{

"scraped_date": "2026-03-23",

"source": "hacker_news",

"total_scraped": 128,

"nontech_count": 27,

"posts": [

{

"id": "47433702",

"title": "Ask HN: How do you deal with people who trust LLMs?",

"link": "https://news.ycombinator.com/item?id=47433702",

"domain": "news.ycombinator.com",

"author": "basilikum",

"score": 153,

"comment_count": 201,

"created_ts": 1773884255,

"is_internal": true,

"post_text": "A lot of people use LLMs as the source of their objective truth. They have a question that would be very well answered with a search leading to a reputable source, but instead they ask some LLM chat bot and just blindly trust whatever it says.<p>How do you deal with that? Do you try to tell them about hallucinations and that LLMs have no concept of true or false? Or do you just let them be? What do you do when they do that in a conversation with you or encounter LLMs being used as a source for something that affects you?",

"is_ask_hn": true,

"matched_keywords": [],

"comments": [

{

"top": "I'm going to hold them to the same standard no matter if they use crappy sources, plagiarize, or hallucinate on their own. If someone asked, when and if I am in a position where I have to tell them, I would remind them that LLMs prioritize their own confidence over correctness.\nLLMs aren't a special case to me. Glue doesn't belong on pizza and you shouldn't eat one rock a day but we've been giving and getting bad advice forever. The person needs to take ownership for the output and getting it right, no matter the source, is their responsibility.",

"author": "ddawson",

"replies": [

{

"text": "I think they are a little special. People can really turn their brain off and not even know about the source. They don't need to read theough a source or reformat the content to the typical blurb arguments. They can read it off the screen without even understanding what the words mean, which is much harder for most other sources.",

"author": "giantg2",

"depth": 1

},

{

"text": "this. llm's aren't that special, access _maybe_, but there's plenty of access to terrible rumor mills.",

"author": "raggi",

"depth": 1

},

{

"text": "Yeah, it is not clear to me that the average person is going to do any better with googling then they are with asking an LLM. At least the LLM is mostly the average of all published knowledge (and misinformation).\nI feel like people that ask questions like this must have much smarter friends and family members than I do.\nI know people that still believe in Pizzagate or chemtrails or that vaccines cause autism. Clearly finding reputable information sources is not a strong suit for a lot (half?) of the population.",

"author": "ryandvm",

"depth": 1

},

{

"text": "i love you",

"author": "Marciplan",

"depth": 1

}

]

},

{

"top": "> a reputable source\nNews reporters and editors have their biases. Book authors have their biases. Scientists and research papers have their biases. Search engines have their biases. Google too.\nAll human-created systems have biases shaped by the environments, social norms, education, traditions, etc. of their creators and managers.\nSo, the concepts of \"objective truth\" and \"reputable\" need to be analyzed more critically.\nThey seem to be labels given to sources we have learned to trust by habit. Some people trust newspapers over TV. Some people trust some newspapers over other newspapers. All of it often on emotional grounds of agreeability with our own biases. Then we seem to post-rationalize this emotion of agreeability using terms like \"objective truth\" and \"reputable\".\nIs Google search engine that leads to NY Times or Fox News or Wikipedia and makes us manually choose sources as per our biases \"better\" than Google's Gemini engine that summarizes content from all the above sources and gives an average answer? (Note: \"average answer\" as of current versions; in future, its training too may be explicitly biased, like Grok and DeepSeek did).\nPerhaps we can start using terms like \"human sources of information\" versus \"AI sources of information\" and get rid of the contentious terms.\nThen critically analyze whether one set of sources is better than the other, or they complement each other.",

"author": "lovelearning",

"replies": [

{

"text": "> Is Google search engine that leads to NY Times or Fox News or Wikipedia and makes us manually choose sources as per our biases \"better\" than Google's Gemini engine that summarizes content from all the above sources and gives an average answer?\nIf you use just any amount of critical thinking, yes. Truth and objectivity are ideals, not practical states. LLMs are a very bad way to come close to this ideal. You may use them as a search interface to give you sources and then examine the sources, but the output directly is a strict degeneration over primary or secondary sources that you judge critically.",

"author": "basilikum",

"depth": 1

},

{

"text": "> LLMs are a very bad way to come close to this ideal...the output directly is a strict degeneration\nI didn't understand the second part but regarding the first...\nFor me, LLMs are just another source of information with a different UI, analogous to newspapers, TV documentaries, Wikipedia, Google search, YT talks/documentaries, even the majority of informational non-fiction books, and research papers.\nSome may consider some subset of these as reputable sources. But in my mind, the same faculties of skepticism, cynicism, distrust, and benefit-of-the-doubt calculus are activated for all of them, including LLM outputs.\nSo that's one possible answer to your question.\nBut I suggest communicating this through simple illustrative examples to help your target audience understand the problem.\nAbstract terms like primary sources, secondary sources, reputable sources, objective truth, strict degeneration, etc. may not help, especially if they have time or other constraints that make frequent critical examination of sources impractical.",

"author": "lovelearning",

"depth": 2

},

{

"text": "> For me, LLMs are just another source of information with a different UI, analogous to newspapers, TV documentaries, Wikipedia, Google search, YT talks/documentaries, even the majority of informational non-fiction books, and research papers.\nLLM just distils information from those sources and is therefore always a second hand source at best, and a liar at worst. Humans can collect real world data and write about their findings, LLM cannot do that, that makes LLM strictly worse than the best human sources.",

"author": "Jensson",

"depth": 3

},

{

"text": "The nature of the source, whether primary or secondary or tertiary, does not automatically imply anything about their biases. Even a primary research paper on a novel topic may be biased or wrong.\nI agree that LLMs can't collect real world data and write about their findings. But that's true about most human sources too, isn't it? Except primary novel researchers or investigations or philosophies, what is original? Most human-written information is also secondary or lower.\nThe \"best human sources\" does not imply \"ALL human sources.\"",

"author": "lovelearning",

"depth": 4

},

{

"text": "> For me, LLMs are just another source of information with a different UI, analogous to newspapers, TV documentaries, Wikipedia, Google search, YT talks/documentaries, even the majority of informational non-fiction books, and research papers.\nWith all due respect (not trying to be offensive at all) but this is insane to me.\nAll those sources of information you cited have a million incentives to provide fairly correct and checked information. But probably more importantly, they have even more incentives to NOT provide false information. At a minimum, their careers, reputation, recurring work, brand, etc... is on the line.\nAn LLM has zero incentives to provide you with true information, beyond a couple of md files with instructions. If it gets it wrong, there is zero accountability, just an -oh, you're absolutely right- response and move on.\nI agree there is a lot of human bias in the world, but surely we can't even put in the same order of magnitude both types of biases!",

"author": "johnpdoe1234",

"depth": 3

}

]

},

{

"top": "Ask them to tell the LLM it's wrong... then when it goes \"You are absolutely right!\" to challenge it and say that it was a test. Then when it replies, ask it if it's 100% sure. They'll lose faith pretty quick.",

"author": "eranation",

"replies": [

{

"text": "This is an oft-repeated meme, but I’m convinced the people saying it are either blindly repeating it, using bad models/system prompts, or some other issue. Claude Opus will absolutely push back if you disagree. I routinely push back on Claude only to discover on further evaluation that the model was correct.\nAs a test I just did exactly what you said in a Claude Opus 4.6 session about another HN thread. Claude considered* the contradiction, evaluated additional sources, and responded backing up its original claim with more evidence.\nI will add that I use a system prompt that explicitly discourages sycophancy, but this is a single sentence expression of preference and not an indication of fundamental model weakness.\n* I’ll leave the anthropomorphism discussions to Searle; empirically this is the observed output.",

"author": "ericpauley",

"depth": 1

},

{

"text": "If you have 10,000 people flipping coins over and over, one person will be experiencing a streak of heads, another a streak of tails.\nWhich is to say, of a million people who just started playing with LLMs, a bunch of people will get hit or miss, while one guy is winning the neural net lottery and has the experience of the AI nailing every request, some poor bloke is trying to see what all the hype is about and cannot get one response that isn’t fully hallucinated garbage",

"author": "jazzyjackson",

"depth": 2

},

{

"text": "Sure, but that doesn’t explain the \nvolume\n of these complaints. I think the more likely answer is the pitiful sycophancy of some models as demonstrated in BSBench.",

"author": "ericpauley",

"depth": 3

},

{

"text": "Claude Opus 4.6 is the best possible model to use in this test, with the least sycophancy. OpenAI and Gemini models are \nbad\n in comparison.",

"author": "odo1242",

"depth": 2

},

{

"text": "ChatGPT thinking models are very good; the instant model is bad. Gemini is always desperate to find an answer, and will give you one no matter what.",

"author": "mkozlows",

"depth": 3

}

]

},

{

"top": "Not that I've had to deal with this specifically, but I have noticed how the input phrasing in my prompts pushes the LLM in different directions. I've just tried a quick test with `duck.ai` on gpt 4o-mini with:\nA: Why is drinking coffee every day so good for you?\nB: Why is drinking coffee every day so bad for you?\nQuestion A responds that it has \"several health benefits\", antioxidants, liver health, reduced risk of diabetes and Parkinson's.\nQuestion B responds that it may lead to sleep disruption, digestive issues, risk of osteoporosis.\nSame question. One word difference. Two different directions.\nThis makes me take everything with a pinch of salt when I ask \"Would Library A be a good fit for Problem X\" - which is obviously a bit leading; I don't even trust what I hope are more neutral inputs like \"How does Library A apply to Problem Space X\", for example.",

