HNHacker News
TopNewBestAskShowJobs

sensanaty

5,068 karma · joined August 8, 2020

submissionscomments
sensanaty··on How is the Bun rewrite in Rust going?
I currently work at a company where the average tenure is somewhere around 7 years. Wanna know their crazy secret? They don't treat their employees like disposable cattle, and actually invest in them! Our salaries go up yearly a few % above inflation as the bare minimum, we have great benefits, there's no crunch or overworking culture etc. It's a company with ~5000 employees globally too, not a tiny startup burning cash.

I'm similar to many people, I wouldn't have left some jobs in the past had I been treated with a modicum of respect and if my compensation went up with my tenure. But time and time again, pay ranges go up as new hires join in, but you're treated like you're insane for asking your compensation as a senior to match the compensation of a newly joined junior.

sensanaty··on Professor's invisible prompt trap catches 32/35 students cheating with AI
I see adults at the job who do not even bother reading a single word from the AI slop they shove down everyone else's throat literally daily. They will copy/paste a question, and just copy/paste the answer with 0 proofreading or checking or anything.

I can so easily believe that kids do the exact same thing.

sensanaty··on What is happening to jobs? Separating AI hype from reality
I'm going to copy/paste a comment of mine from another article the other week, since this discussion seems quite relevant to the topic at hand.

https://news.ycombinator.com/item?id=48743713

---

We've just done an official evaluation at work, using extensive statistics on our gigantic monorepo in a company with ~2000 devs over the course of 2 years, everyone from hardware engineers to regular old frontend engineers. It's a highly profitable and mature public company, and has been for going on a decade at this point without missing a beat. We were given infinite access & budgets to basically any and all AI tooling we could imagine, and we have several "AI Native" teams (whatever the fuck that even means). We're doing agentic coding, we have harnesses of all kind, skills, we have many teams doing spec-driven development, designers using all the various things like Figma Make and access to tools like Devin/Factory Droid/Claude Code/Codex/etc.

This is all to say, we as a company are using AI a lot in all possible corners, but thankfully our leadership isn't schizophrenic and isn't mandating everyone hit token limits or whatever, it's more of a "Let's see what works and what doesn't" type of thing, and we measure a lot of statistics. Nobody here really cares whether LLMs are the next coming of Christ or not, as a company there are many people (even in SLT) that are indifferent to LLMs, and many who are reasonably hyped.

I wish I could link to the actual document we were all shown since it has a beautiful breakdown of the methodology and a fine-grained breakdown of the stats and the categories measured, but in the grand scheme of things, ALL the AI tooling we have implemented (at least on the engineering side of the equation) has contributed to a total of... drum roll please... 7 (seven) Percent overall productivity increase! The most productive teams saw a productivity increase of around 20%, while some teams actually saw drops in productivity into the negative percentage points. My team, none of us really give a shit about AI and we're somewhere in the 3-5% range on certain categories of tasks, which I'd say is a fairly good assessment.

Productivity here is measured in many ways, including but not limited to speed of MR review and merge times, feature/ticket/roadmap closure/delivery, rollback/revert incidence rate, how often people interact with the MR review bots and implement their suggestions/fixes, how many times people check back on AI transcriptions/meeting notes (hint: Nobody looks back on any of it, it's all just noise that gets generated and never actually referenced outside a few extremely rare cases) and many more things I'm forgetting. It is an imperfect number of course, because measuring productivity in engineering is a sisyphean task, but in my opinion it is accurate to the reality on the ground and outside of all the hype and marketing bullshit.

