micro_cam

People here also don't understand machine shops.

My dad ran a job shop focused on small jobs and the economics are different.

A lot of his work was keeping other local shops / industrial equipment up and running. So there is a lot of variety of work but very low throughput and kind of by deffintion you have the capabilities to fix your own machines.

Programing a CNC machine makes it east to make a lot of the same part but if you only need one it may be quicker to just knock it out manually.

A 50 year old mill or lathe is easy to keep up and running, can be upgraded with a Digital readout or even CNC controls if desired. A tool in a shop like this likely won't see the cycles one on a factory floor constant uses sees but may be worth keeping around since it offers a unique capability...he had a large ww ii surplus lathe for jobs that wouldn't fit on the smaller more modern machines for example.

I've eaten at Dave's and also gather and eat morel's regularly in Montana.

Dave's is a popular joint but didn't strike me as particularly good even by Sushi restaurants in Montana standards. They had some creative rolls but seemed to fail to get the basic fish cuts and rice seasoning right.

What they were doing marinating the morel's and eating them basically raw is ridiculous. No one who is experienced with morel's does it that way and most mushroom guides contain info on hydrazine contents/risk and suggested cooking methods.

March and April is too early for Montana (or most US) Morel's which are commercially picked in forest fires from the previous year once the soil temperature reaches 54 degrees. I heard at one point these were grown in China and wouldn't be shocked if they are a different [sub]species that might contain more hyrazine. There are also some "false morels" that fruit earlier (one species is known as the snow morel as it comes up right after snow melt) but these are easy to identify contain even more hydrazine.

The g7 is definitely a bit better on all counts except maybe bluetooth reliability. I use them with [open source closed loop](https://loopkit.github.io/loopdocs/) insulin pump software and it is pretty great.

Still a lot of plastic but the sensor comes in a smaller jar/applicator in a cardboard box and it pretty reliably lasts 10 days cutting down waste.

It sounds like this new sensor is just a g7 but with less stringent accuracy standards for non insulin users allowing it to last 15 days.

I rode mountain and road/commuters with rim brakes for years and I would give up gears and suspension before disc brakes and dropper seat posts.

I actually find hydraulic brakes much easier to maintain. They self adjust to account for wear and don’t suffer from cable stretch so I rarely need to adjust them. They need a bleed maybe once every few years but that is actually dead simple for shimano anyways.

The outdoor industry as a whole had a covid boom and upped production/orders. Now it seems a lot of those sales were in a way borrowed from the future and they have a ton of inventory and lagging sales.

I implemented something like this in a [pre xgboost boosting framework](https://github.com/ryanbressler/CloudForest) ~10 years ago and it worked well.

It isn't even that much of a speed hit using the classical sorting CART implementation. However xgboost and ligthgbm use histogram based approximate sorting which might be harder to adapt in a performant way. And certainly the code will be a lot messier.

I've got a ~10 year old implementation that does something similar calling it "three way splitting" here: https://github.com/ryanbressler/CloudForest

And i got the idea from a lab mate, Timo Erkkila's RF-ACE project though neither of us thought it was a particularly novel idea.

Is anyone seriously proposing removing dams on the Colombia or Snake? Or really any dam used for irrigation?

Most of the removals seem to be on very small scale hydro projects on coastal river that produce very little power, aren’t used for irrigation and open up a lot of salmon habitat.

Does anyone make a quality standalone pps with a heat recovery ventilator? I’d buy one I could retrofit in without existing ducting in a second.

Gpus were built for dense math and they ran with it. To the point current best architectures are in part just the ones that run best using the subset of linear algebra gpus are really good at.

There has been a lot of work on sparsity and discovering sparse subnetworks in trained dense networks. And intel even proposed some alternative cpu friendly architectures and torch/tf and gpus are starting to do okay with sparse matrixes so thing are changing.

My step mom did this in the 70s or 80s. She said it was the first time she heard primal screams and that you could just pick up high end outdoor gear along the trail as people would discard it to save weight.

