drudolph914

15 日前議論

I don't think the majority of ppl are giving it hate tbh - I think it's just a HN thing. everyone I know that is into gaming is excited about it. I would chalk it up to consumer misunderstanding around the pricing, but I actually think most consumers that are interested in buying a steam HW product are fully in the know why there's a chip shortage. I am in the group that is waiting for the price to drop, but I am excited to get mine!

drudolph914

16 日前議論

I appreciate ya commenting, and let me tell ya directly that the work is great! I'm an old-head and curmudgeon - my thoughts here are about the world at large. your work here is great, please keep it up!

drudolph914

16 日前議論

yeah, I agree with you. I guess the part that feels disingenuous is that these kinds of projects are some-what cashing in on attention from people who still who see threejs slop as "high-effort UIs." this speaks more about me than other ppl, I'm just personally already at the point where all of this feels closer to slop rather than quality. I guess as this trend plays out, we'll all eventually feel somewhat similar on the yet-another-3d-UI-slop trend

drudolph914

16 日前議論

big fan of oxide, and love the demo overall, but at the same time, I can't shake the feeling that these kinds of 3D demos are a bit gimmicky/cheap nowadays. pages like this used to be a signal for a high-end product and or a benchmark for good engineering. now though, we all know that this kind of work can be vibe-coded with threejs fairly quickly if you have the assets. idk ... it feels like it's trying to capture my attention through flashing lights instead of letting the work stand on its own

I don't want to subtract from the demo too much, b/c I do love oxide, but I do see this as a trend that more people will use to garner attention until it's too overdone - at which point, 3D will revert to being used for more practical use-cases

EDIT: typos

drudolph914

16 日前議論

I feel like this comment is just engagement farming, but I'll bite anyways

there is a larger appetite for something like open source AI mostly b/c of price. we all know these labs have not figured out their pricing model, and we're all holding our breath out of fear of what the prices could be.

also, if you consider that the only toll to knowledge work before was personal time, and now you need to pay $100s month just to keep up with the baseline speed. it makes sense people are looking for something that gets them back to a workflow where the price to do work is near $0.00.

I think for a smaller group though, it's more to do with a certain combination of principles. Some people don't want censorship, other's want ownership, some want the knowledge of working on LLMs to not be gate kept.

drudolph914

19 日前議論

GLM 5.2 is the first time I'm actually excited about AI! I'm not the most bullish on AI code for several few reasons, but the biggest reason is the ownership model. We all know we're near the tail end of the "subsidized pricing" window for AI, and I've been hoping for so long to get an open weight model that is _close enough_ to the SOTA before this window closes - and we actually got it! I'm excited to be able to in the near future run GLM locally, and use these things like a tool instead of living in this for-rent model for the rest of my life. I'm excited to actually enjoy programming again

drudolph914

23 日前議論

I will say, a positive thing that has come out of msft's 20ish year run of consistent incompetence and piss poor leadership, is that there are quite a few former msft engineers (now retired) that are posting great lectures and educational content on youtube.

also, idk when, but the talent level of a "msft engineer" from 90s to early 2000s feels like they runs laps around the msft engineers of today. it's hard to not feel that the suits cannibalized what was at one point an extremely profitable company with great engineering culture for nothing but shortsighted gains

drudolph914

25 日前議論

so much to unpack here and almost poetic that you say this

first is that the model will write out that it “thought” and “double checked” it’s output

Second, this was in a fresh context window of the latest model (that isn’t fable b/c we can’t use for reasons beyond this thread), and it was on it’s second highest thinking mode. I shouldn’t have to double check something that it claimed to have burned more tokens on to double check

Outside of it costing me more money to fix what it claims to do, the main point of this article is that models are implementing things nearly end to end, and if we scale it up, it will only continue to do that. I Intentionally chose the example of something that is < 70 lines to implement in TS (btw, the language with the second most amount of data available to scrape and train on) I would assume a machine that can almost implement things end to end should be able to implement something of 70 lines of code and has been documented for nearly 50 years.

My point is that time and time again on the most trivial examples, under the best of conditions, and with unlimited amounts of money, they can’t do what it claims

Outside of that, this follow up comment(s) that say, “oh you need to ask it to check its own work and be so involved in the process of it writing the code that you need to spot check it” goes against everything the article states

The best analogy I have for this is New speak in 1984, it’s just vibes dictating vibes and trying to make people claim that the vibes are right. and if you try to validate the vibes, your vibes are just wrong because you don’t get the vibes. The claims that it made have no data backing it. And if there is data, it’s cherry picked. Please use your brain and stop outsourcing your ability to think to a machine that is incorrectly thinking on your behalf

Edit: Typos

drudolph914

25 日前議論

Yeah, I keep coming back to a point that the way people talk about AI is still entirely disconnected from what it can actually do. I think of the bell curve meme a lot when I see people talking about AI. the people most bullish to perpetuate that it's going to take over are people that have vested interested, or people that are fall on the bottom half of the bell curve. I mean ... come on, by design an AI is literally a statistical averaging of all the data it's seen. AI is extremely average at nearly everything it does. If you find yourself using AI and it's doing something amazing, that speaks more to your knowledge/ability about a subject more than it speaks about AIs ability

I mean, I guess if all you do is work on implementing CRUD endpoints ... sure I guess you're cooked. but we had tech to automate this already, this isn't anything new. But oh man, if you're doing real engineering, the tools are barely usable.

