after better tech was available because he wanted to keep
662.
by scosman 3y ago
Been a while. IIRC it didn’t work with millions of users, and partitioning with many users per partition would have same issues. Also looked at CLUSTER but there was no workable online version.
663.
by scosman 3y ago
I made same comment elsewhere before finding this comment. I can vouch for this speed up in the ratio of records per page. It is very real. Only applies for small records where you can pack many rows into a page, and where you can cleanly p
664.
by scosman 3y ago
Postgres for almost anything. Robustness, ecosystem, accuracy, all top notch. One exception: I did migrate two very large tables (15B+ rows) from Postgres to MySQL for performance reasons. InnodB (MySQL storage engine) can arrange the recor
665.
Deno SaaSKit: An open source SaaS template by Deno(deno.com)
1 points by scosman 4y ago|0 comments
666.
by scosman 4y ago
I’ll take the opinionated and enforced structure. Even when they aren’t what I would choose. I like ecosystems with standardization. That said… the one bit I can’t stand is every file having the same name. Working on 4 “+page.svelte” files
667.
by scosman 4y ago
Not really - more like a managed wireguard config system, with fallback VPN for NAT punch though when needed (so it always works, no matter the network). Traffic is direct when it can be, but when it can't it still just works. Nothing
668.
by scosman 4y ago
Mailchimp previously cranked up pricing on their Mandrill product dramatically, with minimal warning, no opt in, and unsympathetic tone from C-suite. Mailchimp is pretty hostile to developers. I don't recommend using them after that ex
669.
by scosman 4y ago
Yup. And apple is trying to make an OS safe for typical consumers who make mistakes and can be tricked. Hoops for you are a necessary burden to protect them. If you hate them, lots of Linux distros meet your desires. I personally think they
670.
by scosman 4y ago
Malware teams would loooooove that checkbox.
671.
by scosman 4y ago
That's a bit negative. Like you say, better at carrying kids than stuff (which is how I use mine), but that doesn't make it "quite shitty". I own this bike: it's very good at carrying kids, pretty good at carrying s
672.
by scosman 4y ago
I'd assume $5000?
673.
by scosman 4y ago
Manager perspective here. 1) Best thing you can do is understand the "why" for decisions you don't agree with. I find people get grumpy and quit because they don't like decisions being made, but are surprisingly timid to
674.
Voicebox: Using ML to help people with cerebral palsy communicate in real time(twitter.com)
4 points by scosman 4y ago|0 comments
675.
by scosman 4y ago
"salt $12" yup, def low.
676.
Bitcoins linked to defunct cryptoexchange QuadrigaCX mysteriously transferred(theglobeandmail.com)
3 points by scosman 4y ago|0 comments
677.
by scosman 4y ago
Since it was price driven we went with AWS SES. We were high-volume low-value per email. A high-value low-volume use case might have better options. I’ve heard good things about sendgrid. Someone who gives a damn about delivery rate.
678.
by scosman 4y ago
My take on raising prices (ran this playbook twice now, once at my company, once with a company I'm an angel in): - Grandfather in old users for as long as you can. Forever if it's feasible. More than likely future customer revenu
679.
by scosman 4y ago
I'm the creator, happy to answer any questions! Highlights include: - Low memory: the files are streamed out to the client immediately - Low CPU: the default server doesn't compress files, only packages them into a zip, so there
680.
Stream ZIP archives on the fly, with highly parallel micro-service(github.com)
2 points by scosman 4y ago|1 comments
681.
by scosman 4y ago
As mentioned in the comment, ML training workloads tend to be super chunky (at least in my experience). Some days we want to train 50 models, some weeks we are evaluating and don’t need any compute. I’d rather be able to spin up 200 gpus in
682.
by scosman 4y ago
We did a big analysis of this a few years back. We ended up using a big spot-instance cluster of CPU machines for our inference cluster. Much more consistently available than spot GPU, at greater scale, and at better price per inference (at
683.
by scosman 4y ago
I wonder if sample code using keys that don't look secure ("1111...") would lead to more or fewer incidents. Easier to guess, but more clear you're using guessable keys to the developer importing the code.
684.
by scosman 4y ago
OmniDiskSweeper on mac: https://www.omnigroup.com/more
685.
by scosman 4y ago
Tailscale + pihole is another way to achieve using same config everywhere. It's pretty easy to setup (by a HN definition of easy, would not suggest for any typical user). https://tailscale.com/kb/1114/pi-hole&
686.
by scosman 4y ago
Good find. Disproves the "only useful for web fingerprinting". It's also useful to their users for a fairly common flow. Don't assume malice, but do consider side effects of your decisions. This does add an extra bit to
687.
by scosman 4y ago
Looks awesome. How does the memory cache work? The architecture looks great, but so much of PG efficiency is cache hit rate. Do the pageservers have a distribution plan? Plans to add something like DynamoDB Accelerator?
688.
by scosman 4y ago
This has a few big improvements over Aurora. Aurora has scaleable disk, but is still 1 big master Postgres instance (and optional read replicas), with all the drawbacks that comes with in scaling speed, cost and failure modes. This appears
689.
Stream large ZIP archives without memory inflation(github.com)
1 points by scosman 5y ago|0 comments
690.
by scosman 5y ago
* flappy pong *

More ›