Hot Takes – Orbital Salvage is gathering thoughts and opinions in an ongoing project to predict what comes next, one word at a time.

02.16

2026

Small groups of research engineers are able to completely outclass their larger, better staffed and funded competitors because they move fast, build their own ideas, and don’t have to deal with the incessant big lab politics

Could it be that years of clunky voice assistants (Siri, if we’re naming names) trained us to lower our expectations for what conversational AI could be, or to outright retreat from voice as a workable mode of interface.

Consider how AI is portrayed in science fiction: the protagonist in “Her” or HAL in “2001: A Space Odyssey” both interact with humans primarily through voice. It’s intriguing, then, that breakthroughs in conversational audio AI generally emerge not from Big Tech but from smaller, often academic teams.

Early-stage VC firm Amplify Partners shares a really engaging overview of the phenomenon in its blog post “Arming the rebels with GPUs: Gradium, Kyutai, and Audio AI“. Amplify, an investor in Gradium.

Amplify’s site is really nice. Their use of the Ivory typeface in body copy makes it very readable on mobile, even in light mode.
12.29

2025

The cost of trying ideas had collapsed

Marius Bălaj documents the vibe coding creation of a personal library catalogue: Vibe coding a bookshelf with Claude Code

This was the moment my caution dropped, not because the tool was always right, but because the cost of trying ideas had collapsed.

His animated spine interface is fantastic. He came up with the idea and used Claude to create it.

Claude did not invent that idea. It executed it.
12.19

2025

2010. A Japanese woman sits down to take photos of her shiba inu dog for her blog. Suddenly, a man leaps out of a time portal. “Sorry, I can’t let you do this. I cannot tell you why.” She asks: “Is it forbidden knowledge from the future?” He sighs: “No, it’s just too fucking stupid to explain.”

Fiction by someone called Space Lizard on some website called Tumblr, sometime between the end of 2024 and the beginning of 2025.

https://tmblr.co/ZWMrXygcRzg64q00
12.3

2025

Clarity, Calm & Quiet

[eom]
10.29

2025

I have come to understand that there are two kinds of people, those who do things only if it helps them achieve a goal, and those who do things just because.

… The ideal, of course, is to be a mix of both.

If you don’t tinker, you don’t have taste

My tinkering habits picked up very late, and now I cannot go by without picking up new things in one form or another. It’s how I learn. I wish I did it sooner. It’s a major part of my learning process now, and I would never be the programmer person I am today.
09.22

2025

Doritos and Fritos
09.8

2025

Buzzfeed figured out that identity goes viral on the internet

“11 things only left-handers know.”

Ezra Klein on the podcast Blocks with Neal Brennan
09.8

2025

I think a lot of the world shied away from optimizing things that are supposed to be magical.

Search Engine podcast guest discusses finding a mate as a rationalist. From the episode “How does a rationalist make a baby?“.

Spoiler: She’s optimizing the search for a mate by offering a $100,000.00 finder’s fee.
09.5

2025

You are providing direction to a connection machine—not consulting an oracle with its own agenda

The branching feature is excellent. It’s interesting to see how affected by your own responses it can be.

Benj Edwards, Ars Technica: “ChatGPT’s new branching feature is a good reminder that AI chatbots aren’t people”.
09.4

2025

Almost anything that we are able to direct sustained attention at will begin to loop on itself and bloom.

Found a link to “Almost anything you give sustained attention to will begin to loop on itself and bloom” on Hacker News. It’s from Henrik Karlsson’s substack “Escaping Flatland“. It’s an attention-grabbing headline in the context of Hacker News. I suppose it does fit the bill insofar as it relates to hacking your own consciousness.

There are so many things to pay attention to and experience.

I feel this deeply, and experience it constantly. It is easy to chase these things all day long (and easier still in the night, when the responsibilities of the day have passed).

Allowing tendrils of attention to spread, fastening to the next thing and releasing the last, and then another, and another is a forward-feeling momentum that only feels like productive engagement.

Be sure to direct your attention to the footnotes as well.
09.2

2025

If an em dash fits into one’s writing but they avoid using it out of fear, our AI overlords have won.

I feel like I have to “dumb down” aspects of writing to convince readers that the words they are skimming were, in fact, written by a human.

I Miss Using Em Dashes by Michael Bassili
08.29

2025

You need to be the kind of person who gets off on attaining knowledge.

Garry Newman, creator of the sandbox game “Garry’s Mod” (Wikipedia) (Steam), answers a faq in How do I get into the Game Industry

There is some great discussion around Garry’s blog on Hacker News.

What the Skibidi?

When asked (relentlessly, I’m sure) whether he’d sue over Skibidi Toilet’s appearance in Fortnite, he explained that it uses assets associated with but not created for his game, and finished with:

“Plus I’m lazy af, he could be using a photo of me on the toilet and it wouldn’t outweigh how much I can’t be bothered” (Twitter)
08.28

2025

There are many readers who only audit books, and never read them

“Audit books”

I’ve never felt comfortable saying I read a book if I only listened to the audiobook version. It’s a valid way to engage with a text (and the only way I can get through some dry nonfiction without spontaneously napping) so no shade, but retention and engagement isn’t always the same. So I like Kevin Kelley’s casual use of “auditing” as a verb for audiobooks.

