I just did this recently to reduce the noise I got for using 'superpower's skill for an internal library. Personally trying to find the right git ops balance of observability/transparancy of agent work while keeping human readablity. I haven't seen any accepted standards or workflows, its really interesting to watch dev culture evolve with this
I hate the rust v go wars, its not x vs y is 'best'. Rather is x better than y and by how much for xyz project done by ABC corp in this era?
As a lead I'd love to use rust, I will put in the time on my own, my team won't or can't. They treat this like any other job they signed up to deliver value with what they know. For hiring not everyone has the talent pool and fund access to get the goat-ed engineers that congregate to tech hubs for maximizing their income. Then if you get through that cherry on top is LLM's are only as smart as you guide it to be. There is probably a staggering amount of ways to write 1 approach to business logic, you may not know the ideal pattern so you'll commit to a worse one on the company dollar.
I'm moving my team's projects slowly to go because, its easy to go from novice to advanced in terms of code writing,legibility and patterns. We also don't have deep ecosystem requirements to ts/python in most of our work. It is verbose but I don't mind that on token spend if it gets done with with validation/error handling which it obnoxiously enforces. It runs cheap, ecosystem is good for platform eng, standard library does a ton out of box.
I've personally completely avoided the USA because of ICE, not because of principle of border policies purely from unpredictable treatment of people in custody with no recorse. Even if mistreatment are edge cases I've got no faith in the system to correct the wrong within itself. I've seen too many stories of legal americans and travelers not even getting bare minimum due process, or just being locked away for days no contact. I present and am read as a woman, if I get sent to male holding facility I'm just chum in the water. Which that decision could be made completely at the whim of the officer and the areas commanding officer, regardless of local laws.
>V-coding is the practice of assigning trans women placed in men's prisons to cells with aggressive cisgender male cellmates as both a reward and a means of placation for said cellmates, so as to maintain social control and to, as one inmate described it, "keep the violence rate down".[170] Trans women used in this manner are often raped daily; and this process has been described as so common that it is effectively "a central part of a trans woman's sentence"
America was the bad place, now it's the worse place. I appreciate the sentiment but 99% of the time people say this, I've noticed they do it to try to dismiss any current concerns with our current administration. I'm not saying that's your intention, it's just what I see.
Canada, I've turned down opritunities even in blue states over this being 100% honest. My Job used to involve myself traveling through the mostly the south. I'm lucky they are accomidating and shifted those responsiblties around but still pre-covid and a little 'post'-covid I was basically living month or so at a time in the south without issues.
Sampling bias. Most of those millions of people are white, cis, straight folks who present as the gender they were assigned at birth. The odds of something horrible happening are much higher if you don't check all of those boxes.
It's still rare, but so are most horrible things that can happen to a person. "There's a pretty low risk of you getting into a car accident, so why wear a seatbelt?" because the risk, while low-ish in absolute terms, is still unacceptable - so you do what you can to mitigate it.
I know this is going to be a hot take, but eating >100-120g protien for most people with even a moderate amount of exercisize isn't needed. Ratios passed around for how much protien you should eat are usually from studies on the most optimal amount for muscle growth of serious gym goers and atheltes. A normal person with moderate excersize would still see progress and be healthy at a 0.8 per pound of body weight. It would be faster with 1+g per pound but its not needed. My best cuts have been at 0.8g, and my best recomp was usually around 0.9g .
When I use a calculator, I atleast try to get with in a few digits of what I think the anwser is in my head. Mostly since when I was younger I had a very passionate teacher about how much slower everyone is now because of calculators on simple math. I just apply the same thing with LLMs, just try and think of how and what I would have said and see how close I was. Only thing I change is I don't trust the anwsers and accept some nuance in the given context. It's a double edge sword because then I crash out over it more than if I don't. When it over and under explaining the wrong sections or when it gets to an objectively terrible solution that technically anwsers the question. It feels like a student trying to get brownie points and/or give fluffed anwsers for the sake of not leaving anything blank on a test.
OT, but if we made kids learning math use log tables and slide rules for all their calculations I expect that they would engage their brains more and actually think about what they were doing, ie: form a strategy to solve a problem before they started calculating. Also I think that they would get a better "feel" for working numbers in general. I have no evidence, but I suspect that by abstracting away a lot of the "gruntwork" of calculating, we've really hampered people's development in math.