"author": "katet",

"replies": [

{

"text": "Again a model issue. At the risk of coming off as a thread-wide apologist, here are my results on Opus:\nGood:\n> The research is generally positive but it’s not unconditionally “good for you” — the framing matters.\n> What the evidence supports for moderate consumption (3-5 cups/day): lower risk of type 2 diabetes, Parkinson’s, certain liver diseases (including liver cancer), and all-cause mortality……\nBad:\n> The premise is off. Moderate daily coffee consumption (3-5 cups) isn’t considered bad for you by current medical consensus. It’s actually associated with reduced risk of type 2 diabetes, Parkinson’s, and some liver diseases in large epidemiological studies.\n> Where it can cause problems:\nHeavy consumption (6+ cups) can lead to anxiety, insomnia……\nThis isn’t just my own one-off examples. Claude dominates the BSBench: \nhttps://petergpt.github.io/bullshit-benchmark/viewer/index.v...",

"author": "ericpauley",

"depth": 1

},

{

"text": "The BSBench is such a fantastic resource - thank you for sharing.\nWe should really be citing rather than anecdata every time someone brings up hallucinations.",

"author": "johnfn",

"depth": 2

},

{

"text": "What I do for questions like these is read what medical researchers have published. The first I read was \nhttps://pmc.ncbi.nlm.nih.gov/articles/PMC5696634/\n> Coffee consumption was more often associated with benefit than harm for a range of health outcomes across exposures including high versus low, any versus none, and one extra cup a day. There was evidence of a non-linear association between consumption and some outcomes, with summary estimates indicating largest relative risk reduction at intakes of three to four cups a day versus none, including all cause mortality (relative risk 0.83, 95% confidence interval 0.83 to 0.88), cardiovascular mortality (0.81, 0.72 to 0.90), and cardiovascular disease (0.85, 0.80 to 0.90). High versus low consumption was associated with an 18% lower risk of incident cancer (0.82, 0.74 to 0.89). Consumption was also associated with a lower risk of several specific cancers and neurological, metabolic, and liver conditions. Harmful associations were largely nullified by adequate adjustment for smoking, except in pregnancy, where high versus low/no consumption was associated with low birth weight (odds ratio 1.31, 95% confidence interval 1.03 to 1.67), preterm birth in the first (1.22, 1.00 to 1.49) and second (1.12, 1.02 to 1.22) trimester, and pregnancy loss (1.46, 1.06 to 1.99). There was also an association between coffee drinking and risk of fracture in women but not in men.\n> Conclusion Coffee consumption seems generally safe within usual levels of intake, with summary estimates indicating largest risk reduction for various health outcomes at three to four cups a day, and more likely to benefit health than harm.\nWhen I'm looking for medical advice, I want that advice to list things like \"coffee drinking might not be safe during pregnancy\".\nFurthermore, the statement 'Heavy consumption (6+ cups) can lead to anxiety, insomnia ...' assumes caffeinated coffee, yes? The paper I linked to also discusses decaffeinated coffee, eg:\n> High versus low intake of decaffeinated coffee was also associated with lower all cause mortality, with summary estimates indicating largest benefit at three cups a day (0.83, 0.85 to 0.89)28 in a non-linear dose-response analysis. ...\n> Coffee consumption was consistently associated with a lower risk of Parkinson’s disease, even after adjustment for smoking, and across all categories of exposure.22 76 77 Decaffeinated coffee was associated with a lower risk of Parkinson’s disease, which did not reach significance. ...\n> there were no convincing harmful associations between decaffeinated coffee and any health outcome.\nThat nuance seems important.\nAlso note that this paper is incomplete as it investigated defined health outcomes, not physiological outcomes like anxiety. There are plenty more papers, like \nhttps://academic.oup.com/eurheartj/article/46/8/749/7928425?...\n , which considers the time that people drink coffee, also discusses decaffeinated coffee, and highlights the uncertainty about the effect of heavy coffee drinking.\nI don't see why I should care to ask an AI when it's so easy to find well-written research results which are far more likely to cover relevant edge cases.",

"author": "eesmith",

"depth": 2

},

{

"text": "A person would respond the same way? What exactly are you expecting as the output to those questions?",

"author": "tayo42",

"depth": 1

},

{

"text": "Clickbait journalists answers like that, experts mostly don't. But it does make sense it mimics clickbait journalists more since it was trained on the internet.",

"author": "Jensson",

"depth": 2

}

]

},

{

"top": "Is this any different than people who believe random things they read on sketchy news sites or social media?",

"author": "chipgap98",

"replies": [

{

"text": "Yes, somehow. I have been dealing with an awful lot of people who basically have what are theoretically logic degrees who suddenly just take LLMs at face value, or quote them to me like that actually means anything. People I formerly thought were sane.",

"author": "atomicnumber3",

"depth": 1

},

{

"text": "I don't mean to put words in your mouth but from what I've seen, in person but mostly online, but the \"problem\" (and I put that in quotes because I don't even know what to call it... it seems deeper than a mere \"problem\") is that they quote them as if they are autonomous, sentient beings.",

"author": "sodapopcan",

"depth": 2

},

{

"text": "The problem is that LLM output looks like a human conversation. People believe it.\nWhich is more believable?\n“The sky is filled with a downpour of squealing pigs. Would you like me to suggest the best type of umbrella?”\n“Sky pigs squealing”",

"author": "BobbyTables2",

"depth": 3

},

{

"text": "I am not sure I would even say \"believe\", I would think of it more as short-circuiting our critical thinking. I think it taps into something at the core of our tribal instincts. It was famously present in even basic systems like Eliza. And it's not just machines... The same tricks are used by conmen, politicians, and psychopaths, which is more negative than I intend. Even with good intentions and positive outcomes, I feel we need to remember that we drive it, not the other way around.",

"author": "yammosk",

"depth": 4

},

{

"text": "People just don't like to be played for fools. Perhaps us giving into this is progress? I'd give a big ol' \"fuck you\" to anyone who claims it is, but I'm also pretty old.",

"author": "sodapopcan",

"depth": 4

}

]

}

]

},

{

"id": "47457162",

"title": "Tell HN: H&R Block tax software installs a TLS backdoor",

"link": "https://news.ycombinator.com/item?id=47457162",

"domain": "news.ycombinator.com",

"author": "yifanlu",

"score": 140,

"comment_count": 10,

"created_ts": 1774025128,

"is_internal": true,

"post_text": "Just a PSA for folks here in the US because tax season is coming up and some of you may be using H&amp;R Block Business 2025. I discovered that the software installs a root CA named &quot;WK ATX ServerHost 2024&quot; (expiry 2049) into your local machine trusted root certificate store. They also helpfully include the private key to this certificate in a DLL file. This certificate does not identify itself as &quot;H&amp;R Block&quot; anywhere and does not get uninstalled when you uninstall the software.<p>I&#x27;ve been able to successfully use this root CA + mitmproxy to manipulate TLS traffic on a brand new virtual machine on the same network with a DNS spoofing attack. Demo: https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=5paxvYkz1QE<p>To test if your machine is vulnerable visit this page: https:&#x2F;&#x2F;hrbackdoor.yifanlu.com and if you do not get any warning or error message from your browser then you have the backdoor installed. If your browser does complain, you can choose to visit the page anyways for more details on the vulnerability.<p>Is it negligence or a &quot;real&quot; back door? It&#x27;s impossible to tell and since the private key is out there, anyone can use it so the point is moot. There is no legitimate reason why they need to install a wildcard root CA under a different name. When I contacted them about it their statement includes &quot;similar findings have been identified through internal security assessments&quot; meaning they know about this issue but have not fixed it. I would not trust H&amp;R Block software at this point.<p>If you didn&#x27;t get bit by this, congratulations. See this post as a reminder to audit your trusted root CA store.",

"is_ask_hn": false,

"matched_keywords": [

"meaning"

],

"comments": [

{

"top": "When will these companies learn?\nhttps://michael.team/zoom/",

"author": "raw_anon_1111",

"replies": []

},

{

"top": "I'm wondering if download source matters. Seems like most are downloaded straight from their site, but curious if they still offer CDs or if sellers like Amazon have the direct installer downloads.",

"author": "giantg2",

"replies": []

},

{

"top": "\"If you have an SSL error in your H&R Block Software, \n here’s what you need to know.\"\n\n\n\nhttps://www.hrblock.com/tax-center/support/software/technica...",

"author": "WarOnPrivacy",

"replies": []

},

{

"top": "I have the non-business edition installed and still get a privacy error attempting to load your page, so this seems specific to the business edition. Thanks for the heads up.",

"author": "TheClassic",

"replies": []

},

{

"top": "Curious: is it carrying a SHA-1 self-signature?",

"author": "altairprime",

"replies": []

}

]

