dr_analog
razorboy
No bio...
User ID: 583

On top of that, most major organizations are still barely adapted to using spreadsheets and the most simple algorithmic techniques that were created decades ago. Literally just using excel to automate tasks could save these companies tens of millions of dollars a year. And yet... they don't?
You say this is because of political/irrational reasons but I think it's because the human capital available to them is too stupid to use a spreadsheet to (e.g.) predict when they need to restock the warehouse. Something intelligent agents should be very useful at!
Currently, you can either have Pete walk the warehouse every week and plan a re-order list based on his vibes, or you can buy an expensive business intelligence package, which is risky because it requires you to switch a ton of your existing workflow over. The smart simple solution a college grad in hard sciences with a spreadsheet can come up with is not available to most businesses because that grad doesn't want to work in an import-distribution business.
"satire" is what the Conspiracy calls truth when it’s afraid people might believe it!
Experience shows this doesn't work on cultists. They just move the world end date further in the future without updating anything else.
My favorite instance of this is the Church of the SubGenius prophesizing that the X-Day was July 5, 1998. When the day arrived but without aliens, they did a facepalm and said they had it upside down! X-Day is actually in the year 8661.
They really put every other cult to shame.
I have a nontechnical coworker, who has recently recreated some complex business intelligence tool we purchased not long ago using readily available AI and a little bit of coaching. He had an oh shit moment, when he realized how cannibalized the software industry is about to get.
So many contract renewal conversations with SaaS providers lately are like "they want how much per year for this service? that's three months salary for a typical engineer and our engineers say one of us can build this in-house in 6 weeks with LLM tools"
Thank you for the clarifications (and outright corrections!). I suppose the awkward fitting is to avoid stepping on the President's toes wrt his authority to declare an emergency, which seems to be widely upheld?
Big ups to you if you've heard of the US Court of International Trade before yesterday!
On April 22nd this court denied plaintiff's motion to issue a TRO against the Trump Liberation Day tariffs
The Court of International Trade on April 22 denied a group of five companies' application for a temporary restraining order against President Donald Trump's "reciprocal" tariffs imposed under the International Emergency Economic Powers Act. Judges Gary Katzmann, Timothy Reif and Jane Restani held that the companies "have not clearly shown a likelihood that immediate and irreparable harm would occur" before the court considers their motion for a preliminary injunction against the tariffs.
That seems crushing but then yesterday they issued a summary judgment declaring the tariffs as unlawful.
https://www.cit.uscourts.gov/sites/cit/files/25-66.pdf
(1) What. Also, this was months ago but I remember a ~10 point market swing and business leaders expressing general panic about when/how to invest in the US. Port deliveries ending. Truckers saying they have nothing to truck. One wonders what qualifies as "immediate and irreparable harm" in the US Court of International Trade if not that.
(2) At first I thought this was some administrative agency in the executive branch defecting against Trump, but this is apparently a real court? In the judicial branch? With judges appointed by multiple Presidents (though not Trump)? The SCOTUS might even uphold the ruling? Huh.
(3) The court finds the President overstepped the authority delegated to him by Congress under IEEPA by declaring a fake emergency and using that to impose broad peacetime tariffs, something it was never meant to authorize.
I read >0 amount of political analysis about this and nothing seemed to cover the US CIT striking this down as a possibility. Does the media just suck?
It was. I haven't tried 4 yet. Kind of anxious it might waste my time and send me a bill again. I'll report back if I get proper fucked and try it out of desperation.
At $100 USD, it's a no-brainer. At higher prices, well you might want to look for compounding pharmacies or grey market sources.
I've seen gray market sources selling the "constitute vials yourself at home" dosages for around $30/month. Presumably that's with healthy profit margin baked in due to legal risk.
Once these are out of patent I expect generics to approach $7/mo
My view of GLP-1s has a similar flavor.
me at 35: LSD is so consciousness altering! whoa!
me at 45: GLP-1s are so consciousness altering! whoa!