Unfortunately this adds quite a bit of overhead and would make everything take a lot more time. It might be worth it though.
This is really cool! After entering the software field, I felt uncomfortable not having a sense of relative size while working with digital storage units. Piece of mind was achieved after sitting down and mapping the units (Gb, Mb, etc) to common artifacts(audio file, plain text, database).
I've been planning forever to do something similar with length, duration, power, etc.
> Also I think that they would get a better "feel" for working numbers in general.
This is called "number sense". I'm pretty sure we do have evidence under searches for that term, it was well-enough known as a concept when I was in school decades ago and is the reason we don't use calculators when first learning math.
Having done a bit of teaching/grading for computer science, I don't know how to do it reliably nowadays.
For math we ask the whole reasoning and not just the answer, but here what do you do ?
What I resorted to is a live code review with the student, and it becomes very quickly apparent if they know the code, know what the AI said about the code, or didn't bother to read the code at all.
Yeah, I've always tried to train myself to do calculations in my head as much as reasonably possible when learning about mathematical objects, etc. Like when I was learning linear algebra I made myself invert 4x4 matrices in my head. (Pen and paper is also cheating!) Calculators and computers have been better than me at this sort of thing for my entire life, so in some sense this isn't a change?
We had Mental Math sessions in class. The goal was to teach you how to do math without pen/paper, calculators were not even an option. I try to teach some of this to my 6 y o.
> The goal was to teach you how to do math without pen/paper, calculators were not even an option. I try to teach some of this to my 6 y o.
It works quite well. I do the math lessons during bath-time daily with my 6 yo. He's up to the point were he can add multiply pretty much any number by 2, 3 or 4 as long as the product is under 3 digits.
Going from adding single digits to multiplication of random 2-digit numbers by 4 with lessons only during bath-time (no paper or whiteboard) gives a child a great deal of confidence with numbers.
It's great. Personally I am the worst at mental math and resort to a calculator for any stuff beyond 2 digits multiplications, but at least I know how to do it (very slowly) if I have to, and this means I could have an intuition of an erroneous answer a calculator could have given me (for simple stuff but you get my point).
Computer science teaching really needs to be prepared to do the same progression as math, but even harsher because LLMs are not even deterministic.
The calculator comparisons are truly meaningless imo, a calculator does nothing if you don’t know how to use it and what to input, an LLM circumvents all that, but a lot of people seem to think it’s the same.
It’s not a perfect analogy. I’m approaching it from perspective of a more universal experience of people losing aspects of a skill over time by using a tool. When ultimately the skill learned with out it allows them to use the tool much better
When I (and I assume others) use the calculator analogy, it's mainly about the teaching method.
You learn math without a calculator, then you learn how to do the boring parts with a calculator that you would still be able to do (slower) if needed.
It could be practical, internet speed will be a limitation for sure if it’s hosting a website not just apis. It really depends, also consider vps from hentzer auctions for splitting some services or even a refurbished server from 2016 era they are under valued for sure. How I like to think this through on how much everything is worth is measuring my (personal/organization) capabilities, capacity and quality of life for on going ops and in crisis. How big of an impact is it when things go wrong, how could I fix it and is there a chance it’s irreversible. Basically frame it around risk and your risk tolerance, I prefer to focus spending on ways to make my 2am crisis least stressful budgets allow.
At work I’ve moved us to a hybrid solution where we use cloud for managed database and storage solutions and compute on local servers. I personally prefer to have the budget be spent on the ops/maintenance of the scariest thing losing company data and somehow failing to maintain securing that data. Compute is almost easy to me in comparison, disaster strikes I just switch to another server local or in cloud, let it talk to the database, secure out going access, run my docker compose. More time spent on the write up than the fix. Someone that is a seasoned dba may choose the exact opposite.
I've built and audited medical billing systems and billing practices. It isn't an economics issue in a traditional sense. Infact it would drastically reduce complexity of these systems and payments over time, even allowing private insurers to exist but have to compete with a base general coverage. (many smarter people than me at princeton did economics showing this worked out as a net less expensive than what we are doing now) The biggest reason why its not the simplistic solution, is politics of all the middle men (me) making exponential returns from solutions to these systemic issues.
