anarticle

I'm still surprised we don't have many physical interventions for this, knowing that it's a physical problem. I'm reminded of the older surgeon style medicine whose reputation seems to be pretty coarse now since we have so much knowledge about chemistry and non-invasive interventions.

Wonder what would happen if we pushed the envelope on physical fixes like endarterectomy like procedures and made them safer/more routine. Like getting your teeth flossed.

Maybe, I always say "let's write a plan and refine it" which ends up being a nice rubber duck session to keep claude and friends from going off the rails implementing things like privacy or encryption or some other rabbit hole when I just need it to do a few features at a time so I can review while it starts the next few things. Maybe this is slower to deliver, but I don't spend a lot of time unwinding commits.

I don't necessarily use plan mode explicitly for this, but I might turn it on if I think something destructive is coming up, like a schema change or some other devops thing where claude may decide "oh we copied it all over so I can delete it from source" "oh I'm sorry I didn't catch that error code and assumed success, your data is now gone" lol. This happens a lot less but I've been bit right on the face by it before.

Are plans the wrong abstraction? You can use a big enough plan to spam tickets into linear and have claude burn them down. I find it increases subagent accuracy.

Isometric holds. Tendons grow slow and recover slow, generally not as much circulation as muscle.

Big heavy holds will help encourage blood flow without stretching them / stressing them as hard as big lifts.

It seems weird because it goes again ROM ideas, but it has definitely helped protect my arms for rock climbing as well as less fatigue in my legs for big lifts.

Tools vs services in my mind. There is no guarantee any provider will continue to do what they are doing for you at the price they are doing it.

The object permanence of not having to reinvent the world every time a model gets sunsetted has value.

Really wish they’d host more models for us normies. My guess is OAI will buy / subsidize them with terms that will close off open models.

Turns out we ended up with paid search, but paid for by who? :D

Seems bad.

I think the catch is there's no money to be made in mirror molecules / life... YET! If that changes, say mirror molecules some how extend life or have a therapeutic use then we're like to see the tune change.

Drives are cheap, data is expensive.

I work in several labs where I have to give this advice, which some labs are smart enough to make an ersatz backup that we store somewhere (PI's house) as a cheap offsite. Even a partial recovery is great in the face of fire/flood/explosion, or the most common error: IT fucks it up and then says "oh well".

Hope someone has some copies and they can rebuild, what a nightmare.

hahah this is all well and good until the 20k+ line a day person is considered the good engineer, based on kpis alone. Your VPs aren't reading any of that.

Bonne chance fired senior!

I have depicted you as the 100 line a day weakling, and myself as the 20k+ line chad.

The systems at bigcos are too ruthless to risk gambling your career on this kind of reasoning which you have no recourse to. Best to start small teams and destroy them from the outside.

I feel like we've probably spent more for less! Any interaction with the medical industry for less than 1k feels like a steal to me.

I always tune into: https://kohina.com

Thanks for sharing, I'll def put it on.

Not really sure what you're getting at there, I was a hired gun for about a year, founded a company and am having a great time.

Last time I was at a corporate gig, I got a 2+ year old base model MacBook Pro to do a data science. 16gb of ram is fun and all, but when you're using R you're going to have a bad time.

During my hired gun period I worked with two organizations that used gitlab, and they're doing just fine. One place I had no problem using their cloud for bigger runs, the other had a permission system that was pissing me off so I solved my own problem and billed them enough it didn't matter anyhow. Like I said, having extra cpus for big R runs you don't have to ask anyone for has a value all its own.

Happy computing!

I added sounds: https://github.com/gpurkins/waiting-for-claudot which has become a bit pavlovian. This along with notifications in iTerm is good enough to juggle some claude sessions.

I would have guessed nutrition, we live an in age of vitamins and fortified foods. You can get a lot of zinc and other metals from clams and oysters.

Is it? When you're inside the fence you get some IT cast off from the last guy that was cleaned up and re-spywared from 2-3y ago (sometimes worse!). As a hired gun (data science/engineer), I have an m3 128gb 1tb all sliders to the right MacBook Pro that works great, no spyware, rocket fast, trains/runs small models. With Apple 0%, paid in a year. I generally can name my price so it pays for itself.

If you're riding it every day, every little speed up counts. I don't wait on GitHub actions or external systems for my crunching. For that I have my local slurm cluster of mini PCs, 64 cpus and counting!

I assure you there is a better world outside of the fence, I am unlikely to return.