I hate when people don't give examples, so I am going to throw one here. just the other day, I asked the newest and most expensive claude model to write an LRU and to have a running tally of the capacity of bytes in the cache as the threshold to evict something from the cache. It wrongly implemented the threshold checks and just tracked how many elements were in the cache. this might sound small, but scale that mistake up to a real production system. this is literally unusable. and the expectation to sit there, have it generate 1000s of lines of code for you, and then spot check that small but huge error is not worth it. you have to move so slow to spot check everything - to the point that it's literally faster to type it. This is a model that costs $100s to run per hour and is advertised as "PHD level intelligence" making High school AP computer science to freshman computer science errors - like come on.

If you're reading this, are an expert in your field, and are actually worried about your job - you got be able to have some mental fortitude and not fall for this ...

drudolph914

26 日前議論

I’ve personally reached a point where if I’m saving time typing, but gaining back that time on reviewing and understanding, I may as well have just written the code most of the time. AI is great for boilerplate and learning, but my team has given up on trying to use the tool for entire implementations

drudolph914

26 日前議論

it's possible I'm stating the obvious for this crowd, but if people like this, you will probably also appreciate Pico-8 [1]

[1] - https://www.lexaloffle.com/pico-8.php

drudolph914

先月議論

there is a lot to pull apart here

there is always an aspect of every job that is performative - even small companies. I like to call this perception management. a lot of any job is effectively communicating what you're doing. a lot of effective communication is also not just saying what you're doing, but also how you deliver the information. people are more likely to listen when you communicate things in a more positive tone, make the information concise in a bottom-line up-front style, use a deeper voice (told to me by my wife and women colleagues), and pace the information in a way that lets people ask follow up questions iff needed. no one should _have_ to do all this, but it does change people's perception of how competent you are. I've seen both sides of this coin - amazing engineers that get no promo because they can't communicate, and mediocre engineers that get promoted quickly due to their ability to communicate. I'd almost even argue that this is how should be - as you climb the corporate ladder, communication becomes a lot more important than technical skills and ability

to your point about 1:1s: if you're not getting anything out of your 1:1s, that's a skill issue and is on you IMHO. even when I had bad managers, I was able to effectively communicate my needs, goals, updates, thoughts, as well as give feedback back; in doing so, I've been able to turn horrible manager-team dynamics into a positive experiences. and I'd always argue it came down to the fact that the people perceive you directly correlates with how serious they'll take your word

at the same time, I can empathize with the idea that some middle managers are just bodies that get in the way - everyone's had their fair share of that. but if you're actually good at your job and communicating , you should almost always be able to get around them when it's really necessary

EDIT: and this is coming from a person who is and will always want to stay as an IC engineer

drudolph914

5 か月前議論

this is the “chronological newsfeed to auto curated newsfeed moment” but for ai/anthropic … _great_

drudolph914

6 か月前議論

tbf mac is starting to get pretty bad too

drudolph914

8 か月前議論

I think when the author says

> “We programmers are currently living through the devaluation of our craft”

my interpretation of what the author means by devaluation is the general trend that we’re seeing in LLMs

The theory that I hear from investors is as LLMs generally improve, there will exist a day where a LLMs default code output, coupled with continued hardware speeds, will become _good enough_ for the majority of companies - even if the code looks like crap and is 100x slower than it needs to be

This doesn’t mean there won’t be a few companies that still need SWEs to drop down and do engineering, but tbh, the majority of companies today just need a basic web app - and we’ve commoditized web app dev tools to oblivion. I’d even go as far to argue that what most programmers do today isn’t engineering, it’s gluing together an ecosystem of tooling and or API’s.

Real engineering seems to happen outside of work on open source projects, at the mav 7 on specialized teams, or at niche deeply technical startups

EDIT: I’m not saying this is good or bad, but I’m just making the observation that there is a trend towards devaluing this work in the economy for the majority of people, and I generally empathize with people who just want stability and to raise a family within reasonable means

drudolph914

8 か月前議論

Great, another feature I need to figure out how to turn off

drudolph914

9 か月前議論

80-20 is also a gracious ratio, my experience it’s more like 65-35

drudolph914

10 か月前議論

I am an educator alongside being an engineer, so I've had to think about how to explain this topic to people in ways that give them some kind of intuition/insight. I don't have a good take for non-stem people, but I think I have a better explanation for people who are CS adjacent

