This sounds like someone complaining about how Windows is a black box while ignoring the existence of Linux/BSD.
I'm currently hosting, on very reasonable consumer grade hardware, an LLM that is on par performance wise what every anyone was paying for about a year ago. Including all the layers in between the model and the user.
Llama.cpp serves up Gemma-4-26B-A4B, Open WebUI handles the client details: system prompt, web search, image gen, file uploading etc. With Conduit and Tailscale providing the last layer so I can have a mobile experience as robust as anything I get from Anthropic, plus I know how all the pieces works and can upgrade, enhance, etc to my hearts delight. All this runs from a pretty standard MBP at > 70 tokens/sec.
If you want to better understand the agent side of things, look into Hermes agent and you can start understanding the internals of how all this stuff is done. You can run a very competitive coding agent using modest hardware and open models. In a similar note, image/video gen on local hardware has come a long way.
Just like Linux, you're going to exchanging time for this level of control, but it's something anyone who takes LLMs seriously and has the same concerns can easily get started with.
Yet I still see comments like this that seem to complete ignore the incredible work in the open model community that has been perpetually improving and is starting to really be competitive. If you relax the "local" requirement and just want more performance from an LLM backend you can replace the llama.cpp part with a call to Kimi 2.5 or Minimax 2.7 (which you could feasibly run at home, not kimi though). You can still control all the additional part of the experience but run models that are very competitive with current proprietary SoTA offering, 100% under your control still and a fraction of the price.
Gemma-4-26B-A4B does not require 50+ Gb of vram. It is a MoE model so only 4B of active parameters at a time and not as GPU dependent. I can run it on 16gb of vram and ~20gb of DDR5 regular ram for a 8 bit quant.
Everytime I've tried a local model, and I have tried lots for a couple years now, they just seem like they were overtrained on benchmarks. They consistently perform dramatically worse than even older models from Anthropic/OAI/Google.
That might be true, but still: with Claude Opus I can give a task with 2 lines and it will just do it, with a local Qwen I have to use plan mode for everything even small tasks.
You're spot on btw, not sure why you're getting downvoted. It's funny that a community of supposed "hackers" seems to think your only choice is dolling out money to hyper scalers for what amounts to a code writing SAAS.
> LLMs and LLM providers are massive black boxes... No trust that they won't nerf the tool/model behind the feature... No trust they won't sunset the feature (the graveyard of LLM-features is vast and growing quickly while they throw stuff at the wall to see what sticks)
Doesn't really apply to the article regarding Claude Code Routines in particular. Should this feature disappear, it should be trivially easy to setup a similar pipeline locally, using a cronjob to run opencode configured to use a local LLM. Easy. I have no qualms using a convenient feature I could reimplement myself, it saves me time.
Following along with this, I find the real hits to self-worth post layoff are in the process of finding that next job. Even when you have a job, a serious job search can be exhausting and, depending on the feed back you're getting, really whittle away at self confidence.
But when you can feel the financial timer ticking, you continually start to question yourself and, dangerously, drop your standards. Desperation is a serious trap that can easily lead you to a situation where you are less likely to succeed (despite believing that dropping standards will increase you chances), leading to even further anxiety and insecurity. It's one thing to get rejected from a dream job, but getting rejected from something you internally think is beneath you really stings. Ironically I've found it's in desperate times that confidence and self respect is the most valuable. Clearly, this is much easier said then done.
For people with some financial buffer, you can afford the time to clear your head, and focus on finding something that will lead you to success. Without it, it's possible to have someone who could otherwise end up working for a place like Anthropic getting rejected from a small town startup offering half their previous pay (being a bit hyperbolic here, but I've seen situations like this narrowly avoided).
Being unexpectedly unemployed also starts a virtual timer of sorts not on your terms. Regardless of how you feel about the event, the longer it persists is universally seen as a negative signal to those that would hire you for your next role. It gets exponentially worse as time goes on making it even harder to find a job, because of the increased time you don't have a job.
I'm continually fascinated by the huge differences in individual ability to produce successful results with AI. I always assumed that one of the benefits of AI was "anyone can do this". Then I realized a lot of people I interact with don't really understand the problem they're trying to solve all that well, and have some irrational belief that they can get AI to brute force their way to a solution.
For me I don't even use the more powerful models (just Sonnet 4.6) and have yet to have a project not come out fairly successful in a short period of time. This includes graded live coding examples for interviews, so there is at least some objective measurement that these are functional.
