{"slug": "pangram", "name": "Pangram", "category": "ai", "category_name": "AI models and assistants", "description": "Pangram is an AI-text detection tool that estimates whether writing was AI-generated or human-written.", "recommenders": 9, "commenters": 23, "warners": 6, "score": 4.386, "rising": 1.04, "first": "2026-02-10", "latest": "2026-10-29", "why": "HN commenters recommend Pangram as an AI-text detector with a very low false-positive rate, and some say evasion tricks work less well against it than expected.", "why_is_quote": false, "has_verdict": true, "sentiment": {"recommend": 10, "neutral": 9, "compare": 0, "warn": 6}, "spark": [0, 0, 0, 1, 0, 0, 0, 3, 3, 1, 2, 0], "listed": true, "loved": ["Low false positives", "Overall quality"], "caveat": "Misses AI text", "short_quote": "pangram is pretty good; i use it all of the time and pay for it.", "icon": true, "domain": "pangram.com", "github_repo": null, "official_url": "https://www.pangram.com", "pricing": "freemium", "aliases": ["Pangram"], "verdict": [{"text": "HN commenters recommend Pangram as an AI-text detector with a very low false-positive rate,", "cites": [{"id": 778692320, "item_id": 48668270, "url": "https://news.ycombinator.com/item?id=48668270", "by": "saithound", "date": "2026-06-24", "stance": "recommend", "quote": "if Pangram says something is 100% AI-written, you can trust that.", "excerpt": "I use Pangram quite extensively (burning through my 600 token allowance every month). They managed to get their false positive rate impressively low: if Pangram says something is 100% AI-written, you can trust that. But they need to improve their humanizer dataset. Right now, most models can be given system prompts which cause them to emit text classified as 100% human. It looks like their automated humanizers do wor", "context": "extensive user", "use_case": "AI text detection", "switched_from": "", "story": "Exploring the internal representations of Pangram 3.3.2"}, {"id": 781581392, "item_id": 48848837, "url": "https://news.ycombinator.com/item?id=48848837", "by": "wgd", "date": "2026-07-09", "stance": "recommend", "quote": "Pangram does work, in the specific sense that when it says something was AI authored it is vanishingly unlikely that it was written by a human", "excerpt": "Pangram does work, in the specific sense that when it says something was AI authored it is vanishingly unlikely that it was written by a human (who was not deliberately trying to write like an AI), and IMO getting people to recognize that we actually do have a decent solution in this space now is pretty important if we want the Internet to remain a place for humans and not just bot swarms. > rule of three, em dashes,", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "AI Content Is Everywhere on Social Media, Especially LinkedIn"}, {"id": 787413616, "item_id": 49213351, "url": "https://news.ycombinator.com/item?id=49213351", "by": "no_multitudes", "date": "2026-08-07", "stance": "neutral", "quote": "I agree Pangram's UI is awful and often misleading. Their underlying classifier model is pretty accurate in my experience", "excerpt": "I agree Pangram's UI is awful and often misleading. Their underlying classifier model is pretty accurate in my experience, though, at least in the sense that it has very few natural false positives. If you disagree, please send me some long-form verifiable false positives that were not explicitly written to trick Pangram. I love to learn.", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "Taste Is All That's Left"}, {"id": 788130896, "item_id": 49258181, "url": "https://news.ycombinator.com/item?id=49258181", "by": "nunez", "date": "2026-08-11", "stance": "recommend", "quote": "pangram is pretty good; i use it all of the time and pay for it.", "excerpt": "pangram is pretty good; i use it all of the time and pay for it. surprised that it's not mentioned that often here. they just released a new model that is supposed to lower the fpr (false positive rate) even further than it's already impossibly low score. it also detects AI in images now, though I expect the fpr to be pretty high there given its newness.", "context": "", "use_case": "detecting AI-generated text", "switched_from": "", "story": "How Claude marks AI-generated content"}]}, {"text": " and some say evasion tricks work less well against it than expected.", "cites": [{"id": 793322064, "item_id": 49582629, "url": "https://news.ycombinator.com/item?id=49582629", "by": "bcantrill", "date": "2026-09-06", "stance": "recommend", "quote": "that technique doesn't work nearly as well (at least with Pangram 4) as you think it might", "excerpt": "Well, give it a shot -- you'll likely find that that technique doesn't work nearly as well (at least with Pangram 4) as you think it might. When we had Max on the podcast[0], Adam explicitly asked him about exactly this (after all, you can give an LLM access to Pangram and let it iterate!), and Max reported that someone had attempted to do this -- and ended up burning through $700 in tokens and had a \"sad Claude.\" An", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "The revolt of the reader"}, {"id": 783147840, "item_id": 48946740, "url": "https://news.ycombinator.com/item?id=48946740", "by": "WhitneyLand", "date": "2026-07-17", "stance": "recommend", "quote": "Not convinced just using top n sigma is going to beat a SOTA detector. The fingerprints Pangram uses", "excerpt": "Not convinced just using top n sigma is going to beat a SOTA detector. The fingerprints Pangram uses should in principle be able to detect style above the token selection level. Other papers have tried to beat it with temperature and it didn\u2019t work, although I haven\u2019t seen anyone try insane levels. Give it a shot and let me know if you have any success.", "context": "", "use_case": "detecting LLM-generated text", "switched_from": "", "story": "Detecting LLM-Generated Texts with \"Classical\" Machine Learning"}]}, {"text": " The most common complaint is that it misses AI text, especially from prompted or edited model output,", "cites": [{"id": 781601552, "item_id": 48850097, "url": "https://news.ycombinator.com/item?id=48850097", "by": "dwa3592", "date": "2026-07-09", "stance": "warn", "quote": "Pangram fails spectacularly in detecting AI.", "excerpt": "What I agree with in this article is that there is shit ton of slop on the internet today. What I disagree with is that if pangram says it's human, then it's not AI. That is hardly the case. Pangram fails spectacularly in detecting AI. https://github.com/deepanwadhwa/ai_detector_fails/blob/main/pangram/1_modified_gemini_response.png - I posted this link in another thread the other day. I might do more of these tests ", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "AI Content Is Everywhere on Social Media, Especially LinkedIn"}, {"id": 778692320, "item_id": 