I am branching out from full-stack web design & development to some deeper backend topics especially Machine Learning. There is a steep learning curve that will keep me busy for a while.
The specific project I'll be applying these skills toward is a bit more nebulous, but I'm particularly interested in generative systems for text and images.
I am still plugging away on my snail simulation. Motivation comes and goes, but if I try to work on anything else I just go back to wanting to work on snails.
If I find something interesting, I want to be able to click a button (chrome extension/bookmarklet) that will extract the text from that page, format it nicely and email it to me (preferably to a separate Gmail label).
I can then read it from my email client (iOS/Android email app) during my commute.
I don't want the overhead of a complete app like Evernote or Pocket.
1. Last year I built a sports' klub listing website [1] which shows where you can train different sports on a map in Slovenia. The most challenging part is keeping the data up to date. This year, it's going to be about automating it fully, such that it can operate with almost no supervision, by means for crawling and automatic emailing klub owners to confirm the validity of the data.
2. Something based on hardware -- still thinking what to do with my Raspeberry 3.
As a piece of constructive criticism (as it seems like maybe English isn't your first language) using gendered words like "guys" can seem a little bit sexist or at least exclusionary. You don't want to put women off reading your book just with the blurb :)
What word would you suggest in its place? As a native English speaker I can't think of something similar which is both gender neutral and casual without being quirky (I'd use 'peeps' in conversations with friends, but not in marketing text).
I use "folks" and some people like "y'all" (iirc the Recurse Center took that as their "guys" replacement). Not sure how quirky that is for you though.
The word "guys" isn't so gendered in 2017 as it might have been 50 years ago. Far more importantly, though, gendered advertisements have had thousands of years of success. Don't insult or belittle any demographic, but keep in mind that it's entirely possible that many different sales pages targeting different demographics is the long-term solution.
For a first trial though, I'd narrow to a laser tight focus and pick one person you believe the book will most serve and write all your copy as if you were talking only to that one person. After you get uptake, expand.
It is a lot more neutral these days, you're not wrong but context is an important part of it.
I'm not a radical feminist or anything, I just think in tech it's quite important to be inclusionary at every given opportunity and every level. The male-only culture we've fostered in the tech industry is becoming poisonous. And nowadays it only takes one person outside of your laser-focused demographic to write the right [insert social media format] that gets your book seen. Especially if you don't have the budget/time/inclination to re-write copy.
Wasn't the context "guys who are convinced it will take them months to reach proficiency"? I could easily see a different person upset if it were "gals who are convinced it will take them months to reach proficiency", which indicates ambiguity in the context.
My point was that from a standpoint of odds of success, you're much better off with focused marketing and getting a strong response from a small group than by getting a weak response from a big group. I.e., see if anyone wants the thing at all and then focus on having a broad target to your marketing.
I agree 100% on being inclusive and am no fan of a culture where it's okay to exclude people from events or education based on their gender, race or other demographic identity. That said, lightly throwing around accusations of someone or thing being "___ist" devalues the terms and polarizes people who would otherwise be on the same page (and likely tipped last US elections).
I don't think I've accused anyone of being anything. I've pointed out some context to someone who's not fluent and may not have been aware of that context.
I believe that people want help finding "smart home" devices that will work well with the existing equipment in their home, and that they would pay good money for that because buying the wrong equipment is a frustrating (and expensive) mistake.
I'm building a way to help you pick the connected equipment that will work best for your home, based on what you want to control, how you want to control it, and the equipment that already exists in your house.
For example, I've got an Ecobee3 thermostat, Lutron Caseta lights, a SmartThings hub and an Echo Dot. My system will tell you the best connected door lock for your home that will work with that setup.
Just one suggestion, can you offer information on device security also? A collection of devices may work well together, but if I was buying a smart home device I'd want to be informed of any security issues as they're discovered. The escalations in DDoS size that we saw in 2016 were supposedly driven by a high number of insecure smart home/IoT devices, I'd like to do what I can to stop this becoming a trend.
>The escalations in DDoS size that we saw in 2016 were supposedly driven by a high number of insecure smart home/IoT devices, I'd like to do what I can to stop this becoming a trend.
Primarily driven by home routers, not IoT devices.
I want to make party/board games using smartphones as the game controller (similar to fibbage) and use this as motivation to learn elixir and phoenix together with a frontend framework (maybe elm). I tried to do this with python but I have problems with thinking in objects so I will go with functional programming.
This isn't a "side project" (I'm not going to make money from it, and it's also something that I work on at my job), but I've been working on umoci[1] which is a way to create and manipulate OCI container images. At (open)SUSE we're planning on using it to create container images inside the Open Build Service. I've also got a few ideas about RPM distribution of OCI images that I'm quite excited about, and hopefully I'll have some code that works soon.
> This sounds interesting but the use case seems very specific to your internal infratstruture and tooling.
