i wrote something similar https://github.com/dwrtz/sink

it has some nice features:

- follows .gitignore patterns

- optional file watcher to regenerate the snapshot when files are changed

- supports glob patterns for filter/exclude

- can report token usage

- can use a config yaml file if you have complicated filter/exclude patterns

i find myself using it all the time now. it's great with claude projects

Yeah one needs to distinguish between the optimal state sequence and the optimal state at a particular time. In other words, a joint posterior (state trajectory up to now given all measurements up to now) vs a marginal posterior (state right now given all measurements up to now). the KF only gives you a marginal posterior. they are not the same

yeah the "filter" term means something different in the KF context and is confusing

filter: use measurements up to time t to estimate the state at time t

smooth: use past and future measurements to estimate the state at time t

predict: use measurements up to time t to estimate the state at time t+n

Sorry to hear you stopped working on your tool. My cofounder and I are working on a similar product (teraphone.app). We're focusing specifically on making a discord-like product for microsoft shops. I agree it's partly a culture thing. There's also an awareness issue. A lot of people don't know that pop-in style communication tools exist, especially outside the tech industry.

Interesting, do you think you have a chance at competing against Microsoft's own tools, like Teams?

Regardless how crappy Teams is at certain things, the people that decide "we're using Teams" often barely use it themselves.

Yeah we decided to integrate with Teams because I had that exact experience in the past. I was at a company that decided "we're using Teams". We weren't allowed to install alternatives like Discord or Slack.

Teams itself has some limitations that are easy to improve upon. They have no way of doing a huddle/pop-in-style room. They screw up the presence signaling in a way that discourages pop-ins: if you join a call with someone your and the other participants' status is set to busy/unavailable. From other peoples' perspective this isn't very inviting, and doesn't indicate that the conversation is public. Teams also limits who can share screens to only 1 person at a time.

Look at Microsoft buying sunrise, even though the server based calendar outlook feature was arguably fully fledged. There’s always space:

You may be thinking of put options. Buying calls is a bullish position.

What's wrong with push-to-talk?

It tends to fail to register the key being released for one, even occasionally on the desktop client.

Last time I tried it the browser app was especially prone to this, to the point of rendering push-to-talk completely unusable.

It's obsolete and only needed because their software is broken.

Skype, Google Hangouts, teamspeak, ventrillo, mumble, and tox all support my microphone and voice activation correctly.

Push to talk is most certainly not obsolete. I (and many gamers I know) use push to talk for the privacy of being able to choose when your microphone is activated.

Its even more useful in competitive gaming and/or streaming where many want to chose what to say to their team/friends and themselves/the stream.

Also helps a lot in large communities. There will always be people that are nice and you like to have around, but are a bit "louder" during gaming sessions. We were really successful by teaching them to not hit the PTT-button when they need to drop their load of frustration.

The gaming community that I am a part of that uses Discord holds a weekly big gaming session where almost everyone attends at the same time. It occurs around dinner time on the west coast, so some of our west coast members will be having dinner between rounds. Push to talk allows them to talk between bites without all of us getting the smacking sounds of chewing and eating in our ears.

Is push to talk a better solution than a toggle push-to-mute?

Push to Talk is necessary when you regularly have 200+ people, sometimes 800 people in the same channel. We don't use discord for comms because it doesn't support shout + channel hierarchies unlike mumble. Discord is still useful for pings, as their mobile app w/ push notifications are very helpful.

Definitely not. If you have been in a channel with 25+ people for a raid or large event (exactly the type of gaming community Discord caters to) then you should know push-to-talk is absolutely required and you will often be kicked if you don't use it.

I don't want to hear you chewing. Please consider push to talk.

PTT is useful when you don’t want everyone else on your channel to hear your breathing, coughing and keyboard-clacking

(can you tell I’ve played Overwatch?)

Yeah - there are games for which I will use voice activation, like PUBG (where I do not have the mental availability to hit a PTT key), but for Overwatch I started with PTT with one of the thumb buttons on my mouse and I still do it.

That said, I use a for-realsies dynamic microphone with tons of off-axis rejection and a wall-mounted arm (I play games in my office/recording space, that's the engineer station mic). So I probably could use voice detection and be fine. But habits die hard.

> keyboard-clacking

The rise of mechanical keyboards has made PTT practically the only option. Otherwise you just hear little clicking all the time.

Audio Weaver is a graphical tool for doing audio dsp. You use a drag-and-drop interface to configure your signal flow graph. When you "run" the system your graph is used to dynamically instantiate the necessary low level functions and compose them together. While the design is running you can do live tuning (changing gains, filter coefficients, delays, etc.). The low level functions have implementations on several architectures (PC, ARM Cortex-A and M, Tensilica HiFi 2 and 3, ADI SHARC, TI C66xx, plus a few more). This means that your signal flow graph is cross platform.

DISCLAIMER - I work for the company that makes Audio Weaver.

Awesome article! Here are two things that stuck out for me...

1.) In the first part of the article talking about decomposing a periodic signal into sinusoids of different frequencies - it would be more correct to refer to it as the Fourier _Series_, which is different from the transform.

2.) The image showing three sinusoids summing together into a non-period signal made me cringe (the sum is also periodic - that's the whole point of the Fourier Series).

1) ...which when applied to a sampled signal, is the discrete-time Fourier transform (DTFT)

2.) It will only be periodic if there exist some frequency for which the frequencies of the sinusoids are integer multiple.

It leaves you with an extra minute to decide if you're going to hit snooze again.

Higher inflation also means that it's cheaper for the U.S. government to borrow money. In fact, higher inflation is good for anyone who borrows money (people with mortgages, student loans, etc.) because they're effectively paying back less than they borrowed.

Except that the risk is asymmetric in that the bank gets bailed out if you can't pay and you don't. They loaned out the phony money and you paid it back with your labor.

Which is a problem of politics, not economics. There's a broad agreement that the bank bailouts should not have happened across most of the actual political spectrum. Only the vulgar neoliberals serving the interests of the capitalist class in government actually want to print money as bailout loans to banks, and yet Austrian "economists" and their Believers pretend that's the only alternative to a deflationary gold standard and total privatization of everything ever.

Evidence suggests that it they are correct but the part you are missing is that the whole point of the system is to bail out the large banks when their bets go bad. All in the name of the people. Whoever they ever are.

Which system?