48668270, "url": "https://news.ycombinator.com/item?id=48668270", "by": "saithound", "date": "2026-06-24", "stance": "recommend", "quote": "if Pangram says something is 100% AI-written, you can trust that.", "excerpt": "I use Pangram quite extensively (burning through my 600 token allowance every month). They managed to get their false positive rate impressively low: if Pangram says something is 100% AI-written, you can trust that. But they need to improve their humanizer dataset. Right now, most models can be given system prompts which cause them to emit text classified as 100% human. It looks like their automated humanizers do wor", "context": "extensive user", "use_case": "AI text detection", "switched_from": "", "story": "Exploring the internal representations of Pangram 3.3.2"}, {"id": 780619232, "item_id": 48788702, "url": "https://news.ycombinator.com/item?id=48788702", "by": "dwa3592", "date": "2026-07-04", "stance": "warn", "quote": "The value is that pangram detects AI generated content which it did not in the above example.", "excerpt": "Well, false positives must anyway be zero for a product which is built around detecting AI. let me explain - the value proposition of this product is that we detect AI. Its not we don't call human writing AI generated. The latter is implied and must not happen ever. The value is that pangram detects AI generated content which it did not in the above example. if this explanation doesn't make sense then think of pangra", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "International chess federation sanctions Kramnik"}, {"id": 784355856, "item_id": 49022241, "url": "https://news.ycombinator.com/item?id=49022241", "by": "qingcharles", "date": "2026-07-23", "stance": "recommend", "quote": "It flagged the first and last paras as \"high\" AI and the middle para as \"high\" human.", "excerpt": "I agree. I ran it on a small sample last night, basically three paragraphs of AI text that I thought read 100% human, except for em dashes, which I use in writing anyway. It flagged the first and last paras as \"high\" AI and the middle para as \"high\" human. I have no idea what it is seeing that sets it off.", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "Scanning for Pangram Errors"}]}, {"text": " and several commenters call its evidence display misleading.", "cites": [{"id": 787413616, "item_id": 49213351, "url": "https://news.ycombinator.com/item?id=49213351", "by": "no_multitudes", "date": "2026-08-07", "stance": "neutral", "quote": "I agree Pangram's UI is awful and often misleading. Their underlying classifier model is pretty accurate in my experience", "excerpt": "I agree Pangram's UI is awful and often misleading. Their underlying classifier model is pretty accurate in my experience, though, at least in the sense that it has very few natural false positives. If you disagree, please send me some long-form verifiable false positives that were not explicitly written to trick Pangram. I love to learn.", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "Taste Is All That's Left"}, {"id": 775670608, "item_id": 48479413, "url": "https://news.ycombinator.com/item?id=48479413", "by": "wgd", "date": "2026-06-10", "stance": "recommend", "quote": "The main \"is this AI or human?\" classification is done with a machine learning model that works very well", "excerpt": "Yeah, the evidence feature is so terrible that it actively harms the overall reputation of Pangram. The main \"is this AI or human?\" classification is done with a machine learning model that works very well but has nothing (directly) to do with any of those stylistic cues it surfaces. In any case, the Pangram link was just meant as objective corroboration of what's pretty blatantly obvious if you just read the text. \"", "context": "", "use_case": "detecting AI-written text", "switched_from": "", "story": "A Server Called Mercury"}]}], "praise": [{"label": "Low false positives", "summary": "Commenters say a 100% AI verdict from Pangram can be trusted because human text is rarely flagged.", "quotes": [{"id": 778692320, "item_id": 48668270, "url": "https://news.ycombinator.com/item?id=48668270", "by": "saithound", "date": "2026-06-24", "stance": "recommend", "quote": "if Pangram says something is 100% AI-written, you can trust that.", "excerpt": "I use Pangram quite extensively (burning through my 600 token allowance every month). They managed to get their false positive rate impressively low: if Pangram says something is 100% AI-written, you can trust that. But they need to improve their humanizer dataset. Right now, most models can be given system prompts which cause them to emit text classified as 100% human. It looks like their automated humanizers do wor", "context": "extensive user", "use_case": "AI text detection", "switched_from": "", "story": "Exploring the internal representations of Pangram 3.3.2"}, {"id": 781581392, "item_id": 48848837, "url": "https://news.ycombinator.com/item?id=48848837", "by": "wgd", "date": "2026-07-09", "stance": "recommend", "quote": "Pangram does work, in the specific sense that when it says something was AI authored it is vanishingly unlikely that it was written by a human", "excerpt": "Pangram does work, in the specific sense that when it says something was AI authored it is vanishingly unlikely that it was written by a human (who was not deliberately trying to write like an AI), and IMO getting people to recognize that we actually do have a decent solution in this space now is pretty important if we want the Internet to remain a place for humans and not just bot swarms. > rule of three, em dashes,", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "AI Content Is Everywhere on Social Media, Especially LinkedIn"}, {"id": 793326288, "item_id": 49582893, "url": "https://news.ycombinator.com/item?id=49582893", "by": "bcantrill", "date": "2026-09-06", "stance": "recommend", "quote": "My experience is using Pangram quite often with lots of writing of all flavors", "excerpt": "My experience is using Pangram quite often with lots of writing of all flavors (including a bunch of known origin). As for my own writing, I didn't do this experiment, but one of my co-workers did -- and over 176 posts spanning 22 years, all 176 (well, 177 now with my latest) are 100% human. This is not hugely surprising in that (in addition to me having actually written them!) my voice is very... distinctive. What w", "context": "Oxide author, audits company's public writing", "use_case": "auditing public writing for AI text", "switched_from": "", "story": "The revolt of the reader"}]}, {"label": "Overall quality", "summary": "Users describe the core classifier as good and say detection has improved a lot.", "quotes": [{"id": 795779872, "item_id": 49736242, "url": "https://news.ycombinator.com/item?id=49736242", "by": "aizk", "date": "2026-09-17", "stance": "recommend", "quote": "Give it a shot, it's quite good.", "excerpt": "I don't work for Pangram, but I do a lot of writing at work. API docs, blogs, code, tweets, linkedin, all that. If you've tried AI detectors a couple years ago, they're basically in the same position that coding agents were a few years ago, where everyone was skeptical at first, but the tech has gotten a lot better. Give it a shot, it's quite good. They are slightly tuned a bit towards classifying things as AI, but I", "context": "writes API docs, blogs, code at work; not affiliated", "use_case": "AI text detection for work writing", "switched_from": "", "story": "Pangram \u2013 AI detector for text and images"}, {"id": 775670608, "item_id": 48479413, "url": "https://news.ycombinator.com/item?id=48479413", "by": "wgd", "date": "2026-06-10", "stance": "recommend", "quote": "The main \"is this AI or human?