It’s weird that I can read an article and tell that it was probably generated with an LLM. There are subtle ways that writing is structured along with word usage and argument construction that don’t feel like what a human would write. LLMs have also ruined the word “genuine” and its derivatives.

The thing about doing comprehensive research projects is that the results are often shocking at first, then they gain nuance and feel like they have less punch to them, and then you round the corner of finishing up and you sit in awe at how everything just came together.

I’m deep into some research. I keep running new experiments and proposing new hypotheses. With everyone so far I’m like “this can’t possibly be true” and with each one it winds up, somehow, being worse than what I had hypothesized.

Today has been an exciting day. This morning I sat down and thought “wouldn’t it be cool to generate transcripts for every episode of Fallthrough and Break?” and many hours later I not only have transcripts for each episode, I also have an entire analysis for each episode.

Are people who do DEI called Deities?

I feel like @matthewsanabria.dev would be proud ☺️

So I made chocolate tonight. I immediately fucked it up. But I turns out you can in fact save seized chocolate.

Being such a nuanced individual, not explaining why this is a precisely correct take is far too uncomfortable.

No Nuance November Hot Take: AI doesn’t exist and never will.

This tends to frustrate people. Especially when I can articulate that yes I was wrong, but a multipoint perspective does not necessitate that you were right just because I was wrong, therefore you can still be wrong even if I am wrong 😀

Some people have a lot of pride in their views and opinions. I do not. I’m fine with admitting I was wrong. The way I see it is if I admit I’m wrong and admit that someone else or some other viewpoint is right, then I’m right when I’m right and I’m right when I’m wrong (even if belatedly).

Baker intuition is when you put something the oven, and a couple minutes later realize you forgot to put the baking powder in, shrug for a minute and think “let’s see what happens”, then think better of it, pull it out of the oven, quickly mix in baking powder, then put it back in the oven.

I believe they call this a Humble Brag 🤭

Even after all these years of baking I have some… misconceptions. Like, I thought convection ovens were flat out better. To the point where I was confused why my rather high end oven had a conventional mode.

In today’s “words I’ve only read and never heard pronounced”: lecithin. Likely because I’ve heard “lectin” pronounced, that’s just how my mind hears and says it. Turns out it’s pronounced leh-suh-thin. At least I can pronounce it correctly now 😭🙃

I might be having a little too much fun with these recent titles 😆 You should definitely watch or listen though! This was a really fun episode.

It’s fascinating where the need for software tools can pop up. Maybe something like this exists, but I want some software that allows me to track ingredients and their pricing and then tells me the cost of a particular recipe, based on the usage of each ingredient.

Love when you preheat the oven and then it failed to actually preheat 😭

In college I was initially confused by the cost and need for synchronization signals. In the broadcast television world, a black burst sync gen was similarly confusing. One of the other engineers explained that it synchronizes frames so when you hot punch you don’t have frame tearing or a flicker.

Paging @matthewsanabria.dev, we require your opinions of the unpopular variety. And we could use some Panic & Recovers from @steveklabnik.com too 🤔

Me: Oh, I like this feature in FCPX that’s only in Sequoia and above. Let’s upgrade. Oh wait, there’s a pending update. It says it’ll take 15 minutes. No problem! *five hours and three OS reinstalls later* Me: Why, Apple, why??

We’ve been trying new things with @fallthrough.fm and some of the things I liked, but others I’ve liked a lot less.

Getting rid of old habits is tough. I’m implementing a type that marshals to text. I implemented some decode functions. Called one DecodeTick and the other DecodeTickString. I implemented the underlying function as generic, then realized I could just have a single generic DecodeTick function.

Pretty sure I’m mad about Paxos not being as popular as Raft. Of course I haven’t done a survey to see if this is actually true, but it feels like it. Time to go build some cool Paxos stuff I guess.

Having to reencode video because it’s variable frame rate is annoying (and I don’t have the energy to figure out why FCP makes constant frame rate proxies from vfr source). However, it does give me the opportunity to dig more deeply into FFmpeg. I will gladly take a CLI tool over a GUI, thank you!