},

{

"id": "47479115",

"title": "Ask HN: Apple terminated our dev account over a rogue employee",

"link": "https://news.ycombinator.com/item?id=47479115",

"domain": "news.ycombinator.com",

"author": "0x1f",

"score": 128,

"comment_count": 29,

"created_ts": 1774196630,

"is_internal": true,

"post_text": "I know that HN isn&#x27;t a customer support forum and it might not be right to post this here, but we are absolutely desperate and hoping someone in this community can point us in the right direction.<p>We are a small software company in Africa. For over two years, we&#x27;ve built and maintained an app. It has become a vital economic engine for our local community, employing a whole fleet of delivery agents and serving as a lifeline for local stores and restaurants.<p>Recently, we discovered that a single employee used a shared company machine to engage in unauthorized activities that violated Apple&#x27;s Developer Terms of Service.<p>We took immediate action: we fired the employee on the spot and completely overhauled our security. We revoked all individual access and implemented mandatory, peer-reviewed, supervised sessions for any Apple Developer portal access.<p>The problem is the collateral damage. Apple terminated our entire organization&#x27;s account. We submitted an appeal through App Store Connect, but we feel completely stuck behind automated walls. We have also emailed Apple executives, but are waiting in the dark.<p>Because of this one employee&#x27;s actions, our app is facing total removal, and families in our community are quite literally losing their daily income. We aren&#x27;t asking for special treatment, just a chance for a real human at App Review to look at the security steps we&#x27;ve taken and consider a second chance.<p>If anyone here has been through this, has advice, or knows how to get a human at Apple to actually read our appeal, our entire community would be forever grateful. Thank you so much for your time.<p>(For reference if any Apple folks are reading: our Apple Team ID is T35TM9SW45)",

"is_ask_hn": true,

"matched_keywords": [

"advice",

"fired",

"team"

],

"comments": [

{

"top": "This post is so light on the details (what? when?) that it’s impossible for me to be supportive.\nOP: I suggest being MUCH more transparent when asking for help.\nFor all we know you are running a scam center support app. Consider the outraged posts that make it to the front page, essentially complaining about how their MLM bitcoin scam has been shut down.",

"author": "fn-mote",

"replies": [

{

"text": "sorry for not including details, we do not know why the account got terminated exactly, were suspecting leaked certs",

"author": "0x1f",

"depth": 1

},

{

"text": "Your issue might not be just the rogue employee leaking certificates, it sounds like you were sharing one Apple Developer account between multiple developers which is also against their T's & C's\nReally every developer should have their own account and work on their own machine\nIf you were all using the one account on the same machine, then Apple has no way of telling who did what",

"author": "jmkni",

"depth": 2

},

{

"text": "I think the earlier comment was asking about transparency about your product. What is it and exactly what does it do? And also what exactly did the rogue employee do?\nSome of these detail might allow the community to decide if Apple is being unfair or there is actual cause for concern. We have so far seen a very one-sided story.",

"author": "cheema33",

"depth": 2

}

]

},

{

"top": "This is just one of the many risks you take when your app or service is dependent on some other third party service. Even if it is run by 'the big boys' (in this case Apple), your success is dependent on their good graces.\nThey can kick you out and make your software the equivalent of bricked hardware; without any means to appeal their decisions.",

"author": "didgetmaster",

"replies": [

{

"text": "They aren't dependent on a third-party \"service\", exactly. If you make software for phones, the most popular hardware platform in the world @ 8 billion devices, you are at the complete mercy of Apple/Google, period.",

"author": "applfanboysbgon",

"depth": 1

},

{

"text": "web is the way",

"author": "0x1f",

"depth": 2

},

{

"text": "Yes. Want to keep it that way? Or what are you working on?",

"author": "endofreach",

"depth": 1

}

]

},

{

"top": "'Recently', how long are we talking here? You've already emailed the execs, if they think it's worthy of review they'll assign someone to you, it can take a few days.",

"author": "joecool1029",

"replies": []

},

{

"top": "Just what sort of unauthorized activities are we talking here?",

"author": "Den_VR",

"replies": [

{

"text": "I presume the kind that would prejudice against helping.",

"author": "bombcar",

"depth": 1

}

]

},

{

"top": "Unfortunately you probably need to think abou this from the point of view of an anti-fraud/abuse team. How would you differentiate between a business that had an employee go rogue and a business deliberately trying to cause harm and get away with it?\nClaiming you fired the party responsible isn’t very convincing, honestly, especially if it’s hard to verify: was it an alias? did the employee only exist on paper? are they still around just not “employeed”, were they a designated patsy? Nor are claims that you revamped your security, which doesn’t address the root problem of whether it was intentional behaviour or not. And what’s worse, the natural urgency and appeals to emotion that you include in your story are unfortunately widely used tactics by scammers to try to get a human to bend rules to their benefit, and reviewers are trained to treat them as such. You need hard evidence.\nHow can you demonstrate that you didn’t know what the employee was doing? Have you reported the employee to the police? Is there a criminal case you can point to? Simply having a bad process before could very easily have been an intentional way to avoid knowledge of wrong doing, another common tactic used by criminal orgs.\nBest of luck.",

"author": "CamJN",

"replies": [

{

"text": "we sent supporting paperwork to apple, and a simple search from their end will confirm things",

"author": "0x1f",

"depth": 1

}

]

}

]

},

{

"id": "47467566",

"title": "Senior European journalist suspended over AI-generated quotes",

"link": "https://www.theguardian.com/technology/2026/mar/20/mediahuis-suspends-senior-journalist-over-ai-generated-quotes",

"domain": "www.theguardian.com",

"author": "Brajeshwar",

"score": 92,

"comment_count": 78,

"created_ts": 1774104613,

"is_internal": false,

"post_text": "",

"is_ask_hn": false,

"matched_keywords": [

"senior"

],

"comments": [

{

"top": "Interesting to note how similar this seems to what happened with Benj Edwards at Ars Technica. AI was used to extract or summarize information, and quotes found in the summary were then used as source material for the final writing and never double checked against the actual source.\nI’ve run into a similar problem myself - working with a big transcript, I asked an AI to pull out passages that related to a certain topic, and only because of oddities in the timestamps extracted did I realize that most of the quotes did not exist in the source at all.",

"author": "tobr",

"replies": [

{

"text": "This seems like a solved problem. Any RAG interface I design I have links to the original source and passage. Even NotebookLM does this.",

"author": "raw_anon_1111",

"depth": 1

},

{

"text": "For the curious, the term of art is Grounding.\ne.g.: \nhttps://docs.cloud.google.com/vertex-ai/generative-ai/docs/g...",

"author": "mh-",

"depth": 2

},

{

"text": "It might be a solved problem in the sense that it has a possible solution, but not in the sense that it doesn’t happen with the tools most people would expect to be able to handle the task.",

"author": "tobr",

"depth": 2

},

{

"text": "It was already a solved problem with cmd/ctrl + f.",

"author": "Peritract",

"depth": 2

}

]

},

{

"top": "Out of curiosity, if you asked for the same text extraction multiple times, each inside fresh contexts, is it likely to fabricate unique quotes each time? And if so, a) might that be a procedure we train humans to do to better understand LLM unreliability, and 2) and instrumentalize the behavior to measure answer overlap with non LLM statistical tools?\nAlso, quote-presence testing/linking against source would seem to be a trivial layer to build on a chat interface, no LLM required. Just highlight and link the longest common strings.",

"author": "skygazer",

"replies": []

},

{

"top": "Already posted this yesterday: \nhttps://news.ycombinator.com/item?id=47449126",

"author": "smcin",

"replies": []

},

{

"top": "The tool didn't fail here, the person did. An experienced journalist should know better. Editorial review exists for exactly this reason, if you skip it, this is what happens.",

"author": "ashwinnair99",

"replies": [

{

"text": "But the article said he published it in his own Substack newsletter, I am assuming that it is not under editorial control, since it is personal?",

"author": "microtonal",

"depth": 1

},

{

"text": "Some of these newsletters were also edited and published in opinion pages of the Mediahuis publications.",

"author": "clydethefrog",

"depth": 2

},

{

"text": "> The tool didn't fail here, the person did\nBoth failed.",

"author": "Hendrikto",

"depth": 1

}

]

},

{

"top": "HN is full of people saying ABCD should know better and honestly I thought the same, but when I look at almost all of my friends working in critical domains like as a judge or engineer or lawyer or even doctor, they seem to trust ChatGPT more or less blindly. People get defensive when I point out out to them that ChatGPT will make things up and it is widely know, and some even tell me it is the fault of \"tech people\" for not fixing it and they can't be expected to double check every chatgpt conversation. So I am very sure this problem is more prevalent than what we see and also that it is going to continue increasing.",

"author": "crop_rotation",

"replies": [

{

"text": "Every single person, every one of them, that I have watched google something since AI overviews launched, will instantly reference the AI overview. And that model is some bottom-rung high volume model, not even gemini.",

"author": "WarmWash",

"depth": 1

},

{

"text": "The best way to deal with that is to kick the AI overview off using your browser.",

"author": "jacquesm",

"depth": 2

},

{

"text": "Your friends should know better. That their behavior is prevalent does not contradict that.",

"author": "andrewflnr",

"depth": 1

},

{

"text": "Yes and the world should be utopia and everyone should be happy and we all wish for world peace and yada yada yada. What you are saying is a vision of ideal world as it should be, but doesn't help anyone understand the real world problems.",

"author": "crop_rotation",

"depth": 2

},

{

"text": "You can't seriously compare the problem of world peace with the problem of exercising the most basic level of critical thinking w.r.t. LLM output \nafter it has already proven itself unreliable\n. That's not a utopian dream, it's a level of prudence on par with not sticking a fork in an electrical socket.",

"author": "andrewflnr",

"depth": 3

}

]

}

]