\" classification is done with a machine learning model that works very well", "excerpt": "Yeah, the evidence feature is so terrible that it actively harms the overall reputation of Pangram. The main \"is this AI or human?\" classification is done with a machine learning model that works very well but has nothing (directly) to do with any of those stylistic cues it surfaces. In any case, the Pangram link was just meant as objective corroboration of what's pretty blatantly obvious if you just read the text. \"", "context": "", "use_case": "detecting AI-written text", "switched_from": "", "story": "A Server Called Mercury"}, {"id": 788130896, "item_id": 49258181, "url": "https://news.ycombinator.com/item?id=49258181", "by": "nunez", "date": "2026-08-11", "stance": "recommend", "quote": "pangram is pretty good; i use it all of the time and pay for it.", "excerpt": "pangram is pretty good; i use it all of the time and pay for it. surprised that it's not mentioned that often here. they just released a new model that is supposed to lower the fpr (false positive rate) even further than it's already impossibly low score. it also detects AI in images now, though I expect the fpr to be pretty high there given its newness.", "context": "", "use_case": "detecting AI-generated text", "switched_from": "", "story": "How Claude marks AI-generated content"}]}, {"label": "Mixed authorship", "summary": "It reports what share of a text is AI-touched rather than only giving a yes-or-no answer.", "quotes": [{"id": 751445712, "item_id": 46965357, "url": "https://news.ycombinator.com/item?id=46965357", "by": "usefulposter", "date": "2026-02-10", "stance": "recommend", "quote": "you might want to check out the leading model", "excerpt": "I'm not affiliated with it in any way, but you might want to check out the leading model https://arxiv.org/abs/2402.14873 and the mixed authorship detection https://www.pangram.com/blog/pangram-3-0-technical. LLM detection has come a long way since the first GPTZero. Human heuristics - I've prompted millions of tokens across every frontier model iteration for all manner of writing styles and purposes - also helps gre", "context": "", "use_case": "detecting AI-generated text", "switched_from": "", "story": null}, {"id": 797775184, "item_id": 49860949, "url": "https://news.ycombinator.com/item?id=49860949", "by": "VogonPoetry", "date": "2026-09-26", "stance": "neutral", "quote": "using Pangram free \"credits\"", "excerpt": "Didn't have the same feel about the piece. But based on the parent comment, decided to also test it, using Pangram free \"credits\". For me it reported that only sections of it (30%) have been \"touched\" by an LLM. There is 60% of human written text. Pangram does not report this as 100% LLM generated. Why do I use the word \"touched\" - because there are some specific details that must have come from some existing text an", "context": "", "use_case": "detecting AI-written text", "switched_from": "", "story": "I'm the mom in that viral Giants clip. Let me tell you about my husband"}, {"id": 794246176, "item_id": 49640386, "url": "https://news.ycombinator.com/item?id=49640386", "by": "woolion", "date": "2026-09-10", "stance": "neutral", "quote": "Pangram says 50% human. That seems to check out", "excerpt": "Pangram says 50% human. That seems to check out, heavily edited yet the LLM aha sentences and cadence are still definitely there, but not overbearing. I read some Dostoevsky recently and I found my speech to be substantially altered for a while, so it makes sense that LLM writing would have an influence.", "context": "", "use_case": "detecting AI-written text", "switched_from": "", "story": "Training a 3.8B LLM to 0.384 CORE for $998"}]}, {"label": "Hard to evade", "summary": "Commenters report that attempts to beat it through iteration or sampling tricks largely failed.", "quotes": [{"id": 793322064, "item_id": 49582629, "url": "https://news.ycombinator.com/item?id=49582629", "by": "bcantrill", "date": "2026-09-06", "stance": "recommend", "quote": "that technique doesn't work nearly as well (at least with Pangram 4) as you think it might", "excerpt": "Well, give it a shot -- you'll likely find that that technique doesn't work nearly as well (at least with Pangram 4) as you think it might. When we had Max on the podcast[0], Adam explicitly asked him about exactly this (after all, you can give an LLM access to Pangram and let it iterate!), and Max reported that someone had attempted to do this -- and ended up burning through $700 in tokens and had a \"sad Claude.\" An", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "The revolt of the reader"}, {"id": 783147840, "item_id": 48946740, "url": "https://news.ycombinator.com/item?id=48946740", "by": "WhitneyLand", "date": "2026-07-17", "stance": "recommend", "quote": "Not convinced just using top n sigma is going to beat a SOTA detector. The fingerprints Pangram uses", "excerpt": "Not convinced just using top n sigma is going to beat a SOTA detector. The fingerprints Pangram uses should in principle be able to detect style above the token selection level. Other papers have tried to beat it with temperature and it didn\u2019t work, although I haven\u2019t seen anyone try insane levels. Give it a shot and let me know if you have any success.", "context": "", "use_case": "detecting LLM-generated text", "switched_from": "", "story": "Detecting LLM-Generated Texts with \"Classical\" Machine Learning"}]}], "caveats": [{"label": "Misses AI text", "summary": "Commenters say system prompts or edits can get AI text classified as human.", "quotes": [{"id": 778692320, "item_id": 48668270, "url": "https://news.ycombinator.com/item?id=48668270", "by": "saithound", "date": "2026-06-24", "stance": "recommend", "quote": "if Pangram says something is 100% AI-written, you can trust that.", "excerpt": "I use Pangram quite extensively (burning through my 600 token allowance every month). They managed to get their false positive rate impressively low: if