HNHacker News
TopNewBestAskShowJobs

asa400

1,194 karma · joined April 10, 2013

submissionscomments
asa400··on What years of production-grade concurrency teaches us about building AI agents
> I can't even dev elixir on my N100 minipc, it's too demanding.

Isn't the N100 a quad core machine? It can't run Elixir?

asa400··on What years of production-grade concurrency teaches us about building AI agents
You can (and should) always handle whatever errors you actually want to and are able handle, so if that means catching a lot of them and forwarding them to a model, that's not a problem.

The benefit comes mainly from what happens when you encounter unknown errors or errors that you can't handle or errors that would get you into an invalid state. It's normal in BEAM languages to handle the errors you want to/can handle and let the runtime deal with the other transient/unknown errors to restart the process into a known good state.

The big point really is preventing state corruption, so the types of patterns the BEAM encourages will go a long way toward preventing you from accidentally ending up in some kind of unknown zombie state with your model, like for example if your model or control plane think they are connected to each other but actually aren't. That kind of thing.

Happy to clarify more if this sounds strange.

asa400··on Tesla Sales Down 55% UK, 58% Spain, 59% Germany, 81% Netherlands, 93% Norway
It's a cult. Really. This is a stock devoid of any connection to reality, fundamentals, products, anything. All that matters is that Musk is on the nameplate. Do not bet against the ability of people in a cult to remain irrational in the face of overwhelming evidence. Do not touch the cult.
asa400··on GNU Pies – Program Invocation and Execution Supervisor
If you have to explain the pronunciation of the name of your tool in the first sentence, you've already lost.
asa400··on AI-First Company Memos
It's extremely human behavior. We all do it to some degree or another. The incentives work like this:

    - If all your peers are doing it and you do it and it doesn't work, it's not your fault, because all your peers were doing it too. "Who could have known? Everyone was doing it."
    - If all your peers _aren't_ doing it and you do it and it doesn't work, it's your fault alone, and your board and shareholders crucify you. "You idiot! What were you thinking? You should have just played it safe with our existing revenue streams."
And the one for what's happening with RTO, AI, etc.:

    - If all your peers are doing it and you _don't do it_ and it _works_, your board crucifies you for missing a plainly obvious sea change to the upside. "You idiot! How did you miss this? Everyone else was doing it!"
Non-founder/mercenary C-suites are incentivized to be fundamentally conservative by shareholders and boards. This is not necessarily bad, but sometimes it leads to funny aggregate behavior, like we're seeing now, when a critical mass of participants and/or money passes some arbitrary threshold resulting in a social environment that makes it hard for the remaining participants to sit on the sidelines.

Imagine a CEO going to their board today and going, "we're going to sit out on potentially historic productivity gains because we think everyone else in the United States is full of shit and we know something they don't". The board responds with, "but everything I've seen on CNBC and Bloomberg says we're the only ones not doing this, you're fired".

asa400··on AI-First Company Memos
> I've never seen any tech theme pushed top down so hard in 20+ years working.

> The common theme is C-suite thinks it will save money and their competitors already figured out out, so they are FOMOing at the mouth about catching up on the savings.

I concur 100%. This is a monkey-see-monkey-do FOMO mania, and it's driven by the C-suite, not rank-and-file. I've never seen anything like it.

Other sticky "productivity movements" - or, if you're less generous like me, fads - at the level of the individual and the team, for example agile development methodologies or object oriented programming or test driven development, have generally been invented and promoted by the rank and file or by middle management. They may or may not have had some level of industry astroturfing to them (see: agile), but to me the crucial difference is that they were mostly pushed by a vanguard of practitioners who were at most one level removed from the coal face.

Now, this is not to say there aren't developers and non-developer workers out there using this stuff with great effectiveness and singing its praises. That _is_ happening. But they're not at the leading edge of it mandating company-wide adoption.