So, I remain thoroughly unconvinced of these personal anecdotes of people being "massively" more productive, especially once you factor in the fact that we now have a 2000EUR budget/month/dev for all the AI tooling, those productivity numbers start looking pathetic once you factor in the costs (which are only increasing as the AI companies need to start recouping the gazillions they've burned). Some teams have started begging to disable coderabbit and other similar tools in their MRs because they're producing nothing but walls of noise that makes reviewing any MR a nightmare of sludging through endless slop of useless bullshit, ours included.

sensanaty··on Git rebase -i is not that scary
I use it often as well, but if I'm rebasing more than a handful of commits it becomes very annoying and confusing if there's a conflict. For example, I often run into the situation where I'll get a conflict, I resolve that conflict on the first rebase step, but then the same conflict keeps popping up. Where did the resolution I just did disappear to? Why do I have to keep redoing the resolution?
sensanaty··on Future euro banknote design proposals
I really dislike the back of the notes on pretty much all of them. Generic crowds of people with ugly backdrops, just feels very uninspired especially when some of the front designs are so bold
sensanaty··on Be skeptical of OpenAI's rogue hacker agent story
That's cause many people here work for AI companies and have massive vested interests in propagating & hyping the bubble.
sensanaty··on OpenAI’s accidental attack against Hugging Face is science fiction that happened
Do I believe that a benchmark created by the biggest grifters on the planet is propaganda? No, just like VW with their emissions, I'm sure Anthropic wouldn't dare fake an opaque benchmark that they created to hype their own products, a product they are intentionally marketing as being dangerous (yet they continue to tweak it and profit off of it despite the apparent danger).

It says that 4.7 and 4.8 score well too, well above the human experts. Those have been available for a good while, where are all these crazy engineered viruses created by Opus 4.7?

Every single word spoken by the people working at these companies is a lie. Every single benchmark is gamed, every single statistic they put out has been proven time and time again to be fudged or straight up fake. The only reason they're angling for this idiotic danger angle is so they can go to daddy Trump and beg him to ban those big bad evil Commie models, since people are realizing all this crap is at best a moderately useful tool in software engineering, and they're looking to IPO so they can drop the bag off with the poor shmucks who aren't a part of their cabal of sociopaths.

They can barely make a functional TUI (using fucking React of all things to boot) with infinite money and infinite access to the Deities they've created in their minds, and I'm supposed to believe they're capable of bio-engineering a virus that will somehow break containment?

sensanaty··on OpenAI’s accidental attack against Hugging Face is science fiction that happened
It's hilarious that people find the idea that these 2 would collude in making this elaborate PR scheme ridiculous. As if the psychopaths like Sama and the rest of his ilk wouldn't ever dream of doing something companies have been doing since time immemorial.
sensanaty··on OpenAI’s accidental attack against Hugging Face is science fiction that happened
My favorite part of this discourse is people somehow finding it preposterous that 2 companies filled to the brim with AI sycophants who regularly lie - and in Sam's case, basically every single word he breathes out is a lie - who have massive vested interests in this tech succeeding couldn't possibly collude together to shore up this facade as a marketing stunt.
sensanaty··on OpenAI’s accidental attack against Hugging Face is science fiction that happened
> We are lucky that it wasn't a case of an agent running a virology lab benchmark that decides to hack a lab and tries to synthesize something.

Where are you people getting this crap from? In what universe are these LLMs in the territory of engineering viruses? I beg of you to stop slurping the AI company propaganda and marketing and think critically for 5 seconds about what you're insinuating here.

sensanaty··on “We have information that Moonshot distilled Fable for the development of K3”
For a buncha supposed capitalists, they sure do hate fair competition eh?
sensanaty··on OpenAI and Hugging Face address security incident during model evaluation
Because it's a marketing stunt, and if they did the obvious, secure things like airgapping, they wouldn't have had an event to market their new scary model.
sensanaty··on Judge approves $1.5B Anthropic settlement for pirated books used to train Claude
What a fucking joke of a country the US is, allowing this kind of behaviour with such a pathetic "punishment". Barely even qualifies as a tap on the wrist, Anthropic should be getting gutted into non-existence for this shit and the execs should be given the Aaron Swartz treatment.
sensanaty··on OpenAI and Hugging Face address security incident during model evaluation
People "make fun of them" because they say they're building some uber-dangerous deity, yet take literally 0 steps to, I dunno, slow the fuck down for a bit?