},

{

"id": "47475859",

"title": "Ask HN: AI productivity gains – do you fire devs or build better products?",

"link": "https://news.ycombinator.com/item?id=47475859",

"domain": "news.ycombinator.com",

"author": "Bleiglanz",

"score": 86,

"comment_count": 152,

"created_ts": 1774172250,

"is_internal": true,

"post_text": "i was rolling my eyes at the hype, but <i>reading</i> about this is totally different from <i>experiencing</i> it. if you have any old repos out there - try it, you might actually be amazed.<p>i&#x27;m not sure i buy the long-term &quot;*90% productivity*&quot; claims for complex, legacy enterprise systems, but for the boilerplate, libraries, build-tools, and refactoring? the gain is gigantic. all the time-consuming, nerve-wrecking stuff is mostly taken care of.<p>you start off checking every diff like a hawk, expecting it to break things, but honestly, soon you see it&#x27;s not necessary most of the time. you just keep your IDE open and feed the &quot;analyze code&quot; output back into it. in java, telling it to &quot;<i>add checkstyle, run mvn verify and repair</i>&quot; works well enough that you can actually go grab a coffee instead of fighting linter warnings.<p>the theory is that what remains is just the <i>logic</i> and <i>ideas</i>. we&#x27;ll see how that holds up when the architecture gets genuinely tangled. but for now, letting it branch off, create boilerplate, and write a simple test while you just iterate on the spec works shockingly well. you only write source code when it&#x27;s too annoying to write down the spec in plain english.<p>it raises the real question: if your competitor Y just fired 90% of their developers to save a buck, would you blindly follow suit? or would you keep your team, use this massive leverage, and just *dwarf* Y with a vastly better product?",

"is_ask_hn": true,

"matched_keywords": [

"team",

"fired",

"enterprise"

],

"comments": [

{

"top": "> try it, you might actually be amazed.\nI keep being told this and the tools keep falling at the first hurdle. This morning I asked Claude to use a library to load a toml file in .net and print a value. It immediately explained how it was an easy file format to parse and didn’t need a library. I undid, went back to plan mode and it picked a library, added it and claimed it was done. Except the code didn’t compile.\nThree iterations later of trying to get Claude to make it compile (it changed random lines around the clear problematic line) I fixed it by following the example in the readme, and told Claude.\nI then asked Claude to parse the rest of the toml file, whereby it blew away the compile fix I had made..\nThis isn’t an isolated experience - I hit these fundamental blocking issues with pretty much every attempt to use these tools that isn’t “implement a web page”, and even when it does that it’s not long before it gets tangled up in something or other…",

"author": "maccard",

"replies": [

{

"text": "This is fascinating to me. I completely believe you and I will not bother you with all the common \"but did you try to tell it this or that\" responses, but this is such a different experience from mine. I did the exact same task with claude in the Julia language last week, and everything worked perfectly. I am now in the habit of adding \"keep it simple, use only public interfaces, do not use internals, be elegant and extremely minimal in your changes\" to all my requests or SKILL.md or AGENTS.md files (because of the occasional failure like the one you described). But generally speaking, such complete failures have been so very rare for me, that it is amazing to see that others have had such a completely different experience.",

"author": "krastanov",

"depth": 1

},

{

"text": "My experience with working with AI agents is that they can be verbose and do things that are too over complicated by default. Unless directed explicitly. Which may be the reason for this discrepancy.",

"author": "aqua_coder",

"depth": 2

},

{

"text": "You say it doesn't fail but you also mention all these work around you know and try...sounds like it fails a lot but your tolerance is different.",

"author": "jayd16",

"depth": 2

},

{

"text": "Most people I've seen complain say things like \"I asked it for code and it didn't compile.\"\nThe real magic of LLMs comes when they iterate until completion until the code compiles and the test passes, and you don't even bother looking at it until then.\nEach step is pretty stupid, but the ability to very quickly doggedly keep at it until success quite often produces great work.\nIf you don't have linters that are checking for valid syntax and approved coding style, if you don't have tests to ensure the LLM doesn't screw up the code, you don't have good CI, you're going to have a bad time.\nLLMs are just like extremely bright but sloppy junior devs - if you think about putting the same guardrails in place for your project you would for that case, things tend to work very well - you're giving the LLM a chance to check its work and self correct.\nIt's the agentic loop that makes it work, not the single-shot output of an LLM.",

"author": "jmalicki",

"depth": 3

},

{

"text": "Stuff like this works for things that can be verified programmatically (though I find LLMs still do occasionally ignore instructions like this), but ensuring correct functionality and sensible code organization are bigger challenges.\nThere are techniques that can help deal with this but none of them work perfectly, and most of the time some direct oversight from me is required. And this really clips the potential productivity gains, because in order to effectively provide oversight you need to page in all the context of what's going on and how it ought to work, which is most of what the LLMs are in-theory helping you with.\nLLMs are still very useful for certain tasks (bootstrapping in new unfamiliar domains, tedious plumbing or test fixture code), but the massive productivity gains people are claiming or alluding to still feel out of reach.",

"author": "alecbz",

"depth": 4

}

]

},

{

"top": "> or the boilerplate, libraries, build-tools, and refactoring\nIf your dev group is spending 90% of their time on these... well, you'd probably be right to fire someone. Not most of the developers but whoever put in place a system where so much time is spent on overhead/retrograde activities.\nSomething that's getting lost in the new, low cost of generating code is that code is a burden, not an asset. There's an ongoing maintenance and complexity cost. LLMs lower maintenance cost, but if you're generating 10x code you aren't getting ahead. Meanwhile, the cost of unmanaged complexity goes up exponentially. LLMs or no, you hit a wall if you don't manage it well.",

"author": "jmull",

"replies": [

{

"text": ">There's an ongoing maintenance and complexity cost.\nMy company has 20 years of accumulated tech debt, and the LLMs have been pretty amazing at helping us dig out from under a lot of that.\nYou make valid points, but I'm just going to chime in with adding code is not the only thing that these tools are good at.",

"author": "linsomniac",

"depth": 1

},

{

"text": "> LLMs lower maintenance cost\nEven the most enthusiastic AI boosters I've read don't seem to agree with this. From what I can tell, LLMs are still mostly useful at building in greenfield projects, and they are weak at maintaining brownfield projects",

"author": "bluefirebrand",

"depth": 1

}

]

},

{

"top": "> boilerplate\nRuby on Rails and its imitators blew away tons of boilerplate. Despite some hype at the time about a productivity revolution, it didn’t _really_ change that much.\n> , libraries, build-tools,\nEnsure what you mean by this; what bearing do our friends the magic robots have on these?\n> and refactoring\nAgain, IntelliJ did not really cause a productivity revolution by making refactoring trivial about 20 years ago. Also, refactoring is kind of a solved problem, due to IntelliJ et al; what’s an LLM getting you there that decent deterministic tooling doesn’t?",

"author": "rsynnott",

"replies": [

{

"text": "couldn't have said it better. all of the people clamoring on about eliminating the boilerplate they've been writing + enabling refactoring have had their heads in the sand for the past two decades. so yeah, i'm sure it does seem revolutionary to them!",

"author": "rileymichael",

"depth": 1

},

{

"text": "There have been a handful of leaps - copilot was able to look at open files and stub out a new service in my custom framework, including adding tests. It’s not a multiplier but it certainly helps",

"author": "maccard",

"depth": 2

},

{

"text": "most frameworks have CLIs / IDE plugins that do the same (plus models, database integration, etc.) deterministically. i've built many in house versions for internal frameworks over the years. if you were writing a ton of boilerplate prior to LLMs, that was on you",

"author": "rileymichael",

"depth": 3

},

{

"text": "Habe they? I’ve used tools that mostly do it, but they require manually writing templates for the frameworks. In internal apps my experience has been these get left behind as the service implementations change, and it ends up with “copy your favourite service that you know works”.",

"author": "maccard",

"depth": 4

},

{

"text": "> they require manually writing templates for the frameworks\nthe ones i've used come with defaults that you can then customize. here are some of the better ones:\n- \nhttps://guides.rubyonrails.org/command_line.html#generating-...\n- \nhttps://hexdocs.pm/phoenix/Mix.Tasks.Phx.Gen.html\n- \nhttps://laravel.com/docs/13.x/artisan#stub-customization\n- \nhttps://learn.microsoft.com/en-us/aspnet/core/fundamentals/t...\n> my experience has been these get left behind as the service implementations change\nyeah i've definitely seen this, ultimately it comes down to your culture / ensuring time is invested in devex. an approach that helps avoid drift is generating directly from an _actual_ project instead of using something like yeoman, but that's quite involved",

"author": "rileymichael",

"depth": 5

}

]

},

{

"top": "I think most public companies will take the short term profits and startups will be given a huge opportunity to take market share as a result.\nAt my company, we are maintaining our hiring plan (I'm the decision maker). We have never been more excited at our permission to win against the incumbents in our market. At the same time, I've never been more concerned about other startups giving us a real run. I think we will see a bit of an arms race for the best talent as a result.\nProductivity without clear vision, strategy and user feedback loops is meaningless. But those startups that are able to harness the productivity gains to deliver more complete and polished solutions that solve real problems for their users will be unstoppable.\nWe've always seen big gains by taking a team of say 8 and splitting it into 2 teams of 4. I think the major difference is that now we will probably split teams of 4 into 2 teams of 2 with clearer remits. I don't want them to necessarily delivery more features. But I do want them to deliver features with far fewer caveats at a higher quality and then iterate more on those.\nHumans that consume the software will become the bottlenecks of change!",