Too much money in the system being flawed, look at pricing for any HIPAA safe products and thats just technology. Money is so hard to get for healthcare providers it is its own industry of revenue cycle management and thrid party billers. Most of these physician lead practices charge more is because planning your account around reemburcement cycles from insurance companies are 30-120 days if your lucky is an advanced accounting problem. (Thats excluding complexities of audits, LOPs, network rates etc.) Medicare/medicaid the fraud side has lots of tiny wins through leaning on tax information more, taking the model from the successful basic income studies and trials worked out.
I'm not sure what you mean by pricing for "HIPAA safe products"? It's not particularly expensive to comply with the HIPAA Privacy Rule, and honestly that level of privacy and security is the minimum we ought to expect in any industry that deals with sensitive consumer data.
There aren't as many physician-led practices anymore. Most of them have been rolled up into larger health systems in order to achieve economies of scale and increase negotiating power with commercial health plans. Which is one of the factors driving up overall healthcare system costs.
The ops is crazy expensive, any service you get or buy into also instantly gets at least a 30% mark up. Look at the price of transport EDI clearinghouses vs healthcare! Yeah, it is stuff we should be doing for all data, but one mistake will leave you(business owner or contractor) with intergenerational debt or leave you criminal liable.
> There aren't as many physician-led practices anymore. Most of them have been rolled up into larger health systems in order to achieve economies of scale
There are plenty, in name at least many states have regulations to force this. Those larger roll ups are still clusters of smb doing 1-10m revenue monthly. They constantly fail and change hands behind the scenes, and are usually only protected by government regulations that reduce risk room to investors normally don’t have in other industries businesses. Their margins are like gambling almost and it’s a key driver of the service issues Americans face.
I think it is fair to dog on this for being a bit performative, like announcing you are blocking someone or quiting XYZ social media by posting on that social media. I do give some grace for trying to make an offline magazine/newsletter because I'd like to see more of that in the world, especially passion projects in general. All that to say, I think the attempt to be offline or more disconnected from some of these services is ultimately a healthy habit and is noble self improvement goal in itself. In the same way, trying and failing to quit cigarettes is ultimately healthier than not trying to quit smoking cigarettes. (I am biased; I think the way he formatted the post is quite aesthetic for what it is and would enjoy more analog to mellow out my life)
As the 'sarah' for an org I'm golden handcuffed to, avoid becoming the 'sarah' at all costs it will cost you so much in your career. I am a founding engineer self-taught from it/software security to full stack dev, and I pushed so we aimed to higher better engineers than myself. When they came in it was a full rewrite, a new abstraction and going down the same paths I learned the hard way not to go down. I was pushed out of enforcing hard-earned business logic, and we are still paying for it. Everywhere from not accepting the inherent complexity of the problem and over simplifying software to trying to get same features we had from 3 years ago. This then has made me the scapegoat for why the new engineers made xyz decision, and was the one in charge of fixing the shortcuts, bugs and workarounds. I have received promotions for this to be in charge of the veterans that didn't listen to me, whom still don't until it's a fire drill. Joke among colleges is I was the first 'agent' our company had, endless work, just enough authority to do current task, not enough respect/authority to solve the symptom. I understand this is also a failure of my office politics and am improving, however its hard to balance blunt productivity, slow careful positioning and letting people struggle with items I don't recommend until circle back.
This is such a toxic combination, becuse it requires significant people skills to get out of the "the kid who learned everything here and is grateful for it" and get proper respect as a professional. At some point the only option is changing jobs. I've seen companies matching your offer, finally realizing you actually have value in the market, but don't count on that, don't bluff.
I too ended up as the "go to guy", partly because I had a lot of enthusiamsm for my new job, and partly because the talent pool there wasn't very deep (or maybe they were smarter than me). It's fulfilling until it becomes unrewarding, I had to move on after almost 5 years. Still did consulting for them ocassionally for a couple more years.
That feels more familiar than I'd care to admit - definitely been there, done that. I think what most fail to realize is that it eventually turns into a ball and chain, decreasing your mobility within the organization. So the consequences of running an org on tribal knowledge and Sarahs is far worse, and direct, than that it won't work very well with agents.
It’s very easy to turn into the Sarah - or the Brent if you prefer the Phoenix Project analogy. As exciting as it might initially be to be the go-to person, it’s also, as you so elegantly put it, “endless work, just enough authority to do current task, not enough respect/authority to solve the symptom”.
Much appreciated! Thanks for the blog around ADR, I didn't know there was a full ecosystem for the approach and the AI tie in will help me sell the process
