pacavaca

The time will tell how industry evolves, but it's clear that right now, memory is the biggest bottleneck for AI, and an enterprise-scale Semantic Layer might just be the answer for all things data

I'm trying to pick up some music skills, and part of that is doing ear training. Those who have ever tried it know that it starts with singing a bunch of stuff: just pitch matching, singing scales, singing chords, intervals, etc. Of course, the best way to do it is against a real instrument, but when you're especially bad at it like me, sometimes you want to practice with a tuner to make sure you're even hitting the correct pitches. Singing with a tuner is rather boring, so I made (yes, vibe-coded) a little browser game that is supposed to make it more fun. It's inspired by a TikTok mask with similar mechanics that went somewhat viral. Unfortunately (fortunately), I don't have TickTock, so I had to DIY it with a slightly different style.

Some gotchas: - It's very new, so, might be glitchy - Ideally, you want to sing with a drone, but it'll interfere with pitch-detection so, use headphones - It works on mobile (iphone), but I've seen some rendering issues. If encountered, try refreshing the page.

Otherwise, enjoy! I find the Random mode particularly fun.

Oh my! I've been looking for this comment Will be using it in the future to explain my feelings about Java and Python

Welp, I guess I'm calling it a day off

I remember trying neovim with Atom a while ago. At that time I realized that me using nvim it's not just about motions/plugins/no mouse, etc, but rather the whole terminal experience. For example, I rely heavily on tmux for splitting my screen whichever way suits the current task best and then, I may spontaneously open nvim in some of the splits or run other CLI tools. This experience is not the same with the CLI built into vscode or any other IDE.

Can I pipe it to Zapier?

Funny enough, those emails caught my attention too and I was even trying to `curl -v` my way through the redirect chain before stumbling upon this thread.

Weirdly, in my case the links in the email were pointing at lnkd.in/<some payload>, which is a legit Linked in domain. However, with that payload it was 301-me to some garbage script on google storage and then some shady website. I'm curious how did they make lnkd.in respond with 301 to whatever they want and can't there be a vulnerability of some sort.

What do you do if you have a script or a server that you need to make accessible to others in your team? Do you run it from your machine, or ask devops to deploy it somewhere? Is there something better than these two options?

Thanks, shmoogy!

Hey, HN!

Slack has recently released its Home Tab (https://api.slack.com/surfaces/tabs). Basically, it's a dedicated page within the Slack app where your bot can build a pretty advanced UI (way beyond basic messages).

While implementing the Home Tab for our main product, I realized that - Hey! This could be a perfect place for all my TODOs to live. I'm part of Apple, Google and a bit of Linux ecosystems and there aren't too many apps that I use consistently across the board. Slack is one of them and I often just DM myself to store a quick note or TODO that I can later view somewhere else. Why can't there be a bot for that?

Said - Done! I and the team hacked together the ToDoBot in 3 days and published it in the Slack store for everyone to use for free. Please check it out and let me know what you think!

Hey, HN!

We've spent the last five months playing with BERT, Universal Sentence Encoder, FAISS, and other cool tools, and look at what we've built!

You can point this thing at any of your Slack channels (preferably one with many Q&As), and it will automatically build an FAQ bot out of it^

The bot will then stay in the channel and respond to trivial questions using previous conversations - kind of what helpdesk teams do half of the times but manually.

^ - automatically, but with some human curation

^^ - yes, it screams privacy concerns, but we only scan what you tell us to, and helpdesk channels usually only have boring stuff about laptops, licenses, and VPNs not working. Please don't add our bot to your top-secret channels.

So, what do you think?

Duckduckgo is fine for like 80-85% of searches I do, but there are areas where Google is unbeatable (unbeaten?) yet

I find it crashed every other morning when I wake up my mac

It has to freaking stop crashing first. And stop freezing my computer with just 20 tabs open (for 32gb of ram). And complete their password manager. And implement recent standards to make apps work. And... Yeah, I'm still using it instead of Chrome just because I want to depend less on Google, but dammit Chrome is better.