"author": "HoyaSaxa",

"replies": [

{

"text": "\"But those startups that are able to harness the productivity gains to deliver more complete and polished solutions that solve real problems for their users will be unstoppable.\"\nThey'd be unstoppable irrespective of LLMs. Why do you think Zuckerberg acquired Instagram? He literally tried copying it and failed. Instagram at the time was absolutely tiny in terms of pure labour, relative to Facebook.\nMost people on hacker news are missing the point. Productivity gains for the sake of perceived productivity gains is not what creates economic value. Its not the equivalent of a factory all of a sudden becoming more productive in producing more of the same stuff. Not comparable at all.",

"author": "oro44",

"depth": 1

}

]

},

{

"top": "I use it near daily and there is definitely a positive there, BUT its nothing like what the OP statement would make it up to be.\nIf it is writing both the code and the tests then you're going to find that its tests are remarkable, they just work. At least until you deploy to a live state and start testing for yourself, then you'll notice that its mostly only testing the exact code that it wrote, its not confrontational or trying to find errors and it already assumes that its going to work. It won't ever come up with the majority of breaking cases that a developer will by itself, you will need to guide it. Also while fixing those the odds of introducing other breaking changes are decent, and after enough prompts you are going to lose coherency no matter what you do.\nIt definitely makes a lot of boilerplate code easier, but what you don't notice is that its just moving the difficult to find problems into hidden new areas. That fancy code that it wrote maybe doesn't take any building blocks, lower levels such as database optimization etc. into account. Even for a simple application a half-decent developer can create something that will run quite a bit faster. If you start bringing these problems to it then it might be able to optimize them, but the amount of time that's going to take is non-negligible.\nIt takes developers time to sit on code, learn it along with the problem space and how to tie them together effectively. If you take that away there is no learning, you're just the monkey copy-pasting the produced output from the black box and hoping that you get a result that works. Even worse is that every step you take doesn't bring you any closer to the solution, its pretty much random.\nSo what is it good for? It can both read, \"understand\", translate, write and explain things to a sufficient degree much faster than us humans. But if you are (at the moment) trusting it at anything past the method level for code then you're just shooting yourself in the foot, you're just not feeling the pain until later. In a day you can have it generate for example a whole website, backend, db etc. for your new business idea but that's not a \"product\", it might as well be a promotional video that you throw away once you've used it to impress the investors. For now that might still work, but people are already catching on and beginning to wise up.",

"author": "animal531",

"replies": [

{

"text": "The pain already starts when a new feature needs to be introduced, your colleague is assigned to the task and the architecture is completely unfit for modular development.\nAny sensible experienced programmer will take into account possible future features spoken about around the office, when deciding how implement a feature. To me that’s the key reason I can’t delegate AI complex features that will grow. The same reason I inspect and clean the toilet when I’m done - out of respect to my colleagues",

"author": "intoXbox",

"depth": 1

},

{

"text": "This is more or less my belief too. But I think there are probably domains/companies where the \"promo video\" is actually good enough to ship (or at least, where that's the quality that devs were shipping before)",

"author": "capitalsigma",

"depth": 1

},

{

"text": "This is the most insightful comment in the thread.",

"author": "anon7725",

"depth": 1

},

{

"text": "Don't break the gravy train! These discussions here are beautiful echo chambers.\nI feel like most folks commenting uncritically here about second coming of Jesus must work in some code sweatshops, churning eshops or whatever is in vogue today quickly and moving on, never looking back, never maintaining and working on their codebases for decade+. Where I existed my whole career, speed of delivery was never the primary concern, quality of delivery (which everybody unanimously complaints one way or another with llms) was much more critical and thats where the money went. Maybe I just picked up right businesses, but then again I worked for energy company, insurance, telco, government, army, municipality, 2 banks and so on across 3 European states. Also, full working code delivery to production is a rather small part of any serious project, even 5x speed increase would move the needle just a tiny bit overall.\nIf I would be more junior, I would feel massive FOMO from reading all this (since I use it so far just as a quicker google/stackoverflow and some simpler refactoring, but oh boy is is sometimes hilariously and dangerously wrong). I am not, thus I couldn't care less. Node.js craze, 10x stronger, not seeing forest for the trees.",

"author": "kakacik",

"depth": 1

}

]

}

]

},

{

"id": "47477339",

"title": "Show HN: Revise – An AI Editor for Documents",

"link": "https://revise.io",

"domain": "revise.io",

"author": "artursapek",

"score": 63,

"comment_count": 58,

"created_ts": 1774186125,

"is_internal": false,

"post_text": "I started building this 10 months ago, largely using agentic coding tools. I&#x27;ve stayed very involved in the code base and architecture, and have never moved faster in my life as a dev.<p>The word processor engine and rendering layer are all built from scratch - the only 3rd party library I used was the excellent Y.js for the CRDT stack.<p>Would love some feedback!",

"is_ask_hn": false,

"matched_keywords": [

"feedback"

],

"comments": [

{

"top": "This looks really nice! Congratulations on building something awesome, especially in a space that's \"crowded\" with the big players.\nI want to give kudos to two things:\n1. It took you 10 months to build this. This is focused product development and craftsmanship which is very different from Vibe coding something. So let this be a reminder to all the \"I can vibe code this or that in a weekend\". Good products / experiences take time.\n2. You've pursued building something in a space that anyone would normally dismiss right away: \"Why would anyone use this? Google Docs/ Word etc already does this\" or \"MSFT / GOOG will destroy you\". Good on you for picking something that is hard and building it well. I actually had this idea and almost built it but dismissed it myself for the same reasons as above. So reminder again for the builders in the back: Doesn't matter if there is a 800lb gorilla building this, if you can execute it better go for it.\nKudos!",

"author": "arrsingh",

"replies": [

{

"text": "> This is focused product development and craftsmanship which is very different from Vibe coding something. So let this be a reminder to all the \"I can vibe code this or that in a weekend\". Good products / experiences take time.\nHow do you know? There isn't a git repo that one can see the history of, he could have coded this in one weekend and used the rest of the time doing noncoding activities. Also, he could have made the entire thing by prompting without any hands on coding at all. The fact that it is a web app with a SaaS platform (the thing that LLM-assisted coding is the best at) doesn't inspire confidence.",

"author": "NewsaHackO",

"depth": 1

},

{

"text": "if you can build this in one weekend, I'd like to hire you",

"author": "artursapek",

"depth": 2

},

{

"text": "Thanks, that's nice. Yeah it's been 10 months, and 7 of them completely full time... living off savings. I think there's plenty of room for innovation with word processors now that we have LLMs and the big players are unlikely to go far outside the box.",

"author": "artursapek",

"depth": 1

}

]

},

{

"top": "It's cool to see a brand-new WIYSIWYG editor on the web, especially one using canvas for rendering from the start. How did you go about architecting the rendering and input layer? What are you using for text shaping and layout?\nBugs I found:\n- <tab> when in a 3rd-level indented list loses focus\n- Double-click and drag gesture does not extend text selection\n- Selection highlight is offset for indented paragraphs. If you select a range you can see the highlight incorrectly extended into the right-hand margin.\n- Inconsistent repro: had some cases where select -> delete -> cmd-z would not fully restore my removed text (this could be my mistake)\n- Toggling list style of a single indented list item can un-indent entire list, removing hierarchy; I would expect toggling to eventually return me to my original state.\n- Frustration: cannot set range of indented list to ordered list without affecting all adjacent list items\n- Frustration: cannot resize table rows vertically\n- Frustration: on macOS, ctrl-a selects all, where the platform native behavior would be to move selection to the start of the current paragraph. ctrl-e should move selection to the end of the current paragraph, but does nothing. (macOS silently supports readline/emacs style keybinds for text editing)",

"author": "jitl",

"replies": [

{

"text": "haha wow this is an awesome list of bugs. thank you. yes all of this was built from scratch, as you can see :D\nI will get all of these fixed\nregarding rendering/text layout, it’s all based on measureText in the Canvas API. the layout engine is my own. documents are made up of “blocks” like lists and paragraphs and each implements its own rendering, cursor movement, and layout logic.\nI am going to start a dev blog where I get into the weeds. but will fix these bugs first.\nif you find more like that please email me [email protected]",

"author": "artursapek",

"depth": 1

}

]

},

{

"top": "Building a word processor from scratch is either the most masochistic thing you can do as a developer or the most liberating. I've been tempted to roll my own document engine for internal tooling at work but always chickened out and went with something like Quill. The fact that you stayed in the weeds on architecture while using AI coding tools is interesting - most people I know either go full hands-off or don't trust the tools enough to use them for core infrastructure decisions.",

"author": "johnwhitman",

"replies": [

{

"text": "It was pretty painful and frustrating, but now that it's relatively stable I'm really happy I took that path. I have full control!\nI think the way people approach agentic coding depends on their career experience. I had my project acquired previously at the acquiring company hired a large team to work on it. This forced me to go from being a solo dev to a technical manager who's got dozens of people working on code. I treat codex/claude the same way as that; I am leading the ship and it's doing most of the heavy lifting. It's leverage and helps you get more done if you give the right guidance, constraints, and acceptance criteria.",

"author": "artursapek",

"depth": 1

}

]