What we are seeing now is, to a first approximation, the result of herd behavior at the C-level. It should be incredibly concerning to all of us that such a small group of lemming-like people should have such an enormously outsized role in both allocating capital and running our lives.

asa400··on Eight more months of agents
That these facets of use exist at all are indicative of immature product design.

These are leaked implementation details that the labs are forcing us to know because these are weak, early products and they’re still exploring the design space. The median user doesn’t want to and shouldn’t have to care about details like this.

Future products in this space won’t have them and future users won’t be left in the dust by not learning them today.

Python programmers aren’t left behind by not knowing malloc and free.

asa400··on LLMs as the new high level language
The process of delivering useful, working software for nontrivial problems cannot be reduced to simply emitting machine instructions as text.
asa400··on LLMs as the new high level language
If all you (not you specifically, more of a royal “you” or “we”) are is a collection of skills centered around putting code into an editor and opening pull requests as fast as possible, then sure, you might be cooked.

But if your job depends on taste, design, intuition, sociability, judgement, coaching, inspiring, explaining, or empathy in the context of using technology to solve human problems, you’ll be fine. The premium for these skills is going _way_ up.

asa400··on AI is killing B2B SaaS
This story is more common than any of us would like to admit, myself included.

I can say this with some confidence because as a junior developer, I was often that person, and I didn’t really understand the collective lessons contained within e.g., Rails, Celery, etc.

I was enamored with doing something “cool” and “interesting” and I didn’t realize that this was distinct from and not actually as cool as solving someone else’s problem on time and under budget.

asa400··on AI is killing B2B SaaS
> Sometimes you can explain things and not be heard until you demonstrate them.

1000%. In my experience this isn't even a "sometimes". It's the opposite. Explaining/arguing in the abstract why a significant directional change is needed has little persuasive power. If you can demo something, you actually have a fighting chance, even if it barely works. It can be a working product, it can be charts, it can be a semi-functional mockup, whatever, but you have to show _something_ that isn't just talking about it. I learned this the hard way after advocating for things and getting shot down too many times.

> Also, never underestimate an enterprise’s ability to convince itself that it’s too big and complex for off the shelf tools. Sometimes that’s the case. Very often it’s not.

Couldn't agree more with this, also. I've seen an astonishing number of homemade job runner systems, homemade message queues, homemade databases, homemade web frameworks, homemade languages, etc.

One can make the argument that this is how technical progress happens, and in some sense, that's true (we wouldn't have a lot of nice things if this never happened), but it's rare. Like, really rare. Most shops would be better served with e.g. Django, Postgres, and HTML, deployed on Heroku (or your Heroku-equivalent of choice) rather than spending hundreds of human-months building "an ad hoc, informally-specified, bug-ridden, slow implementation of half of $COMMON_OFF_THE_SHELF_TECHNOLOGY".

So, I completely agree with you.

asa400··on xAI joins SpaceX
Why not put these AI/PV installations somewhere out in the ocean instead? A tiny fraction of the energy required to ship them there, you can actually physically get there to fix them, can use seawater for cooling, can use existing Starlink for connectivity, etc. Why/how is space more economical than international waters?
asa400··on Tesla ending Models S and X production
> It seems counterintuitive, but this helped Tesla which is why Musk championed it. Basically when that tax credit came out, a bunch of Tesla owners had their cars underwater - loans were more than new cars were selling for and depreciation thru the roof. Plus the tax credit helped their competitors.

This makes sense if your business strategy is to get existing Tesla owners to trade their current Teslas to buy new Teslas, rather than to convert non-Tesla owners to buy new Teslas. The latter market is WAY bigger and the tax credit was a huge carrot enticing them to look at a brand they'd never try otherwise in a market where ICE vehicle prices were skyrocketing.

As it stands, there are a ton of Tesla owners who bought their cars with the tax refund, are underwater on them, bitter about it and/or dislike Elon personally, and will never buy a Tesla again. This is churn and brand destruction without a corresponding top of funnel increase.