What I was expecting was a reduction in hunger, that makes sense. What I find really interesting is when I do have cravings, or feel hungry, the "seeking" behind what I do about it is gone. I feel a lot more activation energy is needed to pit stop at the bakery, or get a goodie out of the pantry, or even scoop myself up another helping.
I think this is the dopamine system being suppressed. But... only at the extremes?
I haven't become an adhedonic robot though. Sex is still excellent and when I do order something delicious and eat it I still enjoy it.
How does this make any sense?
in paragraph 3 I acknowledge/agree this won't work in online dating
They're useful, except when they're not. It would be amazing if they could actually say "I don't know" instead of hallucinating 200 lines of code that are wrong that had no chance of being right.
I've had Claude code waste an hour of my time before and bill me $20 in tokens.
No amount of game or self improvement will ever get you close to that if you lack the genetic basis for it. It's like thinking a 70 IQ man can become a world class physicist and win the Nobel prize if he just tried hard enough-- the world doesn't work that way.
I've certainly met people who aren't very attractive but some combination of attire and posture and personality gets them laid a lot.
It mostly seems illegible to me (and probably them). When it comes to discussion of game I imagine it's people trying to crack that code for themselves. It's probably not hopeless to try, if that's one's goal, even if they aren't the top 10% hottest in terms of profile pics.
Online dating is probably not your strong suit if you're not in the top 10%. You're probably a lot better off going out and getting drunk and hitting on girls in that case.
Homies: Ride or Die the 3d game.
Hacking away at stuff. Make is missing .hpp file changes, probably something about generating the .d files properly via clang.
Aside from that I added proper bounds checking to the racetrack by constructing a bunch of quadrilaterals to paper over it and simply iterating them and asking "is car inside quad?" and rejecting forward vector if it is not found in any. Can change this to a quadtree once I'm confident this is the right approach.
But weirdly the starting point where I spawn the car (700, 0, 0) doesn't fall inside of any quad even though I'm creating the quads procedurally and they should be sharing neighboring line segments. Wondering if this is some floating point precision issue creating some subtle gaps.
EDIT: I pasted the 1500 quadrilateral coordinates into ChatGPT and said why doesn't 700, 0, 0 lie inside of any of these and it responded that it's probably because it sits exactly on the edge of one line and my bounds checking is not inclusive of the actual boundary line itself. E.g. that sits exactly on the edge of one of the quads and my comparison function may be excluding that. Plausible!
EDIT2: yeah, that was it. I had this
bool Quad::is_point_inside_xz(const bx::Vec3& pos) const {
const auto sign = [](const bx::Vec3& a, const bx::Vec3& b,
const bx::Vec3& c) {
return (b.x - a.x) * (c.z - a.z) - (b.z - a.z) * (c.x - a.x);
};
bool b1 = sign(pos, p0, p1) >= 0;
bool b2 = sign(pos, p1, p2) >= 0;
bool b3 = sign(pos, p2, p3) >= 0;
bool b4 = sign(pos, p3, p0) >= 0;
bool result = (b1 == b2) && (b2 == b3) && (b3 == b4);
return result;
}
and ChatGPT o3 told me to try this. Basically using an epsilon value to check for float equality than exact because of precision, and also being more tolerant to point ordering, I'm not sure I'm definitely doing it clockwise in all cases.
This works!
bool Quad::is_point_inside_xz(const bx::Vec3& pos) const {
// 2-D cross product (signed area *2) in the X-Z plane
auto cross = [](const bx::Vec3& a, const bx::Vec3& b,
const bx::Vec3& c) -> float {
return (b.x - a.x) * (c.z - a.z) - (b.z - a.z) * (c.x - a.x);
};
constexpr float EPS = 1e-6f; // tolerance for “on the edge”
// Determine polygon orientation (+1 = CCW, −1 = CW)
// Any non-degenerate triple works; p0-p1-p2 is simplest.
float orient_sign = (cross(p0, p1, p2) >= 0.0f) ? 1.0f : -1.0f;
auto same_side = [orient_sign](float v) // interior ⇒ v*orient ≥ 0
{
return v * orient_sign >= -EPS; // accept 0 within EPS
};
// Point must be on the “same side” of every edge.
