trailrunner46

One reason to consider is around context usage with LLMs. Less is generally better and README.md files are often too much text some of which I don’t want in every context window.

I find AGENT.md and similar functioning files for LLMs in my projects contains concise and specific commands around feedback loops such as testing, build commands, etc. Yes these same commands might be in a README.md but often there is a lot more text that I don’t want in the context window being sent with every turn to the LLM.

Yes, trucks typically are running with the generator constantly on scene. Also many pumps are run on a PTO system where the transmission is put into a pump gear, further wearing on it since pumps can be run a lot on scene.

These numbers for trucks paired with the 3+ year wait times are very real. It hits small communities the hard because they have a small tax base but still need a certain amount of trucks. You can only consolidate so much before you are to far to respond.

Another good point called out in the article are the floating costs. The manufactures do in fact increase the costs after the fact so not only do you need to order a truck years ahead of time with a budget you don’t have (borrow money) but then you have to cough up an indeterminate amount of money years later. A real sad time for first responders.

There very well may be something more effective than water but the economics are probably a barrier here. Water has some pretty amazing at absorbing heat considering how abundant it is. Moving to anything else becomes pricey quickly in a serious quanitiy. Heck, even the foam used is quite expensive in much smaller quantities.

It's also worth keeping in mind that rural fire departments that all have tanker trucks typically carry only 1,500-3,000 gallons and need to continuously fill up from a pump site, drive to the scene, dump in a pond, and continue. Typical gasoline car fires will take less than 1,000 to extinguish. EVs are challenging from this perspective, and departments are just now starting to learn how to deal with them, it's a whole new ballgame.

Source: Firefighter in a rural dept

This was a good read, as someone using K8s a lot in the last two years but not service meshes yet, it gave me a lot to think about.

I understand there are various advantages like metrics, etc, but the encrypted traffic between pods and services is the one that I can see many orgs demanding. If not a service mesh what other options are there?

Might seem like and odd one to some but I became a volunteer firefighter and it has been very rewarding for many reasons but the connection with the community is a big one (most members grew up in the town and a central part of it).

100% agree, taking huge unhedged duration risk and then loading up on negatively convex MBS again unhedged is so crazy.

The chart titled “Impact of unrealized securities losses on capital ratios” really shows just how inadequate the tier 1 capital ratio is (what regulators use). Ignoring the impact of unrealized losses in assets marked as held to maturity is crazy. Seems like a regulator problem to me, no bank taking deposits should be able to make high duration and negatively convex (from high MBS holdings) without hedges.

I like the goal of reducing dependency on SaaS product. I use these Strava/Garmin apps a lot and while Strava was cool for a while it hasn’t progressed and doesn’t feel worth the money. Curious to hear what you had in mind, I am in the geospatial space and work with routing tools, etc.

Im not sure I agree with this. If you have a lot of extra taxable money just put it in a taxable investment account and keep some lowing amount in checking to handle bills. Keeping it all in checking and making the entire thing investments seems like an odd approach, the low interest you are getting in like <10k in your checking is not a big deal.

I can see how you came to wanting this but I think it could lead to dangers for many.

For most people (this is not financial advice for any one person) money in checking and savings should have a low rate of return and therefore low volatility because they need or may need that money to actually be there to pay bills or in times of crisis (emergency savings). Once you have these two pools of money, then you should invest in retirement and finally extra taxable investments. Most people should automate the money going into retirement and investments I agree but turning your entire checking account into a volatile/uninsured pool of money I think is the wrong direction.

There is a lot of skepticism around financial advisors and rightful so given the track record of many (picking insanely expensive funds, charging high fees themselves, etc.)

However, I do think there is a place for advisors not to pick investments but rather actually plan for life events (including tax stuff), which is right up the alley of this app). Having someone who could take all of your inputs, ask the right followup questions and appropriately model the likely range of outcomes on this kind of app could be quite attractive.

This is very slick and well executed. I especially appreciate the thought that went into data transfer (making it opt in, otherwise stays client side). I have a CFA and have been in finance for a while and have maintained a spreadsheet but honestly this does it better so I will use and support you. I wish more people would take a long view with their financial wellness, maybe I can convince a few friends to see the impact of their debt binges :)

On the UI side there are a few odd things like trying to paste a number that has decimal points makes it so nothing gets pasted in, the colors on some of the graphs are hard to read (especially for people with color vision deficiencies), etc. As I use it I will try to give more specific feedback. Can't wait to dig in further.

Python is certainly more popular and for job prospects I always tell that to newer data folks. That being said if you want to load in some data do some SQL like manipulation, run some stats and make a graph or output a report I would argue R is way better experience than Python but that’s much more about the package ecosystem and less a comment on the language. Dplyr is just more friendly to use than pandas (often 3-5 ways to do something and as a beginner this can be disorienting) and ggplot2 vs matlibplot. For interactive graphs you are probably going to use plotless anyway from both languages.

One other thing I would mention is knowing SQL well is the most translatable skill. A lot of dplyr and pandas are doing SQL like operations (in fact dbplyr will generate SQL equivalent commands for your dplyr code for various backends).