Strangely I find traditional software engineers, especially experienced ones, are generally the worst at achieving success. They often treat working with an agent too much like software engineering and end up building bad software rather than useful solutions to the core problem.
"AI" tools I've got at work (and am mandated to use, complete with usage tracking) aren't a wide-open field of options like what someone experimenting on their own time might have, so I'm stuck with whatever they give me. The projects are brown-field, integrate with obscure industry-specific systems, are heavy with access-control blockers, are already in-flight with near-term feature completion expectations that leave little time for going back and filling in the stuff LLMs need to operate well (extensive test suites, say), and must not wreck the various databases they need to interact with, most of which exist only as a production instance.
I'm sure I could hack together some simple SaaS products with goals and features I'm defining myself in a weekend with these tools all on my own (no communication/coordination overhead, too!), though. I mean for an awful lot of potential products I could do that with just Rails and some gems and no LLM any time I liked over the last 15+ years or whatever, but now I could do it in Typescript or Rust or Go et c. with LLMs, for whatever that's worth. At work, with totally different constraints, the results are far less dramatic and I can't even feasibly attempt to apply some of the (reputedly) most-productive patterns of working with these things.
Meanwhile, LLMs are making all the code-adjacent stuff like slide decks, diagrams, and ticket trackers, incredibly spammy.
[EDIT] Actually, I think the question "why didn't Rails' extreme productivity boost in greenfield tiny-team or solo projects translate into vastly-more-productive development across all sectors where it might have been relevant, and how will LLMs do significantly better than that?" is one I'd like to see, say, a panel of learned LLM boosters address. Not in a shitty troll sort of way, I mean their exploration of why it might play out differently would actually be interesting to me.
> The projects are brown-field, integrate with obscure industry-specific systems, are heavy with access-control blockers
These are cases where I've seen agentic solutions perform the best. My most successful and high impact projects have been at work, getting multiple "obscure industry-specific systems" talking to each other in ways that unblocks an incredible amount of project work.
> I always assumed that one of the benefits of AI was "anyone can do this". Then I realized a lot of people I interact with don't really understand the problem they're trying to solve all that well
I've been through a handful of "anyone can do this" epiphanies since the 90s and have come to realize the full statement should be "anyone can do this if they care about the problem space".
If every project you have tackled has come out successful, then you are managing to never tackle a problem that is secretly literally impossible, which is a property of whatever prefilter you are applying to potential problems. Given that your prefilter has no false positives, the main bit of missing information is how many false negatives it has.
> Strangely I find traditional software engineers, especially experienced ones, are generally the worst at achieving success. They often treat working with an agent too much like software engineering and end up building bad software rather than useful solutions to the core problem.
This feels a bit like a strawman. How do you assess it to be bad software without being an engineer yourself? What constitutes successful for you?
If anything, AI tools have revealed that a lot of people have hubris about building software. With non-engineers believing they're creating successful work without realizing it's a facade of a solution that's a ticking time bomb.
When did I say I'm not a software engineer? I have a software engineering background (I've written reasonably successful books on software), I've just done a lot of other stuff as well that people tend to find more valuable.
> What constitutes successful for you?
The problem I need to solve is solved? I'm not sure what other measure you could have. Honestly, people really misunderstand how to use agents. If you're aim is to "build software" you're going to get in trouble, if your aim is to "solve problems" then you're more aligned with where these tools work most effectively.
Yeah, for those you can just relax and trust the vibes. It's for complex software projects you need those software engineering chops, otherwise you end up with a intractable mess.
If it's for a complex software project the first question you need to ask is "does this really need to be software at all?"
Honestly this is where most traditional engineers get stuck. They keep attacking the old problem with new tools and being frustrated. I agree that agents are not a great way to build "complex software projects" but I think the problem space that is best solved by a "complex software project" is rapidly shrinking.
I've had multiple vendors try to sell my team a product that we can build the core functionality of ourselves in an afternoon. We don't need that functionality to scale to multiple users, server a variety of needs, be adaptable to new use cases: we're not planning to build a SaaS company with it, we just need a simple problem solved.
But these comments are a treasure trove of anecdotes proving exactly my point.
I think AI really pushes this higher up the abstraction layer:
> What problem are you solving?
I've spent a good amount of my careering using engineering and math to solve specific problems, I'm usually adjacent to software teams.
What I've seen happen with agentic coding is that traditional software engineers keep focusing on using it to build software, while ignoring the problem they're trying to solve.
Meanwhile I've seen junior data analysts start interfacing with applications and tools they never dreamed of before, and delivering results to stakeholders in record times. Things that were previously blocked by engineering no longer are.