// Use edges p0→p1, p1→p2, p2→p3, p3→p0.
return same_side(cross(p0, p1, pos)) && same_side(cross(p1, p2, pos)) &&
same_side(cross(p2, p3, pos)) && same_side(cross(p3, p0, pos));
}
What am I even doing here. Feels like I'm one of those people playing chess still even though computers totally crush humans at it.
I mean, not really though. I had Claude Code write a lot of boilerplate for me w.r.t. get to triangle but it mostly produced buggy shader code and I had to sit down and think pretty hard a lot of the time. Even the model loading code was buggy and model files are pretty standard.
Helmet that I think looks better in matte black. Not final layout.
Tron bike helmet update.
Waiting for parts from Aliexpress still, but did a mock-up of LED strips on my bike helmet. Looked like crap, figured it was because the helmet was silver and black and draping LED strips over it had too much visual noise. Bought a matte black bike helmet and tried the strips and it looks significantly better, I think. Though I think this helmet will require more cuts to actually line the LED strips up along the solid parts and not cross the vents.
Helmet that I think looks wack attached: /images/17483641109905748.webp
Does there exist a class of research that is so speculative and/or long-term that it's beyond the quarterly target cycles of companies? Who would fund that kind of research if so?
If these people are real (they are) and they still read the NYT (they do), then the piece is understood as some uncomfortable nuance from an insider with a comfortable conclusion. That's not out of place in the NYT.
Yes that part fits like a glove. I still think it required (e.g.) more IQ points than the median NYT essay to follow though. But perhaps that's part of today's performance.
I considered the point of that statement as connecting political beliefs of students with their fashion choices and not something that professors influence compared to other cultural forces (including and especially social media).
This essay feels out of place in the NYT. Which is to say it's well argued, nuanced, a bit witty, requires more than twenty seconds of short-term memory and it advances claims that readers are not going to like. Also it's about 5x longer than usual. I am curious how many readers actually even get through it. The carrot and stick of the article (Harvard good but also bad but also good and Trump bad but also has a point but also bad) is potent but attention spans are so short and nobody is open to ideas.
Which is to say I think the article is excellent!
coding
They can be quite hit or miss. If you ask it something that's very well covered in its training data it can get you going fast. If your problem is tricky it's sometimes a huge waste of time to ask since it will confidently send you on a wild goose chase.
One failure of the craft of coding is that it's not very modular despite enormous work done on code reusability. People write very similar code over and over again, so it helps considerably to have a tool that can rip out boilerplate with your specializations added. Huge win here, especially if coding isn't your primary job.
I suppose I thought some integrations would conveniently fall out of meson for free. Like making it work with neovim LSP, and uh, being able to wrap distributed other libraries easily. But I still seem to have to use a Makefile on top of it to monkey patch things like downloading third party deps and building related tools, like for the shader compilation.
Being a young cranky old man in training does suggest I should move in the build.sh direction though...
EDIT: okay well I just asked ChatGPT to pluck the relevant bits of my meson.build out and incorporate them into my Makefile and it worked. and I can use a tool called "bear" to run "make build" and that tracks and outputs the compile_commands.json needed for neovim/LSP to work with clangd. this went pretty smoothly and I could delete icky Python tools like meson from my workflow now. yassss
Thanks for the inspiration @John_Doe_Fletcher !
Pic of game progress. Still mostly the same as last week but can't come here empty-handed.
I'm closer to having a prototype for a single helmet working. Right now I'm doing the grunt work of cutting WS2815 LED strips into segments, putting them into sleeves (because I accidentally bought one without sleeves, oops) and then wiring GND DO and 12V on the contacts of one to the GND DI and 12V on the contacts of the next. In some cases I have to use a coffee stirrer stick that my kids have been stealing from coffee shops to nudge the strips through the sleeve since they get so stuck.