Pangram says something is 100% AI-written, you can trust that. But they need to improve their humanizer dataset. Right now, most models can be given system prompts which cause them to emit text classified as 100% human. It looks like their automated humanizers do wor", "context": "extensive user", "use_case": "AI text detection", "switched_from": "", "story": "Exploring the internal representations of Pangram 3.3.2"}, {"id": 781601552, "item_id": 48850097, "url": "https://news.ycombinator.com/item?id=48850097", "by": "dwa3592", "date": "2026-07-09", "stance": "warn", "quote": "Pangram fails spectacularly in detecting AI.", "excerpt": "What I agree with in this article is that there is shit ton of slop on the internet today. What I disagree with is that if pangram says it's human, then it's not AI. That is hardly the case. Pangram fails spectacularly in detecting AI. https://github.com/deepanwadhwa/ai_detector_fails/blob/main/pangram/1_modified_gemini_response.png - I posted this link in another thread the other day. I might do more of these tests ", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "AI Content Is Everywhere on Social Media, Especially LinkedIn"}, {"id": 784355856, "item_id": 49022241, "url": "https://news.ycombinator.com/item?id=49022241", "by": "qingcharles", "date": "2026-07-23", "stance": "recommend", "quote": "It flagged the first and last paras as \"high\" AI and the middle para as \"high\" human.", "excerpt": "I agree. I ran it on a small sample last night, basically three paragraphs of AI text that I thought read 100% human, except for em dashes, which I use in writing anyway. It flagged the first and last paras as \"high\" AI and the middle para as \"high\" human. I have no idea what it is seeing that sets it off.", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "Scanning for Pangram Errors"}]}, {"label": "False positives", "summary": "One user says it flagged their handwritten text as AI, and another says it leans toward calling text AI.", "quotes": [{"id": 794075120, "item_id": 49629695, "url": "https://news.ycombinator.com/item?id=49629695", "by": "saejox", "date": "2026-09-09", "stance": "warn", "quote": "Pangram and GPTZero also think it is written by a bot", "excerpt": "my handwritten very detailed 10 lines comment block is %87 robot. and i use broken english with wrong grammer all the time. you are not the only one, Pangram and GPTZero also think it is written by a bot. ai detection may be an unsolvable issue.", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "Better AI code comment detector"}, {"id": 795779872, "item_id": 49736242, "url": "https://news.ycombinator.com/item?id=49736242", "by": "aizk", "date": "2026-09-17", "stance": "recommend", "quote": "Give it a shot, it's quite good.", "excerpt": "I don't work for Pangram, but I do a lot of writing at work. API docs, blogs, code, tweets, linkedin, all that. If you've tried AI detectors a couple years ago, they're basically in the same position that coding agents were a few years ago, where everyone was skeptical at first, but the tech has gotten a lot better. Give it a shot, it's quite good. They are slightly tuned a bit towards classifying things as AI, but I", "context": "writes API docs, blogs, code at work; not affiliated", "use_case": "AI text detection for work writing", "switched_from": "", "story": "Pangram \u2013 AI detector for text and images"}]}, {"label": "Misleading UI", "summary": "Commenters call the UI and evidence feature awful and misleading, even when the underlying model is accurate.", "quotes": [{"id": 787413616, "item_id": 49213351, "url": "https://news.ycombinator.com/item?id=49213351", "by": "no_multitudes", "date": "2026-08-07", "stance": "neutral", "quote": "I agree Pangram's UI is awful and often misleading. Their underlying classifier model is pretty accurate in my experience", "excerpt": "I agree Pangram's UI is awful and often misleading. Their underlying classifier model is pretty accurate in my experience, though, at least in the sense that it has very few natural false positives. If you disagree, please send me some long-form verifiable false positives that were not explicitly written to trick Pangram. I love to learn.", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "Taste Is All That's Left"}, {"id": 775670608, "item_id": 48479413, "url": "https://news.ycombinator.com/item?id=48479413", "by": "wgd", "date": "2026-06-10", "stance": "recommend", "quote": "The main \"is this AI or human?\" classification is done with a machine learning model that works very well", "excerpt": "Yeah, the evidence feature is so terrible that it actively harms the overall reputation of Pangram. The main \"is this AI or human?\" classification is done with a machine learning model that works very well but has nothing (directly) to do with any of those stylistic cues it surfaces. In any case, the Pangram link was just meant as objective corroboration of what's pretty blatantly obvious if you just read the text. \"", "context": "", "use_case": "detecting AI-written text", "switched_from": "", "story": "A Server Called Mercury"}]}, {"label": "Pricing", "summary": "One user calls the models pricey, and another runs through a monthly token allowance.", "quotes": [{"id": 795779872, "item_id": 49736242, "url": "https://news.ycombinator.com/item?id=49736242", "by": "aizk", "date": "2026-09-17", "stance": "recommend", "quote": "Give it a shot, it's quite good.", "excerpt": "I don't work for Pangram, but I do a lot of writing at work. API docs, blogs, code, tweets, linkedin, all that. If you've tried AI detectors a couple years ago, they're basically in the same position that coding agents were a few years ago, where everyone was skeptical at first, but the tech has gotten a lot better. Give it a shot, it's quite good. They are slightly tuned a bit towards classifying things as AI, but I", "context": "writes API docs, blogs, code at work; not affiliated", "use_case": "AI text detection for work writing", "switched_from": "", "story": "Pangram \u2013 AI detector for text and images"}, {"id": 778692320, "item_id": 48668270, "url": "https://news.ycombinator.com/item?id=48668270", "by": "saithound", "date": "2026-06-24", "stance": "recommend", "quote": "if Pangram says something is 100% AI-written, you can trust that.", "excerpt": "I use Pangram quite extensively (burning through my 600 token allowance every month). They managed to get their false positive rate impressively low: if Pangram says something is 100% AI-written, you can trust that. But they need to improve their humanizer dataset. Right now, most models can be given system prompts which cause them to emit text classified as 100% human. It looks like their automated humanizers do wor", "context": "extensive user", "use_case": "AI text detection", "switched_from": "", "story": "Exploring the internal representations of Pangram 3.3.2"}]}, {"label": "Signup restrictions", "summary": "One user could not sign up with custom or Gmail email domains because of a suspicious-activity error.", "quotes": [{"id": 793380896, "item_id": 49586306, "url": "https://news.ycombinator.com/item?id=49586306", "by": "usr1106", "date": "2026-09-06", "stance": "warn", "quote": "The error was always the same, there had been suspicious activity from that domain.", "excerpt": "> but they simply don\u2019t allow signing up with my custom email domain. Tried 4 different domains. 