Maybe people would take the threats more seriously if the hypemen weren't simultaneously claiming that we have to go at warp speed with all of this.

sensanaty··on The Kimi K3 Moment
> In this world too, the punishment was, ballpark, 100x direct damages.

Which ballpark are you playing in to come to a number like that? I have friends whom are authors, and they've certainly not seen a penny from Anthropic. I somehow doubt they're the only ones out there that haven't been compensated for having their work blatantly stolen. In a just world we'd be ensuring Anthropic was destroyed as a result of their actions, since their entire existence hinged on en-masse piracy.

sensanaty··on Claude Code uses Bun written in Rust now
Fair, that rules out "Anthropic pays you directly."

But look at your own disclosure page [0]: OpenAI paid you for your time at a GPT-5 preview, and you regularly get early access, embargoed previews, and free API credits from OpenAI, Anthropic, Gemini, and Mistral. So "I've made no money from Anthropic" is narrower than it sounds, access and preferential treatment are compensation too, even if cash didn't directly change hands.

Same goes for the sponsor list. Microsoft is OpenAI's investor and infrastructure (or was anyways, I've got no clue what the status of that is these days), Amazon and Google are both major Anthropic investors, and most of that list runs on Azure, AWS, or GCP, or sells tools built on these labs' models. Taking a Microsoft sponsorship isn't really separate from an OpenAI one, it's just a single hop removed. The entire software ecosystem and all the players in it are entangled in this bubble.

I'm not accusing you of consciously hyping any of them. Your posts are high quality even when I disagree with the contents of them. You often list downsides too, prompt injection, security, slop issues but the throughline still repeats: this is inevitable, get on board. Product-market fit found, November 2025 as the inflection point. That kinda framing shows up again and again throughout all your posts.

It's refreshing that you're upfront about the previews and credits and all that jazz, but surely you understand why some people stay skeptical even with that disclosed as it confirms the access exists.

Whether you realize it or not, I think they are indeed paying you in some sense, which is part of why I'm skeptical when your posts consistently land at the top of HN. Not because the writing isn't good, but because I think the same incentive shaping the coverage probably shapes some of that reach too.

[0] https://simonwillison.net/about/#disclosures:~:text=2024%2E-...

sensanaty··on Claude Code uses Bun written in Rust now
[flagged]
sensanaty··on The Kimi K3 Moment
> to someone with rights to serve the contents

Now THAT'S doing some heavy lifting lmao. The vast, vast, VAST majority of the original datasets were from pirated books and the like. Also, arguably a robots.txt is the exact mechanism to follow to do the mass GET-ing, yet the AI cos choose time and time and time again to simply ignore it and be as abusive as they possibly fucking can

sensanaty··on The Kimi K3 Moment
Considering they were the original infringers, I don't know how anyone can expect tears to be shed here. The best we can hope for is for all these cancerous - and they really are the definition of a cancer - money burning entities to all fall apart to distillation attacks like these.
sensanaty··on Kaiser nurses say AI, surveillance are making their jobs and patient care worse
Having seen the type of insane shit these LLM transcription tools shit out on the regular, I wouldn't trust these tools for even the most mundane note taking in a medical setting, especially if there's a summarization feature.