This is actually also part of the openSUSE project[1], and it's all free software. The Open Build Service[2] supports several package formats and distributions (Arch Linux's PKGBUILD, Fedora/RedHat's RPMs, openSUSE/SUSE's RPMs, Debian/Ubuntu's debs).
> Can you elaborate a bit on use cases?
Currently at SUSE we're working on a new product for running Kubernetes on top of our distribution. One of the important pieces is building and distributing official SUSE images (as well as customers distributing images). But this code will be a part of the public Open Build Service instance, so everyone will be able to use it. And SUSE customers will also be able to use to to create their own images.
Effectively the "cool thing" is that you can create OCI images without needing to use something like Docker -- it's all just modifying a root filesystem and umoci will generate the diff layers for you.
I'm also working on distributing OCI images through RPMs, which may end up being a really cool way of distributing images (because all of SUSE's tooling is built to work with RPMs).
I probably shouldn't share these because they're nowhere near worth sharing yet but...whatever. I can come back in a year and see whether or not I actually finished this time.
- A HN-like forum written in Hack[0], mostly for self education and to teach myself how to work with Vagrant.
- Actually finish a game in the pseudoframework[1] I (mostly finished) in C++.
I want to build a set of tools to build a full private cloud on bare metal with zero single points of failure. Essentially a full replacement of Fuel and Openstack.
I've already started on a full multi master dhcp server to assign ip adresses to hosts and instances.
Tbh it's been a while since I looked into Triton, but it should operate in the same "feed it hardware and run virtual machines" space. I might experiment with running the components on top of smartos because of all the niceness that brings (i.e. crossbow, dtrace and zfs), but for now I’m building on top of Ubuntu.
well please make it so that it's truly easy to install. And I mean REALLY easy. I have tried at least 10 cloud solutions and I havent managed to successfully install any of them. The best one yet was tectonic but their error messages were too non-existent or vague at the end and the emphasis here is on non-existent.
Yes! I've been thinking of ways to do this, and the idea on top of my list is a live usb image that asks for stuff like network parameters, then configures the live image as a 90% functional node from which the first real host can be pxe booted. After the first host(s) have been installed the admin should be able to reboot the live host and add it to the cluster for real.
Right now i have a running Openstack cluster with Fuel for deploying new nodes over pxe. It works ok-ish, but it has some strange glitches every now and then. nothing production critical going wrong, but it still doesn't inspire confidence in me.
Yes I think your idea sounds very convenient, I basically just want to type the IPs of my machines in somewhere and the rest should be explained to me on screen. If SSH cant be established, tell me why. If i didnt set up ipxbe properly or at all, tell me why (heck, even tell me how you got that information ala 'we tried establishing an ssh connection with ssh@184.4882.1 -v but it resulted in this error log: '. Any other connection problems, tell me why. Generate all the SSL certs automatically, I dont want to type in any commands myself. I dont care if it's a test SSL setup but frankly I dont understand why these tutorials always give me test SSL certs. Just generate something that makes sense for production or tell me what I need and why for production.
If you think that my nodes should have DNS names, go start some internal DNS server for me and set it up in the background for me. I dont get why I should have to do any of that stuff myself.
At the end there should be a screen where all the configs were saved down to in a textual format, so I can have a look at what was done, which processes were started and which ports are now open, what the firewall looks like. For example CoreOS I believe has some cloudconfig stuff and I dont want to figure that file format out myself but I still want to see it after it's been generated.
I like magic but I also like to see what it actually did / is doing.
Personally i don't even want to care about the IPs of the physical machines, just put together a box, plug power and network in (given a properly configured switch) and boot it up from the rest of the cluster.
in the "boot up first box" scenario you should be able to enter all subnets you allocate to the cloud like "my wan range is 15.26.37.0/24, router at .1" And "for the host management network i want you to use 10.67.0.0/16 with router at 10.67.0.1" from there on you should be able to plug in and boot up machines. Just be careful to not plug in a laptop that boots from the network :P
Basically my philosophy is sane defaults, some magic where i know i wouldn't want to care, and introspection everywhere.
Yes i have, and i'm running MAAS on a colocated box as well to install ubuntu on virtual machines. But when you pair it with Juju the licensing becomes a bit expensive as far as i remember, So that's the layer I’m working on to replace first.
I would like to build a learning platform to cover the basic features of moodle using Phoenix, to learn both web development and functional programming.
120 comments
[ 3.0 ms ] story [ 221 ms ] thread2. An Investment portfolio management platform.
The specific project I'll be applying these skills toward is a bit more nebulous, but I'm particularly interested in generative systems for text and images.
EDIT: Never mind, Googled for "snail simulation" and found an old comment of yours explaining it. Looks cool!
If I find something interesting, I want to be able to click a button (chrome extension/bookmarklet) that will extract the text from that page, format it nicely and email it to me (preferably to a separate Gmail label).
