TYPE_FASTER

Of course there is one available on Amazon: https://www.amazon.com/live/video/0a7d56413360489dbf25611edf...

Also: https://www.amazon.com/Kooteyybhyy-Ornithopter-Generation-He...

Which may be re-branded version of this:

https://www.youtube.com/watch?v=v6nuMDV6n18

https://www.amazon.com/stores/Hanvon/page/426FE09E-17DC-4FBB...

I found this forum: https://www.rcgroups.com/ornithopters-101/

Which led to this video: https://www.youtube.com/watch?v=yWqBoA5CfPs - really neat to see it climb and glide.

There is decent RTK coverage in the US these days: https://e38surveysolutions.com/pages/ntrip-rtk-network-acces...

The AI bubble feels similar to the Dot.com bubble. The pattern is the same, right? There's some tech, a hype wave that people try to surf (if you paddle fast enough, you can start something and cash out in time), and when the water recedes before the next wave people/organizations either rode the wave cashed out and walked off the beach to their next thing, didn't make the wave and will try next time, or got pounded and their board broken. Or they ripped a great line, did a backflip like Medina, threw their hands in the air, and paddled back out to catch the next wave.

But you don't always know what the wave is going to look like when it's building. You just know there's a wave, and either you get on it or you don't. The connectivity of the web was obvious, the monetization wasn't super obvious (remember K-Tel anybody?). The power of modern AI is obvious, the monetization and final form of the tech isn't. I mean, using natural language is cool and all, and I think there is a lot of value in using models to help/assist engineering and other work, but I think it's too soon to say what the end game is.

Go for a quick walk. I realized this after fighting something for a few hours, then figuring it out on the walk out the door to go home several times.

I think Apple's success at using power management data from their mobile products to make computer hardware with really good power management is highlighting just how bad computer power management has been.

When you close the lid on a laptop, there are a lot of layers that all have to do the right thing. How is Windows configured? How do the drivers installed on that laptop handle the Windows state transitions? How do all the pieces of hardware on that laptop (CPU, etc.) work together to implement the various states?

I think it is possible for a computer manufacturer like Framework to work with operating system vendors like Microsoft and Canonical, and hardware vendors like Intel and AMD to improve how power management is implemented in their hardware.

There is always some level of "friction" involved when you are trying to integrate across different vendors. Some of the best Windows hardware I've used was made by Microsoft. The Surface line, at least in my experience, is really good.

It will require an investment of course, but I think it is possible.

> He gets to go down looking like he’s fighting instead of unentertaining.

Even better: he uses the opportunity to make it clear that any kind of comment that is out of line with the administration will not be tolerated.

He didn't go down fighting. He went out demonstrating the consequence of not agreeing.

THANK YOU

I find this really cool for some reason.

Exactly what I was wondering.

Also, they are probably trying to figure out how to replace search traffic that switched to ChatGPT.

While the article refers to Florida and Louisiana, due to increased precipitation density we're seeing more intense flooding happening more frequently in areas that aren't near the coast but are near a river. Many small towns in New England were established near a river for power and transportation. In some cases, they've now been flooded on the same day three years in a row.

These didn't used to be disaster-prone cities, but they are now.

And seeing as my premiums have been doubling every 8 years or so, I'm assuming that cost is distributed across everybody.

...by then it might be working again anyway, or the user figured out what they were doing wrong.

"Hey, is it still broken? No? Great!"

Also see https://github.com/Mozilla-Ocho/llamafile.

I used to know a bunch of phone numbers by heart. I haven't done that since I got a cellphone. Has that had an impact on my ability to memorize things? I have no idea.

I have recently been finding it noticeably more difficult to come up with the word I'm thinking of. Is this because I've been spending more time scrolling than reading? I have no idea.

I was lucky to have a high school math teacher who derived calculus with us. Understanding calculus made understanding physics so much easier. Being able to solve some physics problems using calculus seemed like magic.

Of course, I've never had to use any of that knowledge since, but I'm glad I went through the process to acquire it.

I’ve used feature flags to manage logging verbosity and sample rate. It’s really nice to be able to go from logging very little to incrementally pump up the volume when there’s an incident.

Similar experience here. I usually take the weekend off from the medication. Sometimes it serves as a reminder of what it was like before.