I've seen summaries that implied the opposite of what was actually said, and even the transcripts themselves often contain egregious errors. I can so easily imagine these tools summarizing "I drink coke often" to "I use cocaine often", as a random simple example. All nuance gets lost, all context gets lost and unless you're vigilant and looking out for these types of errors it's so easy to slip by the cracks.

sensanaty··on $100 AI Music Video: Claude Fable 5 vs. GPT-5.6 Sol
Truly impressive that we're burning trillions of dollars for this abject garbage. I'm so glad people can't buy RAM anymore so that we can shit out some more soulless slop - and this really is the definition of slop - like this
sensanaty··on What will be left for us to work on?
I'm going to copy/paste a comment of mine from another article the other week, as I see the discussion here is trending towards AI productivity and all that kinda stuff.

https://news.ycombinator.com/item?id=48743713

---

We've just done an official evaluation at work, using extensive statistics on our gigantic monorepo in a company with ~2000 devs over the course of 2 years, everyone from hardware engineers to regular old frontend engineers. It's a highly profitable and mature public company, and has been for going on a decade at this point without missing a beat. We were given infinite access & budgets to basically any and all AI tooling we could imagine, and we have several "AI Native" teams (whatever the fuck that even means). We're doing agentic coding, we have harnesses of all kind, skills, we have many teams doing spec-driven development, designers using all the various things like Figma Make and access to tools like Devin/Factory Droid/Claude Code/Codex/etc.

This is all to say, we as a company are using AI a lot in all possible corners, but thankfully our leadership isn't schizophrenic and isn't mandating everyone hit token limits or whatever, it's more of a "Let's see what works and what doesn't" type of thing, and we measure a lot of statistics. Nobody here really cares whether LLMs are the next coming of Christ or not, as a company there are many people (even in SLT) that are indifferent to LLMs, and many who are reasonably hyped.

I wish I could link to the actual document we were all shown since it has a beautiful breakdown of the methodology and a fine-grained breakdown of the stats and the categories measured, but in the grand scheme of things, ALL the AI tooling we have implemented (at least on the engineering side of the equation) has contributed to a total of... drum roll please... 7 (seven) Percent overall productivity increase! The most productive teams saw a productivity increase of around 20%, while some teams actually saw drops in productivity into the negative percentage points. My team, none of us really give a shit about AI and we're somewhere in the 3-5% range on certain categories of tasks, which I'd say is a fairly good assessment.

Productivity here is measured in many ways, including but not limited to speed of MR review and merge times, feature/ticket/roadmap closure/delivery, rollback/revert incidence rate, how often people interact with the MR review bots and implement their suggestions/fixes, how many times people check back on AI transcriptions/meeting notes (hint: Nobody looks back on any of it, it's all just noise that gets generated and never actually referenced outside a few extremely rare cases) and many more things I'm forgetting. It is an imperfect number of course, because measuring productivity in engineering is a sisyphean task, but in my opinion it is accurate to the reality on the ground and outside of all the hype and marketing bullshit.

So, I remain thoroughly unconvinced of these personal anecdotes of people being "massively" more productive, especially once you factor in the fact that we now have a 2000EUR budget/month/dev for all the AI tooling, those productivity numbers start looking pathetic once you factor in the costs (which are only increasing as the AI companies need to start recouping the gazillions they've burned). Some teams have started begging to disable coderabbit and other similar tools in their MRs because they're producing nothing but walls of noise that makes reviewing any MR a nightmare of sludging through endless slop of useless bullshit, ours included.

sensanaty··on What's slowing down the AI buildout
> To say that paid off nuclear is profitable without subsidies and then mention France is deeply funny.

Care to elaborate on what you find funny?

France's energy regulator (CRE) puts the actual cost of running its nuclear fleet at about ~€61/MWh for 2026-2028. EDF's (Électricité de France) own revenue estimate for 2026 nuclear generation is around €66/MWh. They are currently earning a profit on their power generation, the fleet pulls in over €5 billion annually just from selling the surplus to its neighbors. If that's the punchline of your joke I'm not seeing it.

French industrial electricity runs about €69/MWh vs £168/MWh in the UK, as an example.

> The government (read us, the taxpayers) in most countries will foot the bill for decomisioning

What better use is there of taxes than the running and decommissioning of power plants, nuclear or not? I'd much rather my tax money go to covering all the costs involved with nuclear, from initial construction and everything else involved than pretty much anything else most gov'ts waste their money on.