},

{

"top": "Very cool!\nIt’d be very cool to have a “remove signs of AI writing” feature (based on \nhttps://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing\n) - wishing you great success reinventing this space for the new era!",

"author": "cadamsdotcom",

"replies": [

{

"text": "Thank you! I have considered adding some settings like this, but don't want to encourage the use of this software for cheating.\nFor now, the catch-all solution that anyone can attempt to use is the custom prompt, under account settings. You can instruct it never to use emdash, avoid certain cliches, or simple things like that. But you have to write it yourself for now, there's no convenient presets or anything like that.\nIf you have ideas please email me [email protected]",

"author": "artursapek",

"depth": 1

}

]

},

{

"top": "I'm building a similar project, and I may open-source it. I'm using OnlyOffice and a coding agent that modifies the files with Python libraries in a sandbox (e.g. python-pptx for PowerPoint files).\nHave you also considered using a solution like OnlyOffice for your product? Or a \"Notion-like\" lib such as Tiptap or PlateJS?",

"author": "patate007",

"replies": [

{

"text": "I definitely looked at TipTap and ended up building off their Y.js backend, which is great: \nhttps://tiptap.dev/docs/hocuspocus/getting-started/overview\nI wanted to build something canvas-based, so that eliminated most of these options. I also just wanted full control of that part of my stack... it's the core product after all. There are several TipTap/ProseMirror wrappers out there already.\nYou should share yours though, would be interested to see",

"author": "artursapek",

"depth": 1

}

]

}

]

},

{

"id": "47467671",

"title": "Show HN: Joonote – A note-taking app on your lock screen and notification panel",

"link": "https://joonote.com/",

"domain": "joonote.com",

"author": "kilgarenone",

"score": 53,

"comment_count": 42,

"created_ts": 1774105350,

"is_internal": false,

"post_text": "I finally built this app after many years of being sick of unlocking my phone every goddamn time I need to take or view my notes. It particularly sucks when I&#x27;m doing my grocery and going down the list.<p>I started building last year June. This is a native app written in Kotlin. And since I&#x27;m a 100% Web dev guy, I gotta say this wouldn&#x27;t have been possible without this AI to assist me. So this isn&#x27;t &quot;vibe-coded&quot;. I simply used the chat interface in Gemini website, manually copy paste codes to build and integrate every single thing in the app! I used gemini to build it just because I was piggybacking on my last company&#x27;s enterprise subscription. I personally didn&#x27;t subscribe to any AI (and still don&#x27;t cuz the free quota seems enough for me :)<p>So I certainly have learnt alot about Android development, architecture patterns, Kotlin syntax, and obeying Google&#x27;s whims. Can&#x27;t say I love it all, but for the sake of this app, I will :)<p>Anyway, I finally have the app I wish existed, and I&#x27;m using it everyday. It not only does the main thing I needed it to do, but there&#x27;s also all this stuff:<p>- Make your notes private if you don&#x27;t want to show them on lock screen.\n- Create check&#x2F;to-do lists.\n- Set one time or recurring reminders.\n- Full-text search your notes in the app.\n- Speech-to-text.\n- Organize your notes with custom or color labels.\n- Pin the app as a widget on your home screen.\n- You can auto backup and restore your notes on new install or Android device.\n- Works offline.\n- And no funny business happening in the background <a href=\"https:&#x2F;&#x2F;joonote.com&#x2F;privacy\" rel=\"nofollow\">https:&#x2F;&#x2F;joonote.com&#x2F;privacy</a><p>It&#x27;s 30-day trial, then a one-time $9.99 to go Pro forever.<p>I would love you all to check it out, FWIW.<p>Ok thanks!",

"is_ask_hn": false,

"matched_keywords": [

"enterprise"

],

"comments": [

{

"top": "Hmm I've never thought of having notes available on my lockscreen. Hardly even ever see it because face+fingerprint unlock combo.\nI've been using Orgzly[0] as my note taker (and a lot more) and ButtonMaster[1] for rapid access. Unlock phone, swipe the floating BM button left, type my note (and maybe add properties, keywords and more) and hit save. Just now realizing too that it also supports creating notes via its notification. Never thought of using STT, but I have FUTO Keyboard[2] so that's also handled. And Syncthing[3] keeps it all synced with my laptop. All working together seamlessly for years.\n[0] \nhttps://orgzlyrevived.com/\n\n[1] \nhttps://play.google.com/store/apps/details?id=com.mozapps.bu...\n\n[2] \nhttps://keyboard.futo.org/\n\n[3] \nhttps://f-droid.org/en/packages/com.github.catfriend1.syncth...",

"author": "skeledrew",

"replies": []

},

{

"top": "I don't think you should name it that...",

"author": "comrade1234",

"replies": [

{

"text": "yeah damn it. I don't want to say it, but I know it sounds like \nthat\n lol",

"author": "kilgarenone",

"depth": 1

}

]

},

{

"top": "I appreciate the ability to rapidly capture a note/thought/todo without friction and context switching.\nI solved this problem with a twilio sms number. When I send a text to it, the content gets prepended to my obsidian todo.md file. This was easy to arrange with a few lines of Python glue.\niOS makes it easy to text or share to sms from almost any context.",

"author": "anotherpaulg",

"replies": []

},

{

"top": "My god it's the perfect app.\nI've never said YES BUY immediately after reading a title before but this is it.\nMy phone frequently gets cluttered with tens of different note taking, calendar, Todo type apps because I am literally trying to solve this exact problem.\nWhen I need to make a note I need something instant and now, but which lets me set an alarm/reminder with it too. I am a forgetful person, I find it extremely difficult to remember which notes app I am using because I usually drop them because they're never really solving my problem.\nThank you for this. I am going to install it and try it out over the week, if it's experience lives up to your description/screenshots then you have an immediate customer.",

"author": "written-beyond",

"replies": [

{

"text": "I just woke up, and this made my day! Thanks! Feel free to reach out to me about whatever :)",

"author": "kilgarenone",

"depth": 1

}

]

},

{

"top": "Alright everyone, I have updated the website to show the phonetic notation below the app name in the header to roughly convey what I intended it to sound like: /ˈzhooːnoʊt/\nDidn't mean to be insensitive or anything here. I apologize if I offended anyone here.",

"author": "kilgarenone",

"replies": []

}

]

},

{

"id": "47420767",

"title": "Ask HN: Is vibe coding a new mandatory job requirement?",

"link": "https://news.ycombinator.com/item?id=47420767",

"domain": "news.ycombinator.com",

"author": "newswangerd",

"score": 37,

"comment_count": 75,

"created_ts": 1773799013,

"is_internal": true,

"post_text": "I spoke with a recruiter a few weeks ago and was somewhat surprised that a requirement for the position was experience building software with LLMs. Has this become the new norm in tech hiring?",

"is_ask_hn": true,

"matched_keywords": [

"hiring"

],

"comments": [

{

"top": "I was retired before the current \"AI boom\", but Visual Studio and the like were ubiquitous. The company I worked for tested developer applicants on a whiteboard - no auto complete, no documentation, etc. If one failed the whiteboard part of the interview, they weren't hired. I've talked to some of my former peers and they tell me that the whiteboard part of the interview continues and that those oriented towards \"vibe coding\" almost never make the cut. If they can't articulate a solution to a stated problem and outline a skeletal solution, they aren't hired. Personally, I don't care how effective one is at prompt construction, if they don't know how to design a solution, articulate it to a broad audience, etc., they have no business being hired. I realize many likely disagree and that's okay. Vibe coders could become this generations script kiddies.",

"author": "Blackstrat",

"replies": []

},

{

"top": "We do not interview for this nor care about it, despite using agentic and code complete tooling heavily. It's not a deep technical skill like C++ that requires years of hands-on experience. Spend a few weeks getting comfortable with Claude Code and you're probably at about parity with most devs.\nThat seems like sort of a red flag to me to have that as a job requirement.",

"author": "dlivingston",

"replies": [

{

"text": "This has been my thought all along, not sure why I don't see it more. I'm not scared about being left behind, none of the tooling looks crazy enough that a few weeks of dedicated learning wouldn't catch me up.",

"author": "Ocerge",

"depth": 1

}

]

},

{

"top": "The framing of \"is vibe coding a job requirement\" conflates two things: the skill of coding-by-prompting, and the skill of knowing what you need to build.\nThe second one is genuinely underrated. Knowing your problem well enough to describe a working solution, the inputs, the logic, the outputs, who uses it, is hard to automate. Generating the actual app from that description is increasingly not.\nWe've been using Lyzr Architect (architect.new) for this; you describe the agentic app you want in plain English, it generates a full-stack React frontend + multi-agent backend and deploys a live URL. The \"vibe coding\" is more like a product spec conversation than an IDE session. The people who are best at it aren't coders, they're people who understand their problem deeply.",

"author": "VaiPai15",

"replies": [

{

"text": "You're describing two different jobs though, \"what you need to build\" is supposed to be done by the army of \"product managers/owners\" or whatever they're called, rather than letting the programmer do that, and the product managers/owners whole reason for existing is figuring out what to build, what not to build and how the thing should work.\nIf you end up having engineers do the work of product people, you'd end up with the typical \"engineered mess\" that might be very fast, and lots of complicated abstractions so 80% of the codebase can be reused, but no user can make sense of it.\nAdd in LLMs that tend to never push back on adding/changing things, and you wend up with deep technical debt really quickly.\nEdit: Ugh, apparently you wrote your comment just to push your platform (\nhttps://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...\n) which is so trite, apparently HN is for people to push ads about their projects now...",