Wow, thanks for the extended answer! I liked the part about "any code not written by me is a garbage", although I'm mostly trying to drop this attitude :D Jokes aside, I got your point, but it's still about catching errors upfront, during a more tedious code-review.

I've probably over-exaggerated a bit how "terrible" our merged code is but anyway, my point is that at the "review" phase the code is still "imaginary". We think that it'll work, but it hasn't been integrated with anything yet, it hasn't been battle-tested or, often, even used. Obviously, we find a lot of issues after the merge and filing tickets allows us only "not to forget to fix them". I feel like there should be a standard practice aimed at preventing those issues from RE-occurring.

Yeah, but the problems is that I (and I'm sure many other people) just can't pay enough attention to every detail DURING the code review. Yet after the review, when you start using the code, issues just reveal themselves without any extra effort. I'm just looking for ways to prevent them from reproducing I think.

Isn't that the case with pretty much any Slack bot though?

Completely on a side note, but I've just recently realized how much of a noise-cancellation magic comes buil-in with Zoom. My wife was grinding coffee literally next to me, and no one on the call could hear it. Now I'm learning that the whole standalone startup exists dedicated to doing it for all the platforms. Cool!

Anything that fixes the timezones on the list? Flat Earth?

Components used:

    Tensorflow and Simple Tensorflow Serving

    Pre-trained Universal Sentence Encoder

    Docker

    Python 3

    FAISS by Facebook

    ElasticSearch

    gRPC

Like the Toolkit!

Thank you, airocker!

Hello, HN!

Today I'm sharing a case and solution that we've gradually discovered and developed while running the original OneBar idea by our beta-customers. I posted a link to a blog post, because it better describes what we've built, but the whole thing is live and functional, so please feel free to signup and play with it!

Here's a brief summary:

Problem: A lot of internal teams have to deal with a flow of ad-hoc work-requests coming from Slack. In order to organize and track work they need to somehow turn Slack conversations into tickets. Slack offers "slack actions" to solve this problem, but they don't work well with arbitrarily structured conversations (e.g. multiple messages, threads, etc).

Solution We propose a completely new interface separate from Slack, where you can conveniently Select relevant messages from any discussion, turn them into tickets and then track in OneBar. Right now our workflow is mainly UI based, but we provide some AI assistance and plan to build up more of it in the future.

We spent quite some time building this thing, and I'd really love to hear your feedback, good or bad. Thank you! peace

Hello, HN!

We've been working on OneBar since spring, and ran it for about three months in a few companies. We're now ready to show it to the broader world! (it's still in beta though ;) )

A few things OneBar can do for your team:

- Store Q&As (just like Quora or StackOverflow)

- Search the KB right from Slack (via a bot)

- Source questions from Slack, and assign them to team members

- Save Slack threads as Q&A

- Automatically cross-link content

Here's one particular thing that worked well during the test: https://blog.onebar.io/crowdsource-your-companys-glossary-in...

We would highly appreciate HNs feedback on OneBar. Thanks in advance for checking it out!

Interesting. We already use something in our company that gives us "go/" links, but I'm not quite sure what it's powered by :)

Let's see what they do for the real YC W19 :D

Ubuntu 18.04 symlinks /usr/bin/vim to nvim when you install Neovim, so it kind of calls them the same. But ok, I agree that technically they are 100% different

But isn't Neovim a legitimate Vim?

But why should it matter for customers who does the job? I mean, if you don't tell anybody, and pretend it's 100% AI then it's bad, but if it "will eventually become AI", and your investors and everybody interested in the technical details know how it's actually done, then what's wrong? A true "AI" should be able to pass a Turing test, so for the customer, it should be indistinguishable, and shouldn't matter.

Of course, the privacy concerns are there, but then again, if it's a real "AI", then it may be worse for the computer to read your data than for a random low paid worker ;)