And 1932! https://www.movingimagearchive.com/sources/san-francisco-by-...

It just sends the PDF to Gemini Flash and asks it to extract semi structured data (semi structured in the sense that it's not fully structured extraction, but its a JSON file to feed into the next step of the LLM design process).

Works pretty well in general, though I should update to the latest Gemini version.

That's not how Metcal fixed temp induction soldering irons work. They don't have a control loop, which is what makes them so much better than other soldering irons.

They exploit the Curie temperature [1] of some alloys, in which the metal loses its inductive properties when it hits a certain temperature. A Metcal power supply pumps a simple 13.56 Mhz signal into the tip which heats it up and when it reaches its Curie temperature, it just stops heating. Since the signal is constant, whenever the temperature of the tip drops it just heats back up without any PID loop. The downside is that you have to switch tips (alloys) with a pair of pliers to change temperatures, but the upside is that there is no control loop delay.

It's also why their station are so reliable. There's basically no "modern" electronics in them and the worst you have to do to fix them is replace an electrolytic capacitor. I've got a power supply manufactured in the late 90s that's still as good as new and like you said, going back to a Hakko or any other soldering iron is downright painful.

(Side note to anyone who cares: when the patents expired, a couple of Metcal engineers left to form Thermaltronics, which sells cheaper stations and Metcal-compatible tips)

[1] https://en.wikipedia.org/wiki/Curie_temperature

> FWIW - Computer vision is also NP complete, but we do that all the time now.

I have no idea what you mean by this. What's your definition of NP complete?

You and everyone else. That's the great mystery of transformer architectures as applied to language.

To be clear though, they're only good at schematic capture, which is very much a textual representation. Most of the data basically boils down to netlists, which are a text based format mapping connections between abstract pins that only later map to physical copper. The actual schematic portion is for human consumption and LLMs don't need to produce those to be useful.

Where LLMs completely break down is the next step, PCB routing. That's an NP-complete research problem that's been ongoing for decades without much progress. I've had some fun playing with using LLMs to better specify DRC rules in Altium so that the "classical" algorithms are more usable, but at the end of the day their geometric intuition is nonexistent.

[1] https://en.wikipedia.org/wiki/Xfinity#Xfinity_WiFi

It depends on what exactly you mean by "commercial value commensurate with the costs involved" but I'd volunteer the 3G/4G/5G specifications and the other documentation required to implement the mobile network protocols. 5G is currently sitting at over 50,000 pages and it's one of the reasons Qualcomm/Broadcom/Apple are the only ones who can realistically make a mobile radio.

I don't think there is a single human to whom more than a few thousand pages would be comprehensible at a time except for the occasional genius.

There is zero useful evidence because there is no one to collect it.

The Institution of Civil Engineers was founded in 1818, after decades of random civil engineering societies in Britain doing the exact same thing we are now (running around like chickens with their heads cut off). It wasn't until after the ICE's Royal Charter a decade later that civil engineering began to get really systematized into the "real engineering" we know today and that charter effectively established them as a regulatory body that allowed that to happen.

It means you put "i.swear.this.is.localhost [remote ip]" in your hosts file.

Bring a software engineer into a courtroom as an expert witness, and the jury's eyes will glaze over. Bring in the PE who told their firm not to cut that corner, and the hammer comes down hard.

Even if the certification for software engineers starts as barebones as knowing what WASP is, it still provides an avenue for the feedback mechanism to work (the rules "written in blood"), so that the entire industry can study and learn from what happened, instead of this mess we have now, the peak of which is postmortem blog posts. Even now we have plenty of examples of regulatory frameworks where the regulations adapt to the field like the FDA where you've got a huge spectrum ranging from diagnostics to medical devices of which where are many classes, and drugs where every clinical trial can be tailored to the exact nature of the disease.

That's also the entirety of the biotech VC industry, except even more so.

Biotech startups require far more capital to make it to market than tech startups, so biotech VCs take the role of early stage funding for R&D. Then the startup IPOs (with zero revenue and unlimited scientific risk) and uses that to fund its clinical trials, while the investors dump their stock. Then, because ramping up manufacturing and quality control takes even more money, the biotech startup sells out to a pharmaceutical company once its passed its trials or the results are promising enough to take the risk, closing the financing loop. Most of these startups stop existing before they even earn a single dollar in revenue, either because they fail or they're acquired by someone who can actually manufacture and distribute their product at scale. This has been the way of the industry for at least the last 20-30 years ever since the small molecule cliff hit like a brick wall.

So, the VC industry whose standard operating procedure is to unload companies onto public investors long before anyone even knows whether the drug or device would work, wouldn't touch Theranos with a thirty foot pole.

I'm wondering what data you're using and if you can impose a more precise coordinate system into the visualizations with ways to track a specific point (perhaps starting from polygons or GPS coordinates). It seems the visualization includes sea level rise but it's hard to figure out if its getting other details right like the California’s Transverse Ranges uplift (~20 mYa to present) which moved the Los Angeles basin several hundred kilometers northwest all while San Diego and Baja California were also rearranged.

Everyone in diagnostic medicine knows this (they've been dealing with this problem since the glucose finger prick test). That's why no biotech VC ever invested in Theranos and they had to get tech VCs and fancy board members like Henry Kissinger. A cursory glance at the proposal by someone qualified would have shut that down immediately, but tech VCs don't do due diligence anymore.

Brett Devereaux has talked about it multiple times on his blog in various different contexts (most recently two weeks ago)[1][2][3]:

[1] https://acoup.blog/2026/07/24/collections-pre-modern-armies-...

[2] https://acoup.blog/2022/07/01/collections-total-generalship-...

[3] https://acoup.blog/tag/cohesion/

Mozilla Firefox merged Servo's Stylo CSS engine into Firefox 57 in 2017 and their WebRender renderer into Firefox 67 in 2019. I guess you could call that a "pretty niche browser" even then but that's a stretch.

No one's rushing to pick it up because Servo was originally an experimental Rust project meant to incrementally replace parts of Firefox. Then Mozilla stopped funding its development, and only recently have people picked it back up again as an independent browser implementation.