3 of email services of various kind. 4th one my private domain which has absolutely no email reputation because I use it only for internal emails and sending is not even possible. In the end I dug out some old gmail address and tried to use that. The error was always the same, there had been suspicious activity from that ", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "The revolt of the reader"}]}, {"label": "Overreliance risk", "summary": "Critics warn against treating a single detector's result as truth and question how it was trained.", "quotes": [{"id": 791612992, "item_id": 49475812, "url": "https://news.ycombinator.com/item?id=49475812", "by": "altmanaltman", "date": "2026-08-28", "stance": "warn", "quote": "This whole company is honestly such a piece of shit", "excerpt": "Checking the following with another AI checker and it says 100% AI generated. (https://originality.ai/). We should run each check on multiple checkers if you want to provide a substantive claim that something is AI-written or not. This habit of just putting text in a \"AI text checker\" and then treating whatever it says as the truth is absolutely one of the worst things to emerge in recent times. You should check out ", "context": "human writer", "use_case": "AI text detection", "switched_from": "", "story": "Gemini-3.5-Transcribe"}, {"id": 781588224, "item_id": 48849264, "url": "https://news.ycombinator.com/item?id=48849264", "by": "dvt", "date": "2026-07-09", "stance": "warn", "quote": "It's trivial to see how many people think Pangram is absolute trash[1] (because it is).", "excerpt": "> Pangram does work It's trivial to see how many people think Pangram is absolute trash[1] (because it is). > You appear to be misinformed about how Pangram specifically works, it is not based on pattern detection of that sort. I recommend reading their whitepaper, it's a pretty understandable explanation of exactly how they trained their classifier. I did read their paper (which is, by the way, very scant on details", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "AI Content Is Everywhere on Social Media, Especially LinkedIn"}]}], "who": [{"by": "bradleygibbs", "context": "author using AI 10-14 hours a day", "url": "https://news.ycombinator.com/item?id=50056088", "date": "2026-10-12"}, {"by": "Imustaskforhelp", "context": "first-year college student", "url": "https://news.ycombinator.com/item?id=49974524", "date": "2026-10-06"}, {"by": "aizk", "context": "writes API docs, blogs, code at work; not affiliated", "url": "https://news.ycombinator.com/item?id=49736242", "date": "2026-09-17"}, {"by": "bcantrill", "context": "Oxide author, audits company's public writing", "url": "https://news.ycombinator.com/item?id=49582893", "date": "2026-09-06"}, {"by": "saithound", "context": "extensive user", "url": "https://news.ycombinator.com/item?id=48668270", "date": "2026-06-24"}, {"by": "scared_together", "context": "free plan user", "url": "https://news.ycombinator.com/item?id=48091295", "date": "2026-05-10"}], "tags": ["for checking AI-written text", "for writers auditing their work", "free tier available"], "use_cases": ["ai text detection", "detecting ai-written text", "detecting ai-generated text", "ai-text detection", "ai text detection for work writing", "auditing public writing for ai text"], "compared": [], "makers": [{"id": 783094208, "item_id": 48943388, "url": "https://news.ycombinator.com/item?id=48943388", "by": "maxspero", "date": "2026-07-16", "stance": "neutral", "quote": "The synthetic mirrors method for data generation used here is the same as what we use at Pangram.", "excerpt": "> Sounds promising, right? I spent some time trying [perplexity], but results were disappointing\u2014plenty of false positives and false negatives, and no reasonable threshold could be set. Perplexity was widely considered SOTA in 2022. One part of it is because everyone was evaluating on open models or closed models that were still close (i.e. GPT-2 vs. GPT-3.5). Today, the gap is so much wider between the models you ca", "context": "works at Pangram", "use_case": "AI text detection", "switched_from": "", "story": "Detecting LLM-Generated Texts with \"Classical\" Machine Learning"}], "threads": [{"url": "https://news.ycombinator.com/item?id=49580939", "title": "The revolt of the reader", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=48847940", "title": "AI Content Is Everywhere on Social Media, Especially LinkedIn", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=48936880", "title": "Detecting LLM-Generated Texts with \"Classical\" Machine Learning", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=50265393", "title": "Code-execution flaw in UEFI firmware threatens most servers on planet Earth", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=50171936", "title": "Symbolic Learning in Python and Julia", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=50044028", "title": "Loop Engineering: Optimizing Code Generation and Verification", "mentions": 1}], "sentiment_by_month": {"2025-11": {}, "2025-12": {}, "2026-01": {}, "2026-02": {"recommend": 1}, "2026-03": {}, "2026-04": {}, "2026-05": {"neutral": 1}, "2026-06": {"recommend": 3, "warn": 1}, "2026-07": {"warn": 2, "recommend": 3}, "2026-08": {"neutral": 2, "recommend": 1, "warn": 1}, "2026-09": {"recommend": 2, "warn": 2, "neutral": 3}, "2026-10": {"neutral": 3}}, "quotes": [{"id": 804562113, "item_id": 50285132, "url": "https://news.ycombinator.com/item?id=50285132", "by": "Imustaskforhelp", "date": "2026-10-29", "stance": "neutral", "quote": "pangram does say 97% human and only 3% AI", "excerpt": "My apologies, I take back what I say but I tested it on https://kouh.me/ai-text-detector because that's the only AI text detector which doesn't have some word limit. (It was recently submitted as a show HN[0]) I then tested it on pangram and gptzero and pangram does say 97% human and only 3% AI and GPTzero says 100% human For some reason, my AI spidey sense had activated and I used the new AI text detector and it had", "context": "", "use_case": "detecting AI-written text", "switched_from": "", "story": "Code-execution flaw in UEFI firmware threatens most servers on planet Earth"}, {"id": 803634193, "item_id": 50227137, "url": "https://news.ycombinator.com/item?id=50227137", "by": "saithound", "date": "2026-10-25", "stance": "neutral", "quote": "Pangram