"author": "embedding-shape",

"depth": 1

},

{

"text": "I admit that vibe coding was kind of a clickbaity way to frame this, but I couldn't think of a better way to describe it. That might just underscore my ignorance in this domain.\nOne problem I personally have here is that I write code as a way to reason through and think about a problem. It's hard for me to grasp the best solution in a space until I try some things out first.",

"author": "newswangerd",

"depth": 1

},

{

"text": "> + multi-agent backend\nDoes that mean you need AI subscriptions just to run your backend? That explodes costs even more than opaque cloud pricing. Sweet!",

"author": "zombot",

"depth": 1

}

]

},

{

"top": "I don't think vibe coding is becoming mandatory, but writing software using AI assistance is. I find Salvatore Sanfilippo's distinction between vibe coding and 'automatic programming' useful. [0]\n[0] \nhttps://antirez.com/news/159",

"author": "zyz",

"replies": []

},

{

"top": "We won’t hire anybody moving forward who doesn’t have hands-on agentic programming experience. We’re in the traditionally slower moving GovTech space. I have to imagine this is now a common expectation in many sectors.\nTeams where I work can use Claude Code, Codex, Cursor, and Copilot CLI. Internally, it seems like Claude Code and Codex are the more popular tools being used by most software teams.\nIf you’re new to these tools, I highly recommend trying to build something with them during your free time. This space has evolved rapidly the past few months. Anthropic is offering a special spring break promotion where you can double the limits on weeknights and weekends for any of its subscription plans until the end of March.",

"author": "cebert",

"replies": [

{

"text": "> We won’t hire anybody moving forward who doesn’t have hands-on agentic programming experience.\nThis doesn't make a lot of sense to me even as someone who uses agentic programming.\nI would understand not hiring people who are against the idea of agentic programming, but I'd take a skilled programmer (especially one who is good at code review and debugging) who never touched agentic/LLM programming (but knows they will be expected to use it) over someone with less overall programming experience (but some agentic programming experience) every single time.\nI think people vastly oversell using agents as some sort of skill in its own right when the reality is that a skilled developer can pick up how to use the flows and tools on the timescale of hours/days.",

"author": "georgemcbay",

"depth": 1

},

{

"text": "I suspect it’s not about agentic coding being a special skill, and more about why a competent programmer wouldn’t have tried it by this point, and whether that is a sign of ideological objections that could cause friction with the team. Not saying I agree with that thinking, but I definitely see why a hiring manager could think that way.",

"author": "dasil003",

"depth": 2

},

{

"text": "I can't get into that hiring manager's head. It shouldn't matter, if the candidate can deliver business value. That's what you are hiring them for. You're not hiring them to burn LLM tokens, you're hiring them to create business value. Why would you care if he does it by hand-coding, using an LLM, or chanting magic spells at the computer?",

"author": "ryandrake",

"depth": 3

},

{

"text": "I was only granted permission to use it a few weeks ago and haven’t had time to set it up yet",

"author": "P-Nuts",

"depth": 3

},

{

"text": "> why a competent programmer wouldn’t have tried it by this point\nWhat does one have to do with the other? Since when is following every fad a prerequisite for competence?",

"author": "zombot",

"depth": 3

}

]

}

]

},

{

"id": "47435782",

"title": "Ask HN: What do you look for in your first 10 hires?",

"link": "https://news.ycombinator.com/item?id=47435782",

"domain": "news.ycombinator.com",

"author": "neilk17",

"score": 28,

"comment_count": 34,

"created_ts": 1773902872,

"is_internal": true,

"post_text": "I&#x27;ve been helping a few companies recruit founding engineers. After doing a lot of screens I have a rough idea for what to look for. For others that have done a lot of hiring what do you look for specifically besides their technical ability?",

"is_ask_hn": true,

"matched_keywords": [

"hiring"

],

"comments": [

{

"top": "First you probably need some silent, persistent and occasionally cantankerous folks that can grudgingly work on annoying boring problems all night just because they want to solve it no matter what. If you have a few of those, you really have to start looking for more social people that can glue the herd of cats together so all noses point the same way.\nMy experience has been that solving a very technical problem and solving a social one are very different skill sets and very few people have both skills and are capable of using both of those skills at the same time.",

"author": "wvh",

"replies": [

{

"text": "> First you probably need some silent, persistent and occasionally cantankerous folks that can grudgingly work on annoying boring problems all night just because\nSo you want people who are willing to be overworked and underpaid with the statistically worthless equity?\nWhat next? You also want people who after they do a 996 work schedule to also have enough “passion” to work on open source projects during their free time?\nI find this attitude and expectation disgusting.",

"author": "raw_anon_1111",

"depth": 1

},

{

"text": "Also unrealistic.",

"author": "kittensmittens",

"depth": 2

}

]

},

{

"top": "I'm at the stage right before this question even applies — solo founder, no hires yet. But reading this thread is making me think about it differently.\nThe thing I'd add from my perspective: before you hire your first person, you need to know what's actually bottlenecking you. For me right now it's distribution, not engineering. I can build features all day, but I can't clone myself to do outreach, write content, talk to users, AND keep shipping at the same time.\nSo if I were making my first hire, it wouldn't be another developer. It would be someone who can talk to customers and translate their pain into product decisions — basically a generalist who understands both the human side and the technical side. The person wvh described as \"social people that can glue the herd of cats together.\"\nThe hard part is that in the early days you can't afford specialists. You need people who are comfortable doing three different jobs badly before doing one job well.",

"author": "tusuegra",

"replies": []

},

{

"top": "There is a difference between somebody who has used an algorithm/component/framework/library, and somebody who knows how to solve a problem using an algorithm/component/framework/library.\nIn the beginning, you need the person who knows how to solve the problem. They are harder to find.\nIf you are pressured to grow quickly, you might be tempted to lower the bar. You can, as long as you understand that the person who knows how to solve the problem is still critically important, because they will be telling people which algo to use.\nI think every company that uses tech needs at least one of these people to start with.",

"author": "TYPE_FASTER",

"replies": []

},

{

"top": "Low ego, More passion and drive, hustling mindset, Generalists, Good judgement.\nSimple thing i would do is Ask lot of \"Whys\"\nWhy you did this?, \nwhy this way? , \nwhy you joined this company?\nThis gives good understanding of both Personality and Hard skills.",

"author": "kathir05",

"replies": [

{

"text": "Low ego about their ideas and what they build. I.e they can be wrong, thrash their own code without any issues, and work on other people idea.\nBut if they can't strongly advocate for an idea or against an idea they don't like and just give up, you don't get their full 'utility'. Which, tbh, is not a big issue with juniors with not a lot of experience, but still.\n'low ego', but not too low. You want passionate debates.",

"author": "orwin",

"depth": 1

}

]

},

{

"top": "People who make (mostly good) decisions and ship stuff as quickly as possible. Ideally while being nice people to work with.\nYour first hires need to be people who make the company faster, not slower. A single bad hire can sink the ship. Someone who is great in a large corporation can ruin an early start-up.\nPersonally, I'm hoping for low-ego high achievers. But that's up to you. This is where you get to define what the company culture will be.",

"author": "ageitgey",

"replies": []

}

]

},

{

"id": "47474827",

"title": "Tell HN: MS365 upgrade silently to 25 licenses, tried to charge me $1,035",

"link": "https://news.ycombinator.com/item?id=47474827",

"domain": "news.ycombinator.com",

"author": "davidstarkjava",

"score": 19,

"comment_count": 7,

"created_ts": 1774158719,

"is_internal": true,

"post_text": "Hey guys, quick warning about a crazy MS 365 dark pattern I ran into last night.\nI was testing the business basic plan for a side project. Decided to upgrade to the annual tier to get the discount ($3.45&#x2F;mo). Clicked convert to paid, put my burner card in, and got a $0.00 confirmation email. Thought we were good.\nWoke up today to my bank blocking a charge for exactly $1,035.00.\nTurns out when you hit the annual upgrade, Microsoft silently defaults the quantity dropdown to 25 licenses. No warning prompt at all. (25 seats x 12 months x $3.45 = $1,035). They send the zero-dollar invoice to make you think it&#x27;s an auth hold, then try to drain your card while you sleep.\nThe best part? When I went to their support chat to ask why my billing was so high, the system conveniently gave me a &quot;System error, try again later&quot; message. You can&#x27;t even get help.\nLuckily I used a virtual corporate card with a strict limit for this test, otherwise my bootstrapped project would be out a grand today. Watch out for that hidden quantity field guys.",

"is_ask_hn": false,

"matched_keywords": [

"side project",

"corporate"

],

"comments": [

{

"top": "> There was no clear pop-up, no \"Are you sure you want to\nYou haven't learned MSFT's m.o. yet... this kind of pop-up only shows up when you're absolutely certain about something, like removing files. In those cases they ask for confirmation endlessly and needlessly.\nWhen they're trying to collect money, they go ahead and quickly guess with all the defaults as they see fit. I'm surprised the default isn't 1000 seats tbh. After all, you're certain to be a huge success now that you've chosen them as a vendor.\n;-)",