Pros buy their own tools. This is why working for yourself is better than working for a corpo, you get to choose your weapon.

Not your problem under the hot potato model. It's not impossible and here's the other thing: it often doesn't matter if things get broken to your megacorp as long as you keep up appearances with clients.

Sorry if this sounds really grim / cynical, I've simply seen enough of these kinds of migrations to know that it is fundamentally opposite of my perception of engineering philosophy. It often becomes more of a question of business rather than correctness. (Can we simply fire the smaller customers? -> yes.)

How is the payvider model better for customers?

I see why it can reduce costs in a runaway cartel based system, but how does it prevent the two wolf scenario? (It's the same wolf in this scheme.)

I think in the case of Optum + UnitedHealthCare being the Scylla and Charybdis of a healthcare situation, we should break up this style of business. Owning both sides of the equation means there is no competition if you are unlucky enough to find this combination.

Feels like two wolves negotiating on how much of the sheep (the sheep is you) they get to eat.

Dare I ask, who is for the "consumer"? If we should even use those words in this system, which in my mind should be for a nation keeping its citizens alive and well both of their own sake and the state's sake.

It's about offloading blame. If a server nukes, it's on infra to get a guy to unscrew it. If a service nukes, infra guy says "welp it's down", keeps on clicking.

It doesn't matter what happens 6m-2y down the road, your odds of being laid off or job hopping are high in the current regime so this all makes sense. You pay some amount of your budget to make your life "easier" in the now.

The trouble comes 2-5y down the line when the service is bought out by <insert MEGACORP here>, and you have to scramble to replace it or hold your nose and pay up.

(tbh, migration is not that hard, but the org will act like it is)

The matrix of authentications, compliances, and intranets will only go up as your company grows and often are enforced by people who do not suffer them daily.

Hi, sorry, this is so disingenuous of a statement I cannot pass by it without commenting. My bona fides are 10y of lab work, specifically in bioenergetics. I can tell you that 5% is a dramatic UNDERESTIMATION on the value of animal models for medicine at large.

This is ignoring at least these benefits: surgery, development, genetic studies, grafts, anesthesia, and many MANY more. Some non-drug related, some drug adjacent, and they definitely have downstream benefits to humans.

Here's a survey paper with myriad examples: https://pmc.ncbi.nlm.nih.gov/articles/PMC9247923/

I really don't like when bioscience articles land here in HN because they are always commented on with:"in mice", as if to say nothing we see from mouse work works. Well, not everything is software and this kind of work takes years, if not decades. It is real science unfortunately which means that most of it doesn't work! Science, and bioscience specifically, are not efficient systems. In general, the things you do are hard and probably won't work. That doesn't mean you give up.

Animal models are not great, but they are the best progression we can do right now from cell models. And as for being disposable, there are controls on how animals are used in labs in the US: every institution that has animal experimentation has an IACUC (institutional animal care and use committee) that every research proposal must go through, and they do not a rubber stamp your proposal. They want to know why you can't use cell models, and why you can't do it with less or even no animals.

It would be nice if people were a bit more even handed when these types of articles come by. I think HN can do better.

An adage from the lab: "If what we did always worked it would be business, not science."

Software is a tool to solve a problem, as long as you keep finding problems that you can solve with it, you're likely to get paid to do it.

If your crowning achievement is: "I can 100% all leetcode hards" I have bad news for you.

I had something similar to this happen to me, where some kid was causing trouble during nap time in kindergarten.

I was an obsessively good kid, my parents took me everywhere with them and treated me like a peer, within reason. I was well behaved for my age. At the end of the day in kindergarten class, if you didn't cause problems, you received a stamp on your hand. The stamp was everything. A brand that I had ACCOMPLISHED that day.

Nap time was a post lunch, thirty minute time when we turned the lights out and laid down. Some kid near to me was making faces and making weird noises behind the teacher's back during nap time. Of course, he's five, maybe six, so this is not going undetected by our teacher. She storms over and asks "who is making all this noise?". I, being a total narc at 5 simply point. Assuming of course, this means I will receive a daily stamp, maybe even more, for my quick and wonderful detective work.

Then the unthinkable happens. His name goes on the board. MY name goes on the board. A wave of confusion sweeps over me. This is a massive blow to my tiny ego, only bad kids get their name on the board, surely there is a mistake!