reports 100% AI on most comment blurbs in the examples directory.", "excerpt": "PySR itself is still alright, mostly because of previous development. But not for long. Quality control has collapsed, and e.g. the examples in the repo are now embarrassing vibeslop. Pangram reports 100% AI on most comment blurbs in the examples directory. That in itself is not the problem, AI-generated code could still be valid. But in this case, the AI-generated examples have the characteristic shape of slop: plau", "context": "", "use_case": "detecting AI-generated text", "switched_from": "", "story": "Symbolic Learning in Python and Julia"}, {"id": 800897408, "item_id": 50056088, "url": "https://news.ycombinator.com/item?id=50056088", "by": "bradleygibbs", "date": "2026-10-12", "stance": "neutral", "quote": "Do I know how to use Pangram, too? Absolutely. I got my Pangram merit badge months ago.", "excerpt": "Did I use AI to help write this article? Absolutely. I use AI 10-14 hours day, seven days a week to do all sorts of things, including trying to get a code generation and verification app off the ground. It would be unnatural for me not to use AI to research and help think through this article. Do I know how to use Pangram, too? Absolutely. I got my Pangram merit badge months ago. So, why didn't I make sure I got a 10", "context": "author using AI 10-14 hours a day", "use_case": "AI-text detection", "switched_from": "", "story": "Loop Engineering: Optimizing Code Generation and Verification"}, {"id": 799592384, "item_id": 49974524, "url": "https://news.ycombinator.com/item?id=49974524", "by": "Imustaskforhelp", "date": "2026-10-06", "stance": "neutral", "quote": "I decided to give it to pangram to find out it was 77% written by AI.", "excerpt": "I think that I might be just 1 year elder than you but I am in first year of college and we had a prof com exam (just yesterday) while preparing for it, aside from the slides, our sir had also given us a pdf and I just had the vibe because of the law of triads (AI talks much more in threes) and so I decided to give it to pangram to find out it was 77% written by AI. The sir himself said that chatgpt is a tool and sho", "context": "first-year college student", "use_case": "AI text detection", "switched_from": "", "story": "AI tutoring with Khanmigo in a two-year school experiment"}, {"id": 798314336, "item_id": 49894646, "url": "https://news.ycombinator.com/item?id=49894646", "by": "snaveen", "date": "2026-09-29", "stance": "neutral", "quote": "this is 100% ai generated as per pangram", "excerpt": "I love this author's past work and contributions, but this is 100% ai generated as per pangram https://www.pangram.com/history/9b3e03f0-6aa0-4516-90f9-507b0b19af07, which makes me wonder.", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "Can a language model run in Linux eBPF?"}, {"id": 797775184, "item_id": 49860949, "url": "https://news.ycombinator.com/item?id=49860949", "by": "VogonPoetry", "date": "2026-09-26", "stance": "neutral", "quote": "using Pangram free \"credits\"", "excerpt": "Didn't have the same feel about the piece. But based on the parent comment, decided to also test it, using Pangram free \"credits\". For me it reported that only sections of it (30%) have been \"touched\" by an LLM. There is 60% of human written text. Pangram does not report this as 100% LLM generated. Why do I use the word \"touched\" - because there are some specific details that must have come from some existing text an", "context": "", "use_case": "detecting AI-written text", "switched_from": "", "story": "I'm the mom in that viral Giants clip. Let me tell you about my husband"}, {"id": 795779872, "item_id": 49736242, "url": "https://news.ycombinator.com/item?id=49736242", "by": "aizk", "date": "2026-09-17", "stance": "recommend", "quote": "Give it a shot, it's quite good.", "excerpt": "I don't work for Pangram, but I do a lot of writing at work. API docs, blogs, code, tweets, linkedin, all that. If you've tried AI detectors a couple years ago, they're basically in the same position that coding agents were a few years ago, where everyone was skeptical at first, but the tech has gotten a lot better. Give it a shot, it's quite good. They are slightly tuned a bit towards classifying things as AI, but I", "context": "writes API docs, blogs, code at work; not affiliated", "use_case": "AI text detection for work writing", "switched_from": "", "story": "Pangram \u2013 AI detector for text and images"}, {"id": 794246176, "item_id": 49640386, "url": "https://news.ycombinator.com/item?id=49640386", "by": "woolion", "date": "2026-09-10", "stance": "neutral", "quote": "Pangram says 50% human. That seems to check out", "excerpt": "Pangram says 50% human. That seems to check out, heavily edited yet the LLM aha sentences and cadence are still definitely there, but not overbearing. I read some Dostoevsky recently and I found my speech to be substantially altered for a while, so it makes sense that LLM writing would have an influence.", "context": "", "use_case": "detecting AI-written text", "switched_from": "", "story": "Training a 3.8B LLM to 0.384 CORE for $998"}, {"id": 794075120, "item_id": 49629695, "url": "https://news.ycombinator.com/item?id=49629695", "by": "saejox", "date": "2026-09-09", "stance": "warn", "quote": "Pangram and GPTZero also think it is written by a bot", "excerpt": "my handwritten very detailed 10 lines comment block is %87 robot. and i use broken english with wrong grammer all the time. you are not the only one, Pangram and GPTZero also think it is written by a bot. ai detection may be an unsolvable issue.", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "Better AI code comment detector"}, {"id": 793380896, "item_id": 49586306, "url": "https://news.ycombinator.com/item?id=49586306", "by": "usr1106", "date": "2026-09-06", "stance": "warn", "quote": "The error was always the same, there had been suspicious activity from that domain.", "excerpt": "> but they simply don\u2019t allow signing up with my custom email domain. Tried 4 different domains. 