Who should be paying for it, in your mind?

> To talk about them not getting subsidies once they are paid off is to ignore the fact that no reactor would ever get built without the government providing them a guaranteed backstop.

Again, I fail to see how or why this is actually a problem. Power generation is probably the poster child for what governments should be handling and backstopping with tax money. You're also talking as if the large majority of power plants, and I mean coal, natural gas and even renewables aren't also covered by the exact same kind of government backstop.

In the US, nuclear's liability for an accident is capped around $16.3 Billion under the Price-Anderson Act, split across private insurance and an industry-funded pool.

Offshore oil liability is capped by law at $75 million (Yes, only MILLION) per incident, funded by a tiny per-barrel tax, and cleanup costs blow through that number constantly, which is why e.g. Deepwater Horizon needed a special one-off arrangement to handle at all. TVA's 2008 coal ash spill cost over a billion dollars to clean up and got passed straight onto ratepayers because TVA hit its own debt ceiling. And right now, in 2026, the EPA is actively proposing to weaken coal ash cleanup rules, meaning the backstop on coal's mess is getting worse, not better, while it's still being burned at scale.

So no, nuclear doesn't get some unique golden treatment here. Every energy source with real catastrophic tail risk gets a government backstop, because no private insurer on earth prices genuinely catastrophic low-probability events without one. The only difference is nuclear's cap is a known, bounded number attached to an industry with an extremely good actual safety record in reality, while coal's backstop is open-ended and attached to something that poisons rivers and drinking water on a rolling basis without needing a single accident to do it.

sensanaty··on SpaceX wants to launch 100k more Starlink satellites for 100x the bandwidth
> Because Indonesia has a massive population

Indonesia has less people, ~280 vs ~340 million for the US.

> much more densly populated.

Technically true, but as I said in my comment I'm talking about remote/small islands, closer to Flores/Papua to the east, not Jawa where the overwhelming majority (as in, literally more than half the country) of the population lives. Jawa alone has ~145 million people, Sumatra around 60 and Sulawesi around 20, the rest are spread out all over the country and the densities are much smaller, not to mention that those smaller regions have much fewer resources available to them to begin with.

> It's internet users are vastly mobile as opposed to desktop or LAN connections.

The comment I was replying to was talking about companies/the gov't being reluctant to erect cell towers though? So I don't see the relevancy here, I was saying that even tiny, sparsely populated islands in some of the poorest regions of Indonesia have cell tower coverage, and with those towers, decent internet.

> The US geographic landscape and computer use landscape are entirely different.

Arguably it's infinitely easier to erect infrastructure when you're not an archipelago. Plus, the US is the richest country on earth, I'm sure you guys can figure out how to string along fiber even in the most rural of areas, you've somehow managed to do it across the atlantic ocean after all.

sensanaty··on FCC approves test of space mirror to light night sky
What the fuck are we doing as a species?
sensanaty··on SpaceX wants to launch 100k more Starlink satellites for 100x the bandwidth
I've been on tiny Indonesian islands far from anything that could be considered civilization, and they'll have cell towers more often than not.
sensanaty··on What's slowing down the AI buildout
I don't think anyone reasonable is saying we should only do nuclear. The actual position is that we need a mix of both nuclear and renewables while ramping down non-renewable sources as fast as possible, because you need both intermittent cheap generation and reliable baseload that doesn't care if the wind stops blowing or clouds roll over the panels.

The pricing story remains unconvincing to me. LCOE comparisons measure the cost of producing a unit of electricity, but they don't account for what it costs to actually integrate that electricity into a working grid. Solar and wind are cheap on paper, but they still need storage or backup generation to cover the hours they're not producing, and that cost gets left out of the simple dollar per MWh number. Once you factor in grid balancing and storage, the gap closes a lot, and in some studies it disappears entirely depending on the region's wind and solar resources. Obviously things have to be considered in the context of where we're putting these power plants, Norwegians can probably afford to rely a lot more on offshore wind and hydro than a place like Serbia could.