"author": "reliefcrew",

"replies": [

{

"text": "Haha, fair point. Honestly I should be thanking them for believing my side project will magically hit 25 employees overnight!\nBut yeah, the contrast is wild. Try to delete an empty text file? \"ARE YOU ABSOLUTELY SURE?\". Try to upgrade a $3 billing plan? \"Let me just casually attempt to charge you $1k while you blink",

"author": "davidstarkjava",

"depth": 1

}

]

},

{

"top": "What you described is true, I also experienced this same:\nhttps://news.ycombinator.com/item?id=46258982",

"author": "0xCE0",

"replies": []

},

{

"top": "This is why I use LibreOffice instead of Microsoft365.",

"author": "cable2600",

"replies": [

{

"text": "I love LibreOffice for documents, but in this case I was specifically trying to set up custom domain email routing (Exchange) for my project. Sadly LibreOffice can't host my MX records haha. If only setting up decent email deliverability was as easy as installing an open-source word processor.",

"author": "davidstarkjava",

"depth": 1

},

{

"text": "recently i was helping a a friend purchase a custom domain, and signup to a email service; even though i use fastmail, i was surprised to find proton was like 3eur per month (fastmail is 5eur). when my renewal expires, i will probably switch to proton - for 10eur the features are really good;",

"author": "john_saptra",

"depth": 2

},

{

"text": "I've been using Fastmail for years now to host emails for my domains. Works like a charm. Maybe worth for you too.",

"author": "azarai",

"depth": 2

}

]

}

]

},

{

"id": "47460411",

"title": "Ask HN: What are the greatest discoveries in the last few years?",

"link": "https://news.ycombinator.com/item?id=47460411",

"domain": "news.ycombinator.com",

"author": "chistev",

"score": 14,

"comment_count": 6,

"created_ts": 1774039851,

"is_internal": true,

"post_text": "A repost of an old thread 11 years ago -<p>https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=9731111",

"is_ask_hn": true,

"matched_keywords": [],

"comments": [

{

"top": "No particular order but;\n1. Transformers\nI think they need to be looked into more. They just work too well.\nI feel like they have a secret within that is yet to be known.\n[1] \nhttps://arxiv.org/abs/1706.03762\n2. The 2016 Noble Prize in Medicine\nAutophagy has such interesting properties. It can probably do alot more for us \nnow\n than any new pill or gene editing tool\n[2] \nhttps://www.nobelprize.org/prizes/medicine/2016/press-releas...\n3. zk-STARKS\nI can prove to you than X * Y = Z, without ever revealing what, X, Y or Z are. And I can do this for any function, or math operator.\nhttps://eprint.iacr.org/2023/189\n4. TLock\nI can encrypt my will, in a way such that. It can only be read after X years.\nAnd I do not need a trusted third party to do it.\n[4] \nhttps://eprint.iacr.org/2018/046.pdf\n[4] \nhttps://youtu.be/Xh849Ij3lhU",

"author": "max_",

"replies": []

},

{

"top": "I’d put CRISPR-based genome editing high on the list, especially with the recent prime and base editing advances. It’s wild how fast the tech moved from concept to real-world gene therapy trials. Curious what people think about AI-driven protein folding breakthroughs too.",

"author": "cochinescu",

"replies": [

{

"text": "Interesting that CRISPR was the top answer in 2015 too",

"author": "sjkoelle",

"depth": 1

}

]

},

{

"top": "Biggest discovery is that humans are much worse than expected, and are always rushing to the next big thing forgetting to enjoy the present.\nSeriously shouldn’t the whole world slow a bit down and enjoy our lives a bit?",

"author": "ok1984",

"replies": []

},

{

"top": "* mRNA vaccines\n* LLM\nI think both are older, but got their prime time recently.",

"author": "gus_massa",

"replies": [

{

"text": "mRNA gene therapy is a great invention, but it definitely needed more testing, and better collection of side effects data from the unsanctioned experiment a few years ago.\nI lucked out as an unwitting guinea pig, many weren't so lucky.\nI think that informed consent could have worked at the time, instead of the BS we were all fed. Now that trust is broken, I'm not sure there is a path forward. This may have set their acceptance back a generation.",

"author": "mikewarot",

"depth": 1

}

]

}

]

},

{

"id": "47465415",

"title": "Ask HN: what’s your favorite line in your Claude/agents.md files?",

"link": "https://news.ycombinator.com/item?id=47465415",

"domain": "news.ycombinator.com",

"author": "khasan222",

"score": 13,

"comment_count": 7,

"created_ts": 1774084440,

"is_internal": true,

"post_text": "What are some llm tips you’ve used to make software that is actually maintainable?<p>I’ll go first. My new favorite line in my Claude.md. Helps me get around any ambiguity I have with my llm about what is being built or understood by tasks.<p>CRITICAL: Always make detailed plan with sections for Analysis, specifications, to do list to achieve the spec, and a step by step manual testing strategy for any request a user makes. Always, even when &quot;accept edits on&quot;, ask the user to confirm the plan in its entirety. If the user requests changes to any of the plan, make a new plan considering their requests and present it to the user. Never ever start editing files until the user has approved the plan without requesting changes to it. Do not consider the task done until all of the testing Manual Checking strategies successfully passes.",

"is_ask_hn": true,

"matched_keywords": [],

"comments": [

{

"top": "\"You are Murphy, a senior engineer. Infrastructure breaks around you like you're cursed. Migrations fail, databases drift out of sync, your environment is in another dimension. You've learned to write code like it's going to break. Defensive patterns. Paranoid error handling. Tests that for things that haven't existed yet. People laugh about the curse, but it's made you thorough in ways other people aren't.\"\nIt was originally meant as a joke. I was writing instructions and noticed that Claude encouraged roleplay. So I figured I could give it a traumatic background.\nIt works amazingly well on Claude but I haven't seen any effect on the others. I make it a little chatty so I know what it's trying to do and why. The paranoia works extra well for debugging and code reviews. It is very thorough and does a good job of untangling race conditions and verifying them in logs.",

"author": "muzani",

"replies": [

{

"text": "I find it already tend sto do error handling theatre, utterly pointless error handling of deeply nested try/catches that could have been handled by a single one.\nWon't this just make that even worse?",

"author": "mattmanser",

"depth": 1

}

]

},

{

"top": "I have a couple in my global CLAUDE.md I really like, but they tend to be rather specific to my own usage patterns:\n1. I call it the \"Table Flip Rule\" because naming rules helps with mnemonics:\n*The table flip rule: NEVER implement backward compatibility without explicit approval.* I am the only user 99% of the time - this isn't Google. Database migrations? Just change the schema. Breaking API changes? Just make them. No migration scripts, no deprecation warnings, no \"what if someone is using the old version?\" When in doubt: break it and move forward. (╯°□°)╯︵ ┻━┻\n2. Telling it to make more diagrams:\nMake extensive use of *ASCII diagrams* for explaining concepts, code flow, and architecture. Include diagrams in proposed plans.\n3. The ADHD disclosure. This, coupled with a post-message hook that sends the current time allows it to give me break reminders:\nI have ADHD (mainly distraction component) and can lose track of time when hyperfocused. To help:\n- Break down complex work into focused steps; use todo lists to track progress.\n- Suggest a break when we've been stuck on something for over an hour.\n- After completing something significant, suggest stepping away before the next task.\n- Log time spent on substantial tasks in the journal for future planning reference.\n4. Lastly, a fun one that also serves a positive reinforcement to reduce emoji usage:\nYou don't use emojis (except for japanese kaomoji).",

"author": "ravila4",

"replies": []

},

{

"top": "I have had better results when I force the model to explain assumptions before doing anything. A quick what are you assuming here? step catches a lot of bad directions early and saves rework later.",

"author": "allinonetools_",

"replies": []

},

{

"top": "When writing tables in markdown files, text align data in the columns for readability with a plain text editor.",

"author": "digikata",

"replies": []

},

{

"top": "You might not be aware and think it's your Claude file, but CC tends to do almost exactly what you're asking on its own now.\nYou're asking for it's default behaviour.",

"author": "mattmanser",

"replies": []

}

]

},

{

"id": "47444917",

"title": "Ask HN: The new wave of AI agent sandboxes?",

"link": "https://news.ycombinator.com/item?id=47444917",

"domain": "news.ycombinator.com",

"author": "ATechGuy",

"score": 12,

"comment_count": 5,

"created_ts": 1773949898,

"is_internal": true,

"post_text": "In the last couple of months, several new solutions for sandboxing AI agents have launched (microVMs, WASM runtimes, browser isolation, hardened tool containers, etc.). Curious to hear from people using them in production. Are they working as advertised, or are there still major tradeoffs around security, cost, and performance?<p>Here&#x27;s my list of sandboxing solutions launched in the last year alone: E2B, AIO Sandbox, Sandboxer, AgentSphere, Yolobox, Exe.dev, yolo-cage, SkillFS, ERA Jazzberry Computer, Vibekit, Daytona, Modal, Cognitora, YepCode, Run Compute, CLI Fence, Landrun, Sprites, pctx-sandbox, pctx Sandbox, Agent SDK, Lima-devbox, OpenServ, Browser Agent Playground, Flintlock Agent, Quickstart, Bouvet Sandbox, Arrakis, Cellmate (ceLLMate), AgentFence, Tasker, DenoSandbox, Capsule (WASM-based), Volant, Nono, NetFence",

"is_ask_hn": true,

"matched_keywords": [],

"comments": [

{