But many engineers today are not really problem solvers, they're software builders. The idea that solving the end users problem is the goal, not building them software, is incomprehensible.
And so they continue to struggle to use AI effectively because they're trying to build software with it. Which it's not terrible at, but it's really the wrong tool for that job.
Sometimes software is necessary to solve a problem, a few years ago, software was necessary for a fairly large problem surface area (though, to your point, even then a lot of software was not really built to solve those problems). Today that surface area is shrinking, and as economic constraints loom on the horizon, I believe it will increasingly be people who are solving problems (with or without AI) that will be the ones surviving.
> I'm very suspicious of the current international agreement that it's time to take action
Especially since, when you look at the behavior of younger people, they're way more careful about social media than millennials were. My teenage child an their friends keep all of their conversations in a massive but private group chat. Any social media consumed by them, is basically 'read only'. They don't post online, none of them of have social media accounts where they post pictures of themselves etc.
Same with all of my younger gen-z coworkers. If they have socials the post very selectively and all content is work friendly.
The people I see that need "protection" are aging millenials that don't really understand how wildly they're exposing themselves and families. I cringe when I see the amount of personal photos and information shared by the view millenials I know who still need their ego-boost from these platforms (and that number itself is much smaller).
Younger people don't share their opinion and anything resembling private photos online any more.
I definitely would not agree with this and the user metrics of platforms like tiktok and instagram definitely would argue otherwise to your anecdote. Many are showing far more of an alleged window to their lives than ever before, key word being alleged as its always greatly curated in an way that oft attempts to make everything look perfect and effortless.
Absolutely are a lot of gen z who avoid social media, but to pretend most are privately hunkered away is completely ignorant of today's social media usage.
Not long ago I left a reasonably cool AI startup to join an ops heavy (like people physically doing work, running warehouses etc) company. There was some adjustment but the ability to deliver real, concrete, monetary value to people working in the field is incredibly rewarding (and oddly the pay is on par with most bay area startups).
I recently talked to a few companies in the AI space, from (smaller) frontier model labs to companies still looking to build "AI products" and my take away was that, if you're not working for one of the big players, the market hasn't really figured out if there is an "AI engineer" job yet.
I'm increasingly starting to believe that the future of work for people that have technical skills (more than just 'software') is likely going to be working in places that are less about "shipping software" and more about supporting teams doing something physical in the real world.
These companies are also the most ripe to truly leverage AI: they have tons of messy problems that need to be solved and iterated on extremely fast. Operations people tend to be "EoD" deadline people, not quarterly planners. Getting solutions solved in an actionable way on time often means really understanding the core business, the technical space surrounding it, and how to leverage AI to pull of some miracles. It can be stressful, but when you pull it off your stakeholder have sincere and real gratitude and you're actually moving the needle for the company.
I don't think the Bay area, even those sniffing the AI vapors the hardest, is really willing to accept what AI is going to do to software and software companies.
I love working for those companies also, where they are used to waiting months for a small software update and I can do it in hours and they think I'm a wizard.
The best outcome is bespoke software for every company and small "ops heavy" (in startup context) startups have a window to grow like weeds. Imagine the culture shock and legal / procurement process for an established player to bring a vendor in to build this for them. It won't work, it needs to be an internal team, but even then, the internal politics, and short term affects to people's bonuses and incentives will make it almost impossible.
I give this example as I previously worked at a big European REIT. My job was to implement renewable energy across the portfolio which on paper was a no-brainer due to legislation and grants / feed in tariffs etc.
We got huge pushback from every angle with the local teams, people paying lip service to drag it out and delay. Eventually I got to the root cause... The capex had to come out of the business unit, and the payback would negatively affect their KPIs and bonus. Next time I came across this kind of issue, I asked to see the incentive structure before approaching anyone.
And "Do what thou wilt shall be the whole of the Law", does not mean "you can do anything you want", it's much, much closer in meaning to the famous quote from the Upanishads (Crowley largely felt that much of esoterism was basically, as Dion Fortune put it "Yoga of the West"):
> “You are what your deepest desire is. As is your desire, so is your intention. As is your intention, so is your will. As is your will, so is your deed. As is your deed, so is your destiny.”
But HN has increasingly been about having vigorous, opinionated discussion on a surface level understanding of topics (plus a growing number of AI participants), so I'm not sure there's much benefit to pointing it out.
This entire discussion is ridiculous. We shouldn't be conflating the serious economic and philosophical work of someone like Adam Smith with the rantings of someone like Aleister Crowley. It's absurd.