3 of email services of various kind. 4th one my private domain which has absolutely no email reputation because I use it only for internal emails and sending is not even possible. In the end I dug out some old gmail address and tried to use that. The error was always the same, there had been suspicious activity from that ", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "The revolt of the reader"}, {"id": 793326288, "item_id": 49582893, "url": "https://news.ycombinator.com/item?id=49582893", "by": "bcantrill", "date": "2026-09-06", "stance": "recommend", "quote": "My experience is using Pangram quite often with lots of writing of all flavors", "excerpt": "My experience is using Pangram quite often with lots of writing of all flavors (including a bunch of known origin). As for my own writing, I didn't do this experiment, but one of my co-workers did -- and over 176 posts spanning 22 years, all 176 (well, 177 now with my latest) are 100% human. This is not hugely surprising in that (in addition to me having actually written them!) my voice is very... distinctive. What w", "context": "Oxide author, audits company's public writing", "use_case": "auditing public writing for AI text", "switched_from": "", "story": "The revolt of the reader"}, {"id": 793322064, "item_id": 49582629, "url": "https://news.ycombinator.com/item?id=49582629", "by": "bcantrill", "date": "2026-09-06", "stance": "recommend", "quote": "that technique doesn't work nearly as well (at least with Pangram 4) as you think it might", "excerpt": "Well, give it a shot -- you'll likely find that that technique doesn't work nearly as well (at least with Pangram 4) as you think it might. When we had Max on the podcast[0], Adam explicitly asked him about exactly this (after all, you can give an LLM access to Pangram and let it iterate!), and Max reported that someone had attempted to do this -- and ended up burning through $700 in tokens and had a \"sad Claude.\" An", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "The revolt of the reader"}, {"id": 791612992, "item_id": 49475812, "url": "https://news.ycombinator.com/item?id=49475812", "by": "altmanaltman", "date": "2026-08-28", "stance": "warn", "quote": "This whole company is honestly such a piece of shit", "excerpt": "Checking the following with another AI checker and it says 100% AI generated. (https://originality.ai/). We should run each check on multiple checkers if you want to provide a substantive claim that something is AI-written or not. This habit of just putting text in a \"AI text checker\" and then treating whatever it says as the truth is absolutely one of the worst things to emerge in recent times. You should check out ", "context": "human writer", "use_case": "AI text detection", "switched_from": "", "story": "Gemini-3.5-Transcribe"}, {"id": 788130896, "item_id": 49258181, "url": "https://news.ycombinator.com/item?id=49258181", "by": "nunez", "date": "2026-08-11", "stance": "recommend", "quote": "pangram is pretty good; i use it all of the time and pay for it.", "excerpt": "pangram is pretty good; i use it all of the time and pay for it. surprised that it's not mentioned that often here. they just released a new model that is supposed to lower the fpr (false positive rate) even further than it's already impossibly low score. it also detects AI in images now, though I expect the fpr to be pretty high there given its newness.", "context": "", "use_case": "detecting AI-generated text", "switched_from": "", "story": "How Claude marks AI-generated content"}, {"id": 787413616, "item_id": 49213351, "url": "https://news.ycombinator.com/item?id=49213351", "by": "no_multitudes", "date": "2026-08-07", "stance": "neutral", "quote": "I agree Pangram's UI is awful and often misleading. Their underlying classifier model is pretty accurate in my experience", "excerpt": "I agree Pangram's UI is awful and often misleading. Their underlying classifier model is pretty accurate in my experience, though, at least in the sense that it has very few natural false positives. If you disagree, please send me some long-form verifiable false positives that were not explicitly written to trick Pangram. I love to learn.", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "Taste Is All That's Left"}, {"id": 786991024, "item_id": 49186939, "url": "https://news.ycombinator.com/item?id=49186939", "by": "arjie", "date": "2026-08-05", "stance": "neutral", "quote": "Their post gets hard-flagged by Pangram haha.", "excerpt": "Haha, I have to agree. Their style is either intentionally similar for humorous reasons or they're the kind of writer who was copied. > The words are the crime scene, not the alibi. If the latter, it must suck haha. It reminds me of the fact that I used to wear masks when I had a cold so that I could do things with my friends like help them move or whatever without risking them getting sick. But then one day, a pande", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "I'm Scared a Stranger Will Call My Novel AI, So I Built GitHub for Words"}, {"id": 784355856, "item_id": 49022241, "url": "https://news.ycombinator.com/item?id=49022241", "by": "qingcharles", "date": "2026-07-23", "stance": "recommend", "quote": "It flagged the first and last paras as \"high\" AI and the middle para as \"high\" human.", "excerpt": "I agree. I ran it on a small sample last night, basically three paragraphs of AI text that I thought read 100% human, except for em dashes, which I use in writing anyway. It flagged the first and last paras as \"high\" AI and the middle para as \"high\" human. I have no idea what it is seeing that sets it off.", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "Scanning for Pangram Errors"}, {"id": 783147840, "item_id": 48946740, "url": "https://news.ycombinator.com/item?id=48946740", "by": "WhitneyLand", "date": "2026-07-17", "stance": "recommend", "quote": "Not convinced just using top n sigma is going to beat a SOTA detector. The fingerprints Pangram uses", "excerpt": "Not convinced just using top n sigma is going to beat a SOTA detector. The fingerprints Pangram uses should in principle be able to detect style above the token selection level. Other papers have tried to beat it with temperature and it didn\u2019t work, although I haven\u2019t seen anyone try insane levels. Give it a shot and let me know if you have any success.", "context": "", "use_case": "detecting LLM-generated text", "switched_from": "", "story": "Detecting LLM-Generated Texts with \"Classical\" Machine Learning"}, {"id": 781601552, "item_id": 48850097, "url": "https://news.ycombinator.com/item?id=48850097", "by": "dwa3592", "date": "2026-07-09", "stance": "warn", "quote": "Pangram fails spectacularly in detecting AI.", "excerpt": "What I agree with in this article is that there is shit ton of slop on the internet today. What I disagree with is that if pangram says it's human, then it's not AI. That is hardly the case. Pangram fails spectacularly in detecting AI. https://github.com/deepanwadhwa/ai_detector_fails/blob/main/pangram/1_modified_gemini_response.png - I posted this link in another thread the other day. I might do more of these tests ", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "AI Content Is Everywhere on Social Media, Especially LinkedIn"}, {"id": 781588224, "item_id": 48849264, "url": "https://news.ycombinator.com/item?id=48849264", "by": "dvt", "date": "2026-07-09", "stance": "warn", "quote": "It's trivial to see how many people think Pangram is absolute trash[1] (because it is).", "excerpt": "> Pangram does work It's trivial to see how many people think Pangram is absolute trash[1] (because