I’m getting far too close to upgrading my 10GbE network to a 25GbE network. Do I *need* it? I mean, all my servers have two 10GbE links aggregated together for ~20Gbps of throughput… and I am doing all this video editing (and motion graphics soon enough)… so maybe 🤔

Got distracted while making an edit, forgot to finish it. Finished editing the rest of the video. Export audio and video versions for two editions. Review final video. Notice incomplete edit. (╯°□°)╯︵ ┻━┻

@matthewsanabria.dev and I set out to make a *short* aftershow for @fallthrough.fm. It’s a massive failure, because we can’t actually stop talking so the “let’s have an extra 20 to 30 minutes of content” has become “here’s an extra hour of content”. That’s not an aftershow. That’s just a show.

Super clamps are my new favorite rigging gear. Also, rigging things is fun. Such a nice feeling when you tug on something and it doesn’t move even a little.

Let’s see how rusty my After Effects skills are 🤔

What do you mean I’m down to 1 box of butter?? Need to start buying this stuff like 10 pounds at a time.

Completely on brand for @steveklabnik.com and @skriptble.me to talk about REST for two hours and still have more to say. Is there anything we didn’t cover but you wish we had? Should we talk more about how things like GraphQL are actually quite RESTful?

We finally did it! After recording an episode of @fallthrough.fm, @matthewsanabria.dev and I recorded our post show discussion (featuring @www.jvt.me!). That’ll be shipping alongside next week’s episode, which includes the wonderful @steveklabnik.com as well!

I love when we record an episode and wind up on such a long, but somehow still related, tangent that we accidentally made an entire extra episode. Also, when you let @steveklabnik.com cook, he will, in fact, continue cooking until you stop him.

The finished product. Sourdough chocolate chip cookies with flaky sea salt. So far they’ve been a big hit. I mean, who doesn’t like fresh out of the oven chocolate chip cookies? 🍪

Come for the hot takes, writing, and podcasting. Stay for the baking content.

Chop chop chop 🍫

Nothing like waiting for brown butter to cool in the fridge at 3:00am to make you question whether getting back into baking was a good idea… 🙃

One of the toughest parts of having a podcast is preproduction; simply because it's difficult to actually plan and schedule episodes. This was one of the most challenging things we faced with Go Time, and we came up short more often that I would have liked.

More baking! This time snickerdoodles. I really enjoy timelapse as an art form. There’s something beautiful about breaking video down into a series of photos, or really building a series of photos into a moving image. The sum tells a story not captured by the individual photos alone.

Back in this baking nonsense 😋

We spent 15 years arguing about how to make error handling better in Go and looking back the premise of that desire feels so false. Any “fix” for error handling would mean separating errors from all other values by necessity, but I don’t think there was an argument against “errors are values”.

Just finished mastering episode 24 of @fallthrough.fm, which is actually our 25th episode since our first was episode 0. For one, I’m amazed that we’ve been able to not only ship 25 episodes, but also do so every week (on the same weekday with a couple exceptions).

When posed as tools to help people in specific situations be more productive and efficient, I see no problem with AI tools. They fit really well alongside a whole bunch of other tools that people may choose to use in a given situations. I dislike the illogical jump to “no more programmers”.

Episode is edited and mastered! Early access supporters will get the episode of 5:00pm ET! For everyone else, it’ll be available on Monday at 8:00am ET 🥳

Just had a fantastic conversation with @steveklabnik.com and @matthewsanabria.dev about AI discourse and more. It’s basically a continuation of the conversation he had on Oxide & Friends, so if you haven’t already you should listen to that before the @fallthrough.fm episode.

The fact that we’ve been able to ship a @fallthrough.fm episode every week for the last 22 weeks is kind of wild. We’ve had some changes (e.g. YouTube decreasing in priority, Angelica doing more hosting, etc…), but I think that bodes well for our future.

Etymologically, the term “computer” has spent more of its existence referring to a human than it has referring to a machine. By a significant margin. Words can change meaning over time but I’m not convinced that a mechanical computer is doing something appreciably different than a human computer.

The more I attempt to define what a computer is, the more elusive the concept becomes. Not because it’s difficult to arrive at a definition, but because it’s become difficult to describe what people colloquially refer to as a computer in reasonable terms.

After many months and many different designs, I think that I’ve finally got the lowest layer of this DBMS designed! In regular fashion, it has wound up precisely nowhere near where I thought it would be. For example, I was absolutely sure that I’d use copy-on-write semantics. Also, snapshotting.