In contrast, the supercharger network was significant not just for the convenience factor for Tesla owners, but also for the fact that it was a social signal that Tesla was serious about growing the addressable market of EV owners generally by not just making a decent car but making the "EV lifestyle" seem possible to non-EV owners.

If Tesla actually is happy that the tax credit is going away, that seems like they're acknowledging that they're satisfied taking shrinking share of a shrinking market, which is their prerogative, but it's a bad business.

asa400··on Oban, the job processing framework from Elixir, has come to Python
How high of throughput were you working with? I've used Oban at a few places that had what pretty decent throughput and it was OK. Not disagreeing with your approach at all, just trying to get an idea of what kinds of workloads you were running to compare.
asa400··on Why senior engineers let bad projects fail
> I’ve been part of some projects where someone external to the team went on a crusade to shut our work down because they disagreed with it. When we pushed through, shipped it, and it worked well they lost a lot of credibility.

I went through this at a corporate job not too long ago for the first time. One of the more insane things I've dealt with so far in my career. I know life isn't always sunshine and rainbows, but up to that point I'd never seen that level of naked, antisocial self-interest in an org that was ostensibly created to allow people to cooperate toward solving some problem for our customers. Call me naive, but it was really disheartening.

asa400··on Why senior engineers let bad projects fail
> I work in manufacturing and spend a lot of time talking to the people on the factory floor.

Before the pandemic, I used to work doing software for a manufacturing startup and _loved_ that part of it.

I loved hacking something up for the folks on the floor that helped them automate some tedium. They were always so appreciative, not just of the result but also the attention we paid them.

Most of them came from "traditional" manufacturing backgrounds and the way they told it, it was like we were the first software people to ever pay attention to them and the issues they were experiencing.

We weren't even building crazy stuff for them. Most of it was pretty simple, but the bar was _so low_ they were always amazed whenever we were able to give them _anything_ that helped. It was awesome.

asa400··on The Palantir app helping ICE raids in Minneapolis
First, all of what you say is true. I'm going to try to add a little context as someone who is here on the ground, in the city in question.

There is the imminent threat of mass death, and no one here is under any illusions about it.

Every ICE agent is armed, and most have ready access to automatic weapons. These are not well-trained members of an elite organization with a storied, patriotic culture. ICE is a personalist paramilitary organization, and the president has indicated that these ICE agents are immune from consequences, even if they kill people. These are people who volunteered knowing they were going to go into American cities and do violence to people they perceive as their political enemies.

Most of these agents are inexperienced, jittery, poorly trained new recruits away from home. They aren't locals. Their nexus of power and governance isn't local. These are not our community members, they aren't from here, they don't know us or care about us, so they do not empathize with us.

In addition to this, the American citizenry is shockingly well armed. Because everyone involved is so well armed, everybody is slightly touchy about this descending into rioting, because there is a very short path from light rioting to what would essentially amount to civil war. The costs of such any such violence will overwhelmingly be borne by the innocent people who live here, and we know it.

So, people are trying to strike a balance of making sure these people know they aren't welcome here while trying to prevent the situation from spiraling into one in which some terrified agent mag-dumps a crowd of protestors and causes a chain reaction that results in truly catastrophic mass death.

Wish us luck, we're trying.

asa400··on AI is a business model stress test
This makes no sense. If they were better and cost less I would use them for even more work. I'd run more tests, more kinds of tests, more builds, etc. This is classic Jevons paradox territory.
asa400··on AI is a business model stress test
> Making GitHub actions runners more reliable, stable and economical would hurt Microsoft.

Can you explain this one a bit? I know some folks who would absolutely increase their spend if Actions runners were better. As it stands I've seen folks trying to move off of them.

asa400··on Cloudflare CEO on the Italy fines
This is one of the consequences of outsourcing this (and other capabilities) to the private sector.