In summary know how to manipulate data in SQL then pick a language (because you will need to do some IO/reporting stuff outside just data work) where the ecosystem of packages feels user friendly to you and your work flow and roll with that.

This is exactly the kind of book I was looking for. I learn best by building projects. It’s not the syntax that is hard, it’s learning how experienced python devs do things (coming from other languages). Look forward to diving in. Thank you for your time writing it.

Cool idea. I played around with something using texting as the primary controller for users and I quickly found Twilio quite expensive. I stumbled upon SignalWire and it made the cost much more manageable for lots of small messages and for lots of individual numbers. Just wanted to mention it incase it is helpful to you!

If you like the Economist then you might like Bloomberg Business Week. I know I expected it to be bad but it’s actually wonderful. The design team is amazing. I find it the perfect mix of business/economy and pop culture. Often there are 2 or 3 long form articles on a truly fascinating subject.

The best thing I ever learned in sales is people must be comfortable to be successful. Find something relatable and that will often lead to comfort which will lead to a productive conversation/chance at a sale.

Makes total sense, I only used LiveView for a new project that I started within the last year, I can imagine it being tough to only use a little bit in an existing app, especially if its SPA hitting an api. Cool to see you evolved as Phoenix evolved.

Yes, Abinsthe is quite incredible! I have used it with react and vue and had some success. However lately whenever liveview is the right fit I find myself fist pumping all over the place. Something magically about writing most of your code in elixir and only sprinkling JS when needed. Your site is very ui heavy so I imagine it needs to have a lot of js strictly client side so a SPA makes sense (I just find the mental overhead of having to deal with a js framework front end and a elixir backend tricky).

Congrats on the launch. As a fellow elixir enthusiast I wanted to say nice work integrating the client side and pulling this together, you are doing a ton of stuff all at once. I don’t have a use case right now but still will poke around to see what you have built!

The closed end funds can be tricky if yields start to rise as the leverage they use gets more expensive. Frankly I am not much of a muni expert so hard for me to speak about munis specifically. In general though bond investors are worried about the binary default scenario, they are either getting paid over time or all the sudden not. However its important to keep in mind what kind of bond is being considered... an IG highly rated company has a large fixed income/duration component and then is at the mercy of the market pricing of the spread from government to corporate (widens when things look riskier, which means the bond price will fall). High Yield companies have less duration to worry about as a % of the return, its much more about the risk of the company (get paid for a while while the company is operating fine, but can get wiped out in a downturn).

Morningstar is just putting information out there, they have all sorts of language saying they are not giving you investment advice so there isnt much recourse there as far as I am aware (but I am far from a lawyer). The pensions may only invest in X are hard guidelines so certainly a manager would be in trouble if they violated that but that would be rare, much more likely a manager would just be holding more borderline IG paper and trying to look like they have better IG paper in aggregate.

Most of the AUM are in portfolios that are benchmarked to something (Barclays/Bloomberg US Aggregate for example). Although recently there has been a surge in unconstrained bond funds which have more latitude in terms of what they hold and with a looser benchmark. Managers start with a benchmark and then make 2 main choices. 1) how much should I deviate in terms of asset type exposure (benchmark is 40% IG corporate, should I do 50% and overweight?) and 2) within each asset type exposure which securities should I select (within my IG corporate sleeve should I choose the newly issues IBM bonds, old issue IBM bonds, Apple?, etc). Most managers are still pretty manual so the are doing deep securit research and are picking company A over company B but there is some movement to a more quant approach where you pick metrics and do a broad screen across all of those companies. The data across all of the bonds especially outside of just investment grade is hard to get access/clean so that has been a barrier to entry for a while. As the data becomes more available more managers will be using screens like has been commonly been done in equities.

While its true there are alternatives now, mutual funds do sometimes have a place (there are some pretty low cost ones that cover certain areas of the market where there isnt a low cost ETF to do the same). I would say pick an area of the market you want to invest in then compare across ways to access that market, mutual funds should be in that comparison still as often times it still makes sense.

As someone who has spent a lot of time in the Fixed Income world of finance I think it's great to see a paper where the researchers spend time and energy looking at the methodology a major player in the space such as morningstar is using. The easiest way to improve your returns in corporate credit is to buy lower rated (riskier) bonds as they are more equity (stock) like since you have a larger component of default risk baked into the security price. Over the long term the lower rated bonds will return more but will do so at higher risk. If you can convince morningstar which puts managers into discreet buckets to put your fund in a bucket that is safer than your actual bonds reflect, you now appear to have more return than your peers (although in reality its just because you have more risk).

The lack of accurate risk (reflected by the average bond ratings of your holdings) is what is really at the core of this argument. The researchers joined together some pretty commonly used datasets in the industry (probably what I would use if I were to do this) and impressively were able to properly take the holdings of managers and come up with a proper risk picture (if you believe that rating agency ratings of bonds reflects the true risk but that for another post). Morningstar basically said that they have a crappy dataset which just doesn't have rating data for many bonds and therefor, when they don't have a value, they just fill in with a default. This makes me think that Morningstar is doing a pretty lazy job in their evaluation of managers (what incentive do they really have, they are a monopoly in this area).

Happy to answer any questions people have.