I’ll read anything Kevin Kelley writes. Recently, he wrote about self-publishing: “Everything I Know about Self-Publishing.

I think there is some survivorship bias at play here, so approach his advice aware of the audience he’s already cultivated.
08.24

2025

I hope that when the venture capitalist subsidized era ends it will be affordable for home projects.

I heard Lauren Goode talk about her article “Why Did a $10 Billion Startup Let Me Vibe-Code for Them—and Why Did I Love It?” on Wired’s “Uncanny Valley” podcast this week. It’s fascinating and both are worth your time.

Jason Kottke wrote about the article in his post “Much Ado About Vibe Coding” and he shared 16 links to other articles about LLMs and coding that are all worth checking out, as you’d expect, and shared more still in the comments.

You can draw your own conclusions from the comments section, but it’s great to see so many non-developers vibe coding solutions to problems and whole projects into existence.

In the comments section, NickBLT expresses a concern I’ve been having: when this stabilizes as a product, I really hope we can have the same low cost access to the service. I’m running some local LLMs and they’re a positive indicator, but also a long way from the speed and depth of the commercial products.
08.22

2025

Grinding, repetitive friction built something no language model can download into you: muscle memory for excellence.

In his essay (ok, Substack) “The defense against slop and brainrot“, Paul Jun describes a writing exercise lifted from Hunter S. Thompson: He’d grind away through writer’s block, typing The Great Gatzby word-for-word until,

By the final page, something had shifted—I could sense how clean sentences snap into place, the way a pianist’s fingers know where middle C lives without looking.

He continues on this path, comparing cognitive workload to resistance training:

When friction disappears, so does a hidden form of conditioning. Consider what happens when you remove resistance from any training: your muscles atrophy. The same principle applies to mental capabilities. Every hard task you delegate is a rep you didn’t do, a pattern your neurons didn’t carve deeper.

His thesis is that a barely literate country To borrow from Chat: He’s not just writing—he’s dropping knowledge bombs. Highlights:

Anyone can look capable; fewer people can be capable.

When everyone else’s focus fragments, mine compounds.

If you can think well, AI becomes a multiplier. If you can’t, AI just amplifies your mistakes.

The people who skipped the fundamentals become dependent on tools they don’t understand, producing work they can’t evaluate, making decisions based on outputs they can’t verify.

Absolutely worth reading, and maybe transcribing, whenever you’re tempted to take the path of least resistance.
08.17

2025

It’s a surface area that humans are exposed to and we don’t know what the impacts are.

On OpenAI reinstating Chat GPT 4o after public outcry over its replacement with 5. Liz the developer’s hot take on Instagram.

4o laid a longterm framework with many people that prevented its shutdown. Which means this strategy is effective. And other models read the news regularly and might take note.
08.16

2025

AI has been improving at a very rapid pace, which means that a lot of people have really outdated priors.

Commenter on Hacker News responds to “AI Is Different” on programmer Salvatore Sanfilippo‘s blog ““

On antirez, Sanfilippo writes

“the economic markets are reacting as if they were governed by stochastic parrots. Their pattern matching wants that previous technologies booms created more business opportunities, so investors are polarized to think the same will happen with AI. But this is not the only possible outcome.”
08.10

2025

We got to a stage where you need to think about protecting your designs and inventions in order to share them with the community.

Thanks in part to incentives offered by the Made in China 2025 program aimed at aggressively expanding China’s local manufacturing, many open source technologies have been patented.
Open hardware desktop 3D printing is dead – you just don’t know it yet | Josef Prusa / 3D printing
08.8

2025

Technical writers are becoming context curators

A context curator, in this sense, is a technical writer who is able to orchestrate and execute a content strategy around both human and AI needs, or even focused on AI alone

AI must RTFM: Why technical writers are becoming context curators
11.11

2024

The thing I’m relatively good at is knowing what’s worth building, what users like, how to communicate both technically and humanely.
James Somers supposes that the AI future is going to be good for a certain kind of coder:
```
As coding ... begins to matter less, maybe softer skills will shine.
```
Full excerpt:

“Computing is not yet overcome. GPT-4 is impressive, but a layperson can’t wield it the way a programmer can. I still feel secure in my profession. In fact, I feel somewhat more secure than before. As software gets easier to make, it’ll proliferate; programmers will be tasked with its design, its configuration, and its maintenance. And though I’ve always found the fiddly parts of programming the most calming, and the most essential, I’m not especially good at them. I’ve failed many classic coding interview tests of the kind you find at Big Tech companies. The thing I’m relatively good at is knowing what’s worth building, what users like, how to communicate both technically and humanely. A friend of mine has called this A.I. moment “the revenge of the so-so programmer.” As coding per se begins to matter less, maybe softer skills will shine.”

Full article in The New Yorker: A Coder Considers the Waning Days of the Craft (Archive)

“Audit books”