I can then read it from my email client (iOS/Android email app) during my commute.
I don't want the overhead of a complete app like Evernote or Pocket.
See https://www.readability.com/
2. Something based on hardware -- still thinking what to do with my Raspeberry 3.
[1] http://www.zatresi.si/
[1] A crappy page about the book: http://jovicailic.org/mastering-vim-quickly/
For a first trial though, I'd narrow to a laser tight focus and pick one person you believe the book will most serve and write all your copy as if you were talking only to that one person. After you get uptake, expand.
I'm not a radical feminist or anything, I just think in tech it's quite important to be inclusionary at every given opportunity and every level. The male-only culture we've fostered in the tech industry is becoming poisonous. And nowadays it only takes one person outside of your laser-focused demographic to write the right [insert social media format] that gets your book seen. Especially if you don't have the budget/time/inclination to re-write copy.
My point was that from a standpoint of odds of success, you're much better off with focused marketing and getting a strong response from a small group than by getting a weak response from a big group. I.e., see if anyone wants the thing at all and then focus on having a broad target to your marketing.
I agree 100% on being inclusive and am no fan of a culture where it's okay to exclude people from events or education based on their gender, race or other demographic identity. That said, lightly throwing around accusations of someone or thing being "___ist" devalues the terms and polarizes people who would otherwise be on the same page (and likely tipped last US elections).
I'm building a way to help you pick the connected equipment that will work best for your home, based on what you want to control, how you want to control it, and the equipment that already exists in your house.
For example, I've got an Ecobee3 thermostat, Lutron Caseta lights, a SmartThings hub and an Echo Dot. My system will tell you the best connected door lock for your home that will work with that setup.
Do you have any devices in your house, yet?
Primarily driven by home routers, not IoT devices.
It's IoT devices too. Here's an example of an IoT device with subpar security discovered in the past couple of days:
http://hackaday.com/2017/01/02/owl-insecure-internet-of-ener...
(YouTubers say they are broken, YouTube says they aren't)
1. a cli spotify controller (control spotify through mpris and consuming the REST api for search/playlist related commands)
2. a cli GTD tool. I have some experience porting OmniFocus to Android (in Java) and I want to use what I learned from that to build my own GTD tool.
And start up a small VPS provider that will (at least) break even every month. (Any tips?)
[1]: https://github.com/cyphar/umoci
Can you elaborate a bit on use cases?
This is actually also part of the openSUSE project[1], and it's all free software. The Open Build Service[2] supports several package formats and distributions (Arch Linux's PKGBUILD, Fedora/RedHat's RPMs, openSUSE/SUSE's RPMs, Debian/Ubuntu's debs).
> Can you elaborate a bit on use cases?
Currently at SUSE we're working on a new product for running Kubernetes on top of our distribution. One of the important pieces is building and distributing official SUSE images (as well as customers distributing images). But this code will be a part of the public Open Build Service instance, so everyone will be able to use it. And SUSE customers will also be able to use to to create their own images.
Effectively the "cool thing" is that you can create OCI images without needing to use something like Docker -- it's all just modifying a root filesystem and umoci will generate the diff layers for you.
I'm also working on distributing OCI images through RPMs, which may end up being a really cool way of distributing images (because all of SUSE's tooling is built to work with RPMs).
[1]: http://openbuildservice.org/ [2]: https://build.opensuse.org/
- A HN-like forum written in Hack[0], mostly for self education and to teach myself how to work with Vagrant.
- Actually finish a game in the pseudoframework[1] I (mostly finished) in C++.
[0]https://bitbucket.org/kennethrapp/basedforum
[1]https://bitbucket.org/kennethrapp/sdl_framework
Possibly not in that order. None of these will set the world on fire, but I think they'll be good practice.
I've already started on a full multi master dhcp server to assign ip adresses to hosts and instances.
http://www.infoworld.com/article/2900832/virtualization/joye...
https://www.joyent.com/triton
Right now i have a running Openstack cluster with Fuel for deploying new nodes over pxe. It works ok-ish, but it has some strange glitches every now and then. nothing production critical going wrong, but it still doesn't inspire confidence in me.
in the "boot up first box" scenario you should be able to enter all subnets you allocate to the cloud like "my wan range is 15.26.37.0/24, router at .1" And "for the host management network i want you to use 10.67.0.0/16 with router at 10.67.0.1" from there on you should be able to plug in and boot up machines. Just be careful to not plug in a laptop that boots from the network :P
Basically my philosophy is sane defaults, some magic where i know i wouldn't want to care, and introspection everywhere.
A ready-to-deploy PostgreSQL Docker image automagically maintained by experts
- Simple, automated backups of your production databases
- We keep your database secure with automatic security updates
- Monitor your DB with a nice UI
- Avoid vendor lock-in: Containable makes it easy to migrate between different cloud services (or local setups)