it is). > You appear to be misinformed about how Pangram specifically works, it is not based on pattern detection of that sort. I recommend reading their whitepaper, it's a pretty understandable explanation of exactly how they trained their classifier. I did read their paper (which is, by the way, very scant on details", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "AI Content Is Everywhere on Social Media, Especially LinkedIn"}, {"id": 781581392, "item_id": 48848837, "url": "https://news.ycombinator.com/item?id=48848837", "by": "wgd", "date": "2026-07-09", "stance": "recommend", "quote": "Pangram does work, in the specific sense that when it says something was AI authored it is vanishingly unlikely that it was written by a human", "excerpt": "Pangram does work, in the specific sense that when it says something was AI authored it is vanishingly unlikely that it was written by a human (who was not deliberately trying to write like an AI), and IMO getting people to recognize that we actually do have a decent solution in this space now is pretty important if we want the Internet to remain a place for humans and not just bot swarms. > rule of three, em dashes,", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "AI Content Is Everywhere on Social Media, Especially LinkedIn"}, {"id": 780619232, "item_id": 48788702, "url": "https://news.ycombinator.com/item?id=48788702", "by": "dwa3592", "date": "2026-07-04", "stance": "warn", "quote": "The value is that pangram detects AI generated content which it did not in the above example.", "excerpt": "Well, false positives must anyway be zero for a product which is built around detecting AI. let me explain - the value proposition of this product is that we detect AI. Its not we don't call human writing AI generated. The latter is implied and must not happen ever. The value is that pangram detects AI generated content which it did not in the above example. if this explanation doesn't make sense then think of pangra", "context": "", "use_case": "AI text detection", "switched_from": "", "story": "International chess federation sanctions Kramnik"}, {"id": 778692320, "item_id": 48668270, "url": "https://news.ycombinator.com/item?id=48668270", "by": "saithound", "date": "2026-06-24", "stance": "recommend", "quote": "if Pangram says something is 100% AI-written, you can trust that.", "excerpt": "I use Pangram quite extensively (burning through my 600 token allowance every month). They managed to get their false positive rate impressively low: if Pangram says something is 100% AI-written, you can trust that. But they need to improve their humanizer dataset. Right now, most models can be given system prompts which cause them to emit text classified as 100% human. It looks like their automated humanizers do wor", "context": "extensive user", "use_case": "AI text detection", "switched_from": "", "story": "Exploring the internal representations of Pangram 3.3.2"}, {"id": 778433728, "item_id": 48652108, "url": "https://news.ycombinator.com/item?id=48652108", "by": "garaetjjte", "date": "2026-06-23", "stance": "warn", "quote": "But definitely no coincidences here.", "excerpt": "Maybe they don't do that for larger destination providers. But definitely no coincidences here. (in the post I replaced address with example.com because I'm curious if I will ever get other spam onto it, but here's another one unmodified) curl --request POST --data '{\"email\": \"pangramdemo@milek7.pl\"}' https://www.pangram.com/api/validate-email https://milek7.pl/mailverifyspam/another.txt", "context": "", "use_case": "", "switched_from": "", "story": "Don't verify email addresses by sending spam to them"}, {"id": 775670608, "item_id": 48479413, "url": "https://news.ycombinator.com/item?id=48479413", "by": "wgd", "date": "2026-06-10", "stance": "recommend", "quote": "The main \"is this AI or human?\" classification is done with a machine learning model that works very well", "excerpt": "Yeah, the evidence feature is so terrible that it actively harms the overall reputation of Pangram. The main \"is this AI or human?\" classification is done with a machine learning model that works very well but has nothing (directly) to do with any of those stylistic cues it surfaces. In any case, the Pangram link was just meant as objective corroboration of what's pretty blatantly obvious if you just read the text. \"", "context": "", "use_case": "detecting AI-written text", "switched_from": "", "story": "A Server Called Mercury"}, {"id": 775129776, "item_id": 48445611, "url": "https://news.ycombinator.com/item?id=48445611", "by": "armchairhacker", "date": "2026-06-08", "stance": "recommend", "quote": "I recommend checking on Pangram.", "excerpt": "HN has comments which are social media. I don\u2019t get why people care, because social media isn\u2019t intrinsically bad; I always say \u201cmainstream social media\u201d or \u201ctoxic social media\u201d to clarify what I\u2019m referring to. > HN makes it quite difficult (technically impossible, but you can link out) to embed media such as videos and pictures, which makes it fundamentally different than \"social media\". What\u2019s the difference? Subm", "context": "", "use_case": "detecting AI-generated text", "switched_from": "", "story": "Anti-social: It's fads, not friends, which now dominate social media feeds"}, {"id": 769460720, "item_id": 48091295, "url": "https://news.ycombinator.com/item?id=48091295", "by": "scared_together", "date": "2026-05-10", "stance": "neutral", "quote": "Pangram indeed claims the OP is 76% AI-generated.", "excerpt": "I never used Pangram before today, but since I've seen it mentioned many times on HN and I enjoyed reading the OP, I decided to try it. I am only using the free plan so let me know if I'm missing something. I am assuming the parent was referring to the tool hosted at pangram.com and not some other tool of the same name. Pangram indeed claims the OP is 76% AI-generated. It has \"high confidence\" (EDIT: some parts are \"", "context": "free plan user", "use_case": "AI-generated text detection", "switched_from": "", "story": "Incident CVE-2024-Yikes"}, {"id": 751445712, "item_id": 46965357, "url": "https://news.ycombinator.com/item?id=46965357", "by": "usefulposter", "date": "2026-02-10", "stance": "recommend", "quote": "you might want to check out the leading model", "excerpt": "I'm not affiliated with it in any way, but you might want to check out the leading model https://arxiv.org/abs/2402.14873 and the mixed authorship detection https://www.pangram.com/blog/pangram-3-0-technical. LLM detection has come a long way since the first GPTZero. Human heuristics - I've prompted millions of tokens across every frontier model iteration for all manner of writing styles and purposes - also helps gre", "context": "", "use_case": "detecting AI-generated text", "switched_from": "", "story": null}]}