Crowley had a penchant for drama and provocation (which, hilariously, still seems effective today!), but writing off his work as "rantings" is really ignoring the scholarly work he did on Western esotericism (and he also did make a lot of progress bringing greater awareness of non-Western traditions as well), which has always been an important part of Western culture (even if commonly underplayed by mainstream academia).
Technology, science and the occult have always had an interesting relationship in the West. Pythagoras, in addition to his contributions mathematics, is famous for founding a Hermetic mystery cult. As I'm sure you know, inventing physics and calculus was basically a part of Isaac Newton's study of alchemy (which has long had a big of a mystical component pulling from the Hermetic tradition as it has a proto-chemical component), and even Jack Parsons followed in that tradition (being himself a student of Crowley).
It's completely understandable if you don't find Crowley work of interest, but plenty of people also don't find Adam Smith's work of much interest either. Dismissing the work of Western esotericism on the history of the West would be similar to dismissing Sufism on the history of the Middle East.
Ah HNs favorite strawman the "dogmatically and ideologically opposed to anything AI" person who, from my experience, largely doesn't exist.
However I was completely unimpressed with this tool when I saw it this weekend for two reasons:
The first is directly related to how this is built:
> These are rough LLM estimates, not rigorous predictions.
This visualization is neat (well except for reason number two), but it's pretty much just AI slop repackaged. There's no substance behind any of these predictions. Now I'm perfectly open to the critique that normal BLS predictions are also potentially slop, but I don't see how this is particularly valuable.
And the second, like 8% of male population I'm colorblind, so I can't read this chart.
For the record, I do agentic coding pretty much everyday, have shipped AI products, done work in AI research, etc.
Ironically, it's comments like yours that keep me the most skeptical. The fact that an attack on a strawman is the top comment really makes me feel like there is some sort of true mania here that I might even be a bit caught up in.
I'm colorblind as well and what's fascinating to me is that this is the second AI created chart in a week I've seen that I can't read. Surprisingly I've found such agressively colorblind-unfriendly charts to be far less common when created by humans.
That all relies on the assumption of petro-dollar, something that could have been taken for granted during the last 50 years but could easily change within weeks now.
No it doesn’t, the petro-dollar isn’t a real thing. Forcing USD denomination for a transaction doesn’t help USD because there is a buyer of USD and an equally sized seller of USD.
However, there does seem to be an outsized effort applied to defending this not-real thing. A leader who defies the petrodollar has a good chance of getting killed or kidnapped. In a way, the same principle makes any god real: he doesn't have to exist, as long as people who will beat up non-believers do.
Total USD reserve involved in that transaction is not zero. They have to already hold dollars to do the transaction at all, which means a benefit has already been provided to the US. The transaction doesn't change the US's position, but enabling the transaction to occur does.
Petrodollar hypothesis is debunked. The total volume of petrodollar trade approximates minutes in stock markets. This simply isn’t a real factor anymore; a lot of people think it is because there are writings from the 70s that are compelling.
As you said, trade volume doesn't drive valuation, it's reserve size.
Which goes up with inflation btw, so you can export inflation. Actors who maintain reserves of your currency will have to keep buying more from you, providing you with benefits.
I'm currently hosting, on very reasonable consumer grade hardware, an LLM that is on par performance wise what every anyone was paying for about a year ago. Including all the layers in between the model and the user.
Llama.cpp serves up Gemma-4-26B-A4B, Open WebUI handles the client details: system prompt, web search, image gen, file uploading etc. With Conduit and Tailscale providing the last layer so I can have a mobile experience as robust as anything I get from Anthropic, plus I know how all the pieces works and can upgrade, enhance, etc to my hearts delight. All this runs from a pretty standard MBP at > 70 tokens/sec.
If you want to better understand the agent side of things, look into Hermes agent and you can start understanding the internals of how all this stuff is done. You can run a very competitive coding agent using modest hardware and open models. In a similar note, image/video gen on local hardware has come a long way.
Just like Linux, you're going to exchanging time for this level of control, but it's something anyone who takes LLMs seriously and has the same concerns can easily get started with.
Yet I still see comments like this that seem to complete ignore the incredible work in the open model community that has been perpetually improving and is starting to really be competitive. If you relax the "local" requirement and just want more performance from an LLM backend you can replace the llama.cpp part with a call to Kimi 2.5 or Minimax 2.7 (which you could feasibly run at home, not kimi though). You can still control all the additional part of the experience but run models that are very competitive with current proprietary SoTA offering, 100% under your control still and a fraction of the price.
reply