It's nap time. I cannot make any noise, else I will risk A CHECKMARK NEXT TO MY NAME, which will only escalate the punishment in 198x to TIME OUT. Bad kids are always in time out. I am NOT a bad kid.

I am crushed. My small brain cannot process the enormity of what has happened. My name is on the board. I am smart enough to know what's not coming.

2pm comes, we're sitting on the square rug, and we're all putting our hands on our heads to receive our daily benediction: the stamp. I desperately keep my hand on my head, hoping I might trick our assistant teacher into giving me what I know is very far away.

She passes right by. I look left and right and realize, there is no mistake.

I held immediately held back a flood of tears, feeling deep failure. I stood up, and slowly gathered my things. I slogged my way to the bus and remember staring out the window thinking, what if the same thing happens tomorrow? I will never receive another stamp under this system, how could they do this to me?! The stamp continued the next day, but a different mark was made.

I had a short villain era after this, realizing a true injustice of the world: no matter how good you are sometimes things will not go your way.

With dynamic pricing and other consumer predatory schemes out there, Costco feels like the only one that fights for the consumer. The checkout is lightyears faster than any local supermarket, return policies are good, and I don't feel like the warehouse is trying to waste my time. The low price of good quality meat alone is worth the price of admission. If you cook, you come out doubly saving money.

"Something about the whole thing always registered to me as, like, lame—too normcore, too boring, perhaps even too cheugy to an informed and taste-driven millennial ur-consumer like me." -> What even is this? Get over yourself.

Remember that the CEO of Costco wears his name tag to work, and eats the Costco hotdog like everyone else. I'd buy that for a dollar!

That's the edge of what's possible, it's quite common even for researchers to have problems replicating results at the edge.

There's sometime implicit knowledge in a technique that either doesn't get written down, or someone is so good at something you don't think certain details will matter.

In my old lab (biochemistry) some people just have good hands and are really good at making something repeatable, others not so much.

Nice. Bye.

A reason for the thick air of paranoia is that now, everyone knows someone that has been laid off. Simply so many that it is starting to hit home. Estimates are near 2008, and if you lived through that you know that help is not coming on a timescale that you could have to massively change your life.

You lose your job, two years go by, time to sell you your house and move. Hiring is a total circus right now as well, being subjected to a five course hiring obstacle course is a lot of time that you're burning your savings and or missing other opportunities. Compare this to nearly any time since 2012 when it was at most three, and maybe ONE was a technical.

Most people do not save in America, and even when you are employed the health care system does not take great care of you. All of this "choice" is presented as capitalism working, but really it's a set of land mines where two large entities decide how much they want to take from you (the hospital, and the insurance company). Since the pricing is opaque and the amount the insurance company pays is capricious, vaya con dios.

The line feels like don't get sick, and your own country has thrown you to the wolves (they're in on it). Similar to unemployment, and the other "safety nets" not managed centrally or well. Massive delays, and your mortgage is due.

Also, you are paying for all of these safety nets all the time when you are making money, but it is deeply gated when you need it. Sorry for the paragraphs, but watching a friend go through this now and it's very wild.

If you're able to save more than 10%/m, you are very ahead of the game.

As for USA losing the Mandate of Heaven, even people from other countries seem sad to see it happening. Informally, two different groups of Portuguese people I've talked to in the last two weeks in Lisbon had a sentiment of "how could this have happened to such a great country?" Mostly due to the extreme news reports coming from the US, ICE, war, rhetoric etc.

local models mean it never leaves the fence, I'd much rather do that.

Programming Windows is THE authoritative source on Win32 programming: https://www.charlespetzold.com/books.html

It is a fantastic book, I learned everything I know on Win32 from it. Wrote real time scientific software in windows for ~10y! We did it all, external hardware control, custom UIs, etc. Thanks Ryan Geiss for your timing info.

Right about VC6 was the sweet spot imo, C/C++ with lightning fast UI for docs and more. Tools got out of the way. Once other languages got involved (C#?) the docs got out of control and harder to use, and the UI started to get a little overloaded.

The snappiness of those old windows systems was pretty great.

This also feels like "we're about to build a bunch of datacenters and we cannot meaningfully verify the quality of the concrete at our sites." This enables them to monitor the variables and probably not pay if it's out of spec would be my guess.

In my concrete mixer experience that's just one part of the process, lot of other crap goes wrong, forms, vibing, water, additives. I'm not pouring foundations, so my xp is only to say there's a lot going on. Guess it's a first step?