I like to explain this whole hallucination problem by stating that LLMs are 2 different machines working together. one half of the machine is all the knowledge it was trained on, and you can think of this knowledge as an enormous classic tree you learn in CS classes; and each node in this tree is a token. the other half of the machine is a program that walks through this enormous tree and prints the token it's on

when you think of it like this, 3 things become immediately obvious

1. LLMs are a totally deterministic machine

2. you can make them seem smart by randomizing the walk through the knowledge tree

3. hallucinations are a side effect of trying to randomize the knowledge tree walk

I find it interesting that LLM companies are trying to fix such a fundamental problem by training the model to always guess the correct path. the problem I see with this approach is that 2 people can enter the same input text, but want 2 different outputs. if there isn't always a _correct path_ then you can't really fix the problem.

the only 2 options you have to “improve” things is prune and or add better data to the knowledge tree, or you’re trying the make the program that walks the knowledge tree take better paths.

the prune/add data approach is slightly better because it’s improving the quality of the token output. but the downside is you quickly realize that you need a fire hose of new human data to keep improving - but much of the data out there is starting to be generated by the LLMs - which leads to this inbreeding effect where the model gets worse

the 2nd approach feels less ideal because it will slow down the process of generating tokens.

all of this to say, from this point on, it’s just hacks, ducktape, and bandaids

drudolph914

10 か月前議論

there is a lot of half correct and half in-correct information in this thread. it might be worth reading some of the articles im linking

the high-level facts are

1.) unemployment and number of available jobs is bad right now, and inflation never got back down to 2% after covid. So Powell made the announcement to lower interest rates. this effect will raise inflation, but create more jobs - which is the correct and more important thing to focus on right now

on top of this, tariffs are making things worse for the average american. based on what powell is saying, the current estimates claim the tariff's alongside the planned increase in inflation will lead to about a 20% increase in prices for the average consumer, but this one time 20% increase is better than having no jobs!

2.) the government has been overestimating the amount of available jobs for 10+ years. A large part of why this is happening is because of the gig economy

an example of what I mean is if you sign up to be an uber driver, uber registers you, the driver, as its own company with US government. this kind of thing is fine for uber, but the government doesn't count you becoming an uber driver as 1 new job - they were counting it as roughly 7 newly available jobs. this is because each new company created in the US roughly brings on 7 employees. larger private financial institutions were correcting for this, but the department of labor statistics hasn't corrected for this. this means banks and private institutions have had better data than the government on the job market for years and were calculating that in to the stock market

3.) to add another layer of confusion, the government calculates the unemployment rate by counting the number of US citizens that file for unemployment checks, but many people found it easier/faster to get a gig economy job in between full time jobs - rather than waiting a 1+ month(s) to get on unemployment checks. this means that the number of people who are unemployed is way higher/worse than what the government is reporting. what this means is that method used to count available jobs AND unemployment are wildly wrong - there are less available jobs and more people unemployed by about 5-7x what was reported this summer.

On top of that, if you look at states where there are stricter/more requirements to become an uber driver, it actually shows the unemployment rate in those states is much higher than expected. the avg unemployment rate amongst these states are probably more accurate to how bad the unemployment situation is in the US overall

4.) the current US administration has fired a lot of employees, which has led to even worse labor statistics/estimates compared to previous years

5.) trump specifically has actually caused a lot of confusion for the average person trying to understand this year's US economic status because we use to have quarterly checkins in June, but as of the past 2 years we've been doing it in July. The way the government tracks important economic indicators starts with the US gov announcing their initial stats, but these numbers often over estimate; so the US gov will often have a large correction the following month

trump this year has been making claims like, "this is the best GDP we've seen in July of recent years!" but of course it's the best because he is intentionally doing the comparison wrong

to ELI5 what I mean, June 2024 had the over estimates stats and the US government would correct them in July 2024. but now in 2025, July is the month with over estimate, and August will be the month we correct the estimates

what we should be doing is comparing august 2025's GDP with July of 2024's GDP. doing so would show you that GDP is not better, but essentially stagnant

trump and his administration are intentionally not doing the comparison correctly for better sounding headlines

[0] dept of labor statistic report- https://www.bls.gov/news.release/pdf/empsit.pdf

[1] Deep Dive: The US Jobs Market Is Much Weaker Than it Appears - https://www.financialsense.com/blog/20854/deep-dive-us-jobs-...

EDIT - typos

drudolph914

10 か月前議論

I love lit. I've been using lit in production since 2020 and I have never looked back. There is so much to say about lit, but I think the biggest win is that it's built on a very stable foundation. Building apps on top of native web components coupled with all the modern QoL features with Lit allows me to without the fear of some new framework/update coming along to the ecosystem - which in the FE world means the last X years worth of code becomes outdated. Native web components are a stable feature in all browsers and I can just focus on building - more engineering teams need to give it a try

If you're curious about lit and like longer form content - I recommend watching the [0] http 203 video that talks about lit element and other tools like it

[0] - https://www.youtube.com/watch?v=uCHZJy2n8Qs

コメント