There are tons of successful long term non religious cooperative living arrangements of various sorts but yeah most not fully isolated substance farming because that is hard.

Amish communities also aren't self sufficient though. They'll happily sell and buy goods with outsiders and even sell amish furniture via websites (not sure if they use an intermediate). A ton shop at the Costco near me.

Propane is a bit more expensive then traditional electric heat depending on the year.

It used to be much cheaper so a lot of old houses that aren't near a natural gas line were built with propane heat. You might save $200-1000 a year by switching to a heatpump but that means 10+ years to offset a ~$10000 install cost for a large unit professionally installed with a new 220 circuit run for it etc so most people just keep paying for propane or maybe use small electric space heaters.

And if you are somewhere you need to worry about pipes freezing in power outage a while you are away or temps bellow the heatpumps minimum -15f range a propane stove is a nice option. Relatively cheap to install and they can be setup to run on a thermostat with no grid power.

If i was building a new house i'd go heatpump + a propane stove for back up heat and a dual fuel induction/propane range.

It’s not bad where I am now in rural Montana but where I grew up near Seattle the power was off for a week plus most winters.

Lots of above ground power lines in heavily forested suburban sprawl means lots of branches in lines every wind storm and they (private) electric co has to visually check all of them before they can turn power back on.

Despite being much more rural the power coop that serves me now seems to never have an outage that lasts more then a few hours. Probally mostly because they have much less pine to check.

It needs good air flow to work well and the noise is only really a problem inside the house in the room next to it. I may put some sound insulation on the wall at some point.

We are running a large single head Mitsubishi ductless h2i in Montana. It easily provides most of our heating needs down to about -5F and some heat down to -15F, we have our old propane boiler / in floor system that kicks on when its really cold.

And it provided ac in the summer which we didn't have previously.

There are some downsides, it was expensive to put in and the outdoor unit is loud. While it does save us money, it probably wouldn't make sense as something that will pay for itself in heating costs with current propane prices and natural gas is much cheaper if you can get it. But definitely makes sense for new construction especially when paired with a wood or propane stove for back up heat in power outages/ambiance/quick heating.

Does this parse in common he software?

Just stick it in a multi page doc or pdf with minimal formatting. Resumes don’t need to be one page.

Most of the resumes I see for candidates that make it to the interview stage are multi page with very simple formatting. Lots of keywords.

Not super common but its reasonably frequent for someone to go to industry for 2-3 years and then go for a masters/phd.

Those candidates tend to be great as they are up on industry practices/technology like source control and databases where as some phds can have only academic coding experience. And they tend to have studied something they really knew they were interested in and really been the driver on their thesis project vs just contributing to their advisors research.

You also see great phds who didn't wait but really took ownership of their project, used source control, figured out distributed computing, contributed to open source, scraped/built their own datasets, understood the real world implications and hacked on side projects to develop coding skills.

And you see some who just completed a theoretical + computational project their advisor suggested on an existing dataset with the minimal amount of coding needed and little thought to implications/applications.

I've been in the ML/data space for 20 years and went through the 2008 cycle early in my career. I'm not hiring at the moment but am always happy to review resumes and give career advice as part of networking so feel free to ping me.

Reaching out to managers at big cos won't get you much. They are getting a ton of twitter/etc candidates and have stricter hiring policies and rubrics to prevent nepotism and favoritism. If you can network and ask for referrals through your friends / professors that can work better. Or reaching out to early stage startup cofounders can work really well.

ML/AI is less frozen then some other areas so that is good. Most really competitive new PHDs will have either a couple of internships, strong academic contributions or previous engineering experience demonstrating strong coding ability.

You should also consider post docs or academic engineer postings. The grant cycle insulates these a bit from the economic cycle and they can be a good place to gather some experience while you ride out the cycle.

And definitely consider very early stage start ups. A startup that just raised and has 2 years of runway is probably one of the safest places to be at the moment as they are still focused on growth. A lot of great companies proved themselves as startups during the 2008 cycle and grew rapidly after. Networking can mean a lot more here as early stage founders often literally just hire their friends or people they get along with without a ton of process.