I'm sort of getting the hang of the soldering crafting. I need to make absolutely sure I split the contacts in half evenly, which is tricky since they reflect on the cutting tool and the cutting tool is like, the size of the pads so it's hard to see around them[1]. Then I need to fill the contact with solder, then add solder to the exposed wire end, then kind of press the wire against the contact and put the solder iron tip on top until it kind of sinks into the pool of solder on the contact, then pull the iron away while still keeping the wire and pressure in place for a few more seconds. Then I can tug on it and it'll remain nice and together.
I can't quite find a way to do this while they're suspended on helping hands[2], I have to keep them all pressed against a silicone mat and weighted down with other random tools, which doesn't look very pro?
Anyway soldering to contact pads is resistant to tugs but not invincible. I put so much strain on one that a contact pad ripped out of the LED strip. I guess the problem is the 18AWG wire from the hardware store I bought is too stiff, and I should have bought silicone wire for that since it's more supple and bouncy and won't put pressure on them. Yet another thing where I order $7 worth on Amazon to keep things moving then order a big batch from aliexpress.com[3] for $4 to arrive whenever.
Man waiting for stuff from Aliexpress is torture. One batch is in the US as of 4 days ago but it's still waiting to clear US customs.
One thing I'm looking forward to with the kids' helmets is that I ordered ESP32-C3s[4], which are RISC-V based and also much smaller profile. They're like, a whole board with USB-C connector that's postage stamp sized whereas the ESP32 Xtensa I was using is more like, uh, the size of a bic lighter with 1.5x the width? It's kind of sexy how small you can make controllers. I don't know why I find this so impressive, the typical iPhone smartphone is a much bigger miracle of engineering but not as cool for some reason.
- I'm sure there's a perfect small precision cutting tool for $2 at Aliexpress
- Or maybe my helping hands are just not full featured enough. Maybe I should check Aliexpress for incredible deals
- Aliexpress!
- Surprisingly, the Aliexpress price for 3 was equivalent to 3 from Amazon. Huh?
--
Back to the Homies: Ride or Die 3d game, I got a track rendered now and have been doing some minor C++ refactoring since the last time I touched C++ was in 2014 and there's some new features IG. I'm using meson to build but thought I would have more compatibility with external libraries if I used cmake, but cmake is so much more baroque I'm backing off.
Now that I have a NASCAR style track rendered I can work on bounds checking, and add some sound effects of metal grinding if I bounce against the wall. ChatGPT suggests I use SDL_mixer.h, which is an easy lift since I'm already using SDL2.
My kid almost died of a tick two years ago. He woke up one morning paralyzed from the waist down, could not walk, just collapsed into a pile if he tried to stand. We took him to the ER and they were totally confused. Out of desperation they decided to admit him with GBS and would have put him on a ventilator in a day or two while he got worse and died.
While this was unfolding, wifey was talking with wife friends and one of them suggested we check him for ticks and we found one in his head, covered by his hair. It was a species of tick that secretes a neurotoxin to numb the host to its presence. It paralyzes and kills small kids, usually girls, because people can't ever spot it in their long hair in time.
We pulled the tick out and he was walking again the next morning, and 100% fine about two days later.
We had been camping about a week earlier. Our area is not supposed to have endemic tick diseases which is why the ER was so confused, though ChatGPT nailed it in the diagnostic when I presented his case to it.
We were already hyper wary about ticks when traveling to other parts of the country but did not expect them when camping locally. CDC guidance is apparently out of date.
Ticks have really increased our anxiety about going out into nature, something we really loved doing. Whenever we do go out mom makes the entire family strip naked at home and we do full crevice checks (including butt cracks, including adults) for ticks.
You have my sympathies, friend.
More options
Context Copy link