I think the reality is, we need identity on both the client and server sides.

At some point soon, if not now, assume everything is generated by AI unless proven otherwise using a decentralized ID.

Likewise, on the server side, assume it’s a bot unless proven otherwise using a decentralized ID.

We can still have anonymity using decentralized IDs. An identity can be an anonymous identity, it’s not all (verified by some central official party) or nothing.

It comes down to different levels of trust.

Decoupling identity and trust is the next step.

Yup, this is the way. Assume everything is AI unless proven otherwise.

My Twitter account was hijacked earlier this year. I created it in 2007 using an e-mail address on a domain I've owned since 1999. Multiple attempts to recover the account have been rejected.

I created a new account. Since I'm not following anybody, I see a random-ish feed of things.

It's kinda jarring going from a feed of tech science music to misinformation getrichquick scams and reposted viral content. Like woah.

Yesterday I did spent about 30-60min reverse engineering some existing code.

A co-worker spent the same amount of time having a LLM do the same thing, and produce documentation that now lives in a wiki.

It would have taken me another 30-60min to produce the same documentation.

The difference is that I, as a person/team member, understand it.

If the model is effectively a team member and is included in all the same conversations ("hey, what about...") then I think it could be roughly equivalent.

Personalization could include showing the same cached content to many people. The personalization here could refer to which content is served from the CDN to which users.

Yup. Content that may be read-only from a user's perspective might get updated by internal services. When those updates don't need to be reflected immediately, a CDN or similar works fine for delivering the content to a user.

When changes to data are important, a library like Hollow can be pretty magical. Your internal data model is always up to date across all your compute instances, and scaling up horizontally doesn't require additional data/infrastructure work.

We were processing a lot of data with different requirements: big data processed by a pipeline - NoSQL, financial/audit/transactional - relational, changes every 24hrs or so but has to be delivered to the browse fast - CDN, low latency - Redis, no latency - Hollow.

Of course there are tradeoffs between keeping a centralized database in memory (Redis) and distributing the data in memory on each instance (Hollow). There could be cases where Hollow hasn't sync'd yet, so the data could be different across compute instances. In our case, it didn't matter for the data we kept in Hollow.

Yeah. I’ve found it helpful to have a server in the basement for weird dependencies like this. Somehow running SQL Server in a container on an old Mac Pro feels like I’m still (somewhat) in control of my own destiny. Plus, it’s old enough that 64G of RAM was cheap.

I have not read the paper yet, but here is something related from Apple: https://machinelearning.apple.com/research/beyond-sensor

Do bicyclists and pedestrians really require the use of sensors? There are a lot of paths around me that cross busy roads with traffic signals. We just press a button. It speaks after you press it for visually impaired users and shows a red light after you press it for audibly impaired users.

There a couple different ways of learning calculus. I was lucky enough to have a high school calculus teacher that introduced us to calculus by showing us how it worked. Buoyed by that experience, I took math for math majors as a college freshman. I spent the semester being either extremely enlightened or completely lost. I ended up re-taking the second semester of calculus, but I took calculus for engineers instead of math majors. Not super enlightening, but enough to understand what to do when prompted for a derivative or integral.

Honestly, I think a big part of the difference is simply how concepts are communicated. Richard Feynman really was that good at explaining physics concepts using every day examples.

It's also not always obvious when you need to stop and go into detail, and when a general grasp of a concept will be enough.

Hey, at least the AI took responsibility and owned the mistake.

We're at the uncanny valley stage in AI agents, at least from my perspective.

Did it transcribe the dog? How long until we get a LLM that can translate dog barks...

How you store the bits, how you transmit the bits, and how you visualize the bits are three very different things.

> Data needs to live in a database - not an Excel file.

> Access should be granted for each according to their needs.

I mean, for enterprise users it's pretty hard to beat an Excel file in SharePoint.

You could model the data, put it in a relational database, and build a simple web app for CRUD operations, but unless you're adding additional business logic, I've noticed a lot of the time you end up with a table view that's not as powerful as a spreadsheet and maybe a form to enter data.

I definitely agree that it doesn't make sense to e-mail spreadsheets around, but persistence and user interface are different layers.

Sounds familiar. One of my first classes as an undergrad started with boolean logic and boolean gates, then IIRC (it's been a minute) assembly for a theoretical processor.