Nuclear plants will obviously be expensive to build, and will likely always be more expensive than renewables to initially build just due to the sheer complexity involved when compared to renewable sources, but I can't think of a better use of that money personally, especially compared to how much gets spent on far worse things like chucking a billion dollars worth of missiles at Iran[0] for quite literally no reason and especially benefit to anyone (other than arms industry I guess) whatsoever.

Also, already built and paid off nuclear plants are consistently some of the cheapest electricity around, often profitable without any subsidy at all. 70%(!!!!) of France's electricity comes from nuclear. They exported more electricity last year than countries like Belgium consumes in an entire year. They enjoy some of the cheapest power prices in the region because of their historical nuclear fleet, and had people had the same attitude back in the day (looking at you, German "green" party), that fleet would've never gotten built. Letting your offspring enjoy the shade of the tree you plant today and all that stuff, it'll always be too expensive until we actually hunker down and just build the bloody things.

And yeah, new builds like Flamanville have been a mess on cost and timeline, that's a fair criticism of new construction, but it doesn't erase what a mature nuclear fleet does for a grid once it's actually built, and can generally be applied to pretty much any infrastructure project anywhere in the world.

[0] https://news.ycombinator.com/item?id=48845514 - US seeks cheaper hunter-killer drones after Iran destroys $1B worth of Reapers

sensanaty··on What's slowing down the AI buildout
This is completely BS. There were no deaths from Fukushima from radiation exposure itself (one worker's lung cancer death was later linked to it, but the dose in question makes even that contested), and the deaths that did happen in Fukushima came overwhelmingly from a botched evacuation, not radiation. Elderly patients dying from disrupted care and panic, not fallout. And Fukushima only happened because of multiple natural disasters aligning in the most unfortunate way possible, including a fucking tsunami hitting the plant directly.

Chernobyl was a disaster caused by Soviet incompetence combined with an RBMK reactor design that literally had no proper containment structure. That's the worst case that has ever occurred, nothing like it has happened since, and it can't really happen again with any reactor built with a real containment vessel, which is standard everywhere outside the old Soviet bloc. Three Mile Island is the perfect example, a partial meltdown caused by a mix of mechanical failure and operator error back in 1979, and the containment held. Zero deaths, zero meaningful radiation release. Even with human screwups and old tech, the actual design of the reactor kept it from becoming a Chernobyl.

And even taking Chernobyl as the worst case scenario we've ever seen, the exclusion zone covers about 2500 km^2 of land area, which isn't small, but it's not "large swathes of land" either. You get irradiated from flying in airplanes more than walking around most of the exclusion zone these days, obviously other than around Pripyat itself. Hell, you can take a walking tour through the zone, I'd very much recommend it (after the war concludes :p)

Meanwhile the reality right now is that fossil fuel sources are polluting the entire planet actively. Coal plants release more radioactive material into the environment through fly ash than nuclear plants do under normal operation. Coal is far from clean in terms of radioactivity on top of everything else it's doing, and it's actively making the world uninhabitable at a much larger and faster scale than nuclear ever has, but yet we're holding nuclear to some higher standard due to years of propaganda.

It's not a perfect technology obviously, but it sure as hell is better than basically all fossil fuel sources.

sensanaty··on What's slowing down the AI buildout
Not to mention completely subjective, I personally love the look of windmills, and I live in the Netherlands in a location where I see windmills (and I don't mean the old-timey ones, the big power generating kind) everywhere around me for miles. It's certainly prettier to watch than smoke stacks.
sensanaty··on Chat Control 1.0 and 2.0 Explained
Yeah 'cause we've never had private corporations like Coca Cola or Chiquita Bananas hire paramilitary forces in South America before.

Both large gov't and large corpo are horrible and both should be equally avoided.

← PreviousPage 7 of 34Next →