Many governments simply don’t have the skill and political will to invest in these kinds of capabilities, which puts them at the mercy of private actors that do. Not saying this is good or bad, just trying to describe it as I see it.

asa400··on Databases in 2025: A Year in Review
Do you have a specific use case you're curious about? It's the most widely deployed database software of all time. https://sqlite.org/mostdeployed.html
asa400··on Databases in 2025: A Year in Review
Also just a note: BEGIN CONCURRENT is not in mainline SQLite releases. You need to build your own from a branch. Not a huge deal but just something to note.
asa400··on How Google got its groove back and edged ahead of OpenAI
Additionally, they have built-in distribution and integration for their products. I don’t know how folks don’t see that as a massive advantage.

It’s like Microsoft and Internet Explorer in the 90s but on a much larger scale both in the breadth (number of distribution channels) and depth (market share of those channels in their respective verticals).

asa400··on Why didn't AI “join the workforce” in 2025?
> Allegedly every dollar you spent on an engineer is potentially worth 10x(?) what it was a couple years ago. Meaning your profit per engineer could soar, but tech companies decided they don't want more profit?

Exactly, so many of these claims are complete nonsense. I'm supposed to believe that boards/investors would be fine with companies doing massive layoffs to maintain flat/minuscule growth, when they could keep or expand their current staffing and massively expand their market share and profits with all this increased productivity?

It's ridiculous. If this stuff had truly increased productivity at the levels claimed we would see firms pouring money into technical staff to capitalize on this newfound leverage.

asa400··on Why didn't AI “join the workforce” in 2025?
> the unexpected overperformance of GDP isn’t directly attributed AI but it is very much in the “how did that happen?” conversation.

We spent an amount of money on data centers that was so large that it managed to overcome a self-imposed kick in the nuts from tariffs and then some. The amount of money involved rivals the creation of the railroad system in the United States. Of course GDP overperformed in that scenario.

Where did AI tool use show up in the productivity numbers?

asa400··on Cold-blooded software (2023)
> I was reading about terminal text editors (em, en, vi, vim, neovim, etc.), and it's interesting how some of the software that "lasts" is more like Theseus' Ship. All the original components replaced over time, but the core concepts last.

There's probably a lesson about interfaces here. The thing itself is able to last and adapt if you're able to replace the components, and the components can be replaced if the interfaces between them are stable (or at least knowable, so you can change them and know what you're changing). A couple of the examples I can think of that try to do this are Linux and Clojure. Both have improved and added a ton over the years, but they've always focused on maintaining stable interfaces.

asa400··on FracturedJson
This is good! There are a number of these, so it seems like it's definitely somthing people want. The most popular of which I think is gron[0]. My own is jstream[1]. One tiny point of friendly feedback: you may want to consider adding an example usage/output so folks can see what it does literally.

[0] - https://github.com/tomnomnom/gron [1] - https://github.com/ckampfe/jstream

asa400··on BYD Sells 4.6M Vehicles in 2025, Meets Revised Sales Goal
Germany would have something to say about that.
asa400··on Can Bundler be as fast as uv?
> Man, it's easy to be fast when you're wrong. But of course it is fast because Rust not because it just skips the hard parts of dependency constraint solving and hopes people don't notice.

What's underhanded about this? What are the observable effects of this choice that make it wrong? They reformulated the a problem into a different problem that they could solve faster, and then solved that, and got away with it. Sounds like creative problem solving to me.

asa400··on BYD Sells 4.6M Vehicles in 2025, Meets Revised Sales Goal
> Did they ever learn to compete then? The only thing that protected them then was that Japan was a US “ally” and could be “persuaded” to go along with protectionism. China has no such need.

Oh, indeed. I was attempting to be generous, but it's arguable whether they deserve that generosity.

> I would argue that the 70s were a trial run for whats happening today but instead of becoming more competitive the automakers focused on lobbying for Government help; a playbook that won’t help them today.

We're still paying for this today with the so-called "Chicken Tax" (and all of the other crash and emissions regulations) that has deprived us so many good Japanese trucks over the years.

← PreviousPage 3 of 10Next →