TikTok’s parent co, bytedance, has multiple social video products in China. They have been genuinely investing in being good at serving and recommending video for a while.

If you search arxiv.org for bytedance you can see some of the work they have published. I’m sure there is lots more they haven’t.

I suspect a lot of learnings from other products went into the product design of tik tok as well in terms of how to balance exploiting a users knon interests with showing them novel stuff, how to give anything the chance to go viral etc.

Montana. Good to know there are other similar states though

In my state we have a lot of breweries that essentially just operate as family friendly restaurants with bars. Like towns of a few thousand people have multiple breweries.

It turns out liquor license laws make it much cheaper to open a tasting room that serves food then a restaurant that serves beer. Some can or bottle their beer but the business model can work fine without major distribution. So people open breweries instead of pizza parlors or steak houses.

A standard heat pump water heater does that but will only provide a couple degrees of cooling to meet standard domestic hot water needs. You can also get pool heater heads for heat pumps/ac units though last time I looked not all were available in the us.

From an employee point of view, YC companies have a weird correlated risk due to investor sentiment ... if YC and other vc's are nervous about the economy an email goes out that says "if you don't have two years of runway do what you can to get it" and a bunch of companies will start rescinding offers and/or doing layoffs.

Investor and CEO sentiment has been a roller coaster the last couple of years and very few ceo's have a steady hand on the tiller. I've seen companies go from hiring like crazy to laying off then back to hiring like crazy in way to short of a time period.

If interviewing right now I would flat out ask about runway, burn rates and hiring plans and be looking for a company that raised at the right time, has a decent amount of cash and has a pathway to being profitable. And make sure the founders seem to have a steady hand on the tiller.

I was an early employee at my last co and we figure out pretty fast the trial work / lengthy process thing doesn't scale at all. YC seems to push this (do things that don't scale) and it might make sense for your first couple of employees (when there is high risk of a bad hire and you suck at interviewing and are scared to fire people). But even short ~1 hour take home projects filter out too many good good candidates. Get good and comfortable at making a decision based on a normal day long interview and get good at firing people who aren't working out to mitigate risk.

No one is talking about farmers at all. Farmers grow crops intensively on their own land.

Ranching is a very different activity which in the west usually involves a cheap grazing lease on thousands of acres of public land... it is the governments land and they balance the concerns of all stakeholders.

It's much more than that even. Without the wolves the elk and deer graze along creeks destroying willows that shade the water leading to warmer water temps and fewer fish, more erosion and less water retention / flood control. This even impacts downstream farmers who take water out for irrigation.

Also the wolves keep mountain lions, coyotes and other predators in check (through competition and even predation) and wary. I suspect mountain lion encounters in California will go way down if the wolves ever come back.

Welfare ranchers are ranchers who rely on cheap/subsidized grazing leases on federal land. It includes a shocking number of billionaires making big money off public land.

As someone who hires a lot of ml engineers a masters can be a good way to learn but isn't a strong signal on a resume.

Hands on experience and side projects (especially ones in investing/betting where you actually put money on the line) count for way more.

A good masters will help you nail down the fundamentals of linear regression, metrics, regularization, gradient descent, metrics etc. That knowledge doesn't go out of date and you do use it and get asked about it in interviews.

Large language models etc are growing rapidly but how most practitioners user them is much slower. People still use BERT and the smaller ones for computational reasons and ways you fine tune, evaluate, prompt engineer etc for the larger ones don't change that quickly.

There is also the large area of debugging, monitoring, performance tuning and improving large training and production systems. Even if generative models write could write great code for an entire system i don't see this area going away any time soon.

This was true 10 years ago and still is for many jackets. but the latest generation of wpb fabrics including gore tex pro and polar tech neoshell actually breath really well, like you can get a shell you can bike up hill in and not sweat out now. It’s crazy.