I\u2019m wondering if that are technical reasons why the newish browsers (such as Brave or Edge) are choosing Chromium instead of Firefox as their starting point.
If so, shouldn\u2019t this be a priority for Mozilla to change that?"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Ask HN: Why do new browsers use Chromium instead of Firefox as their base?"}},"_tags":["story","author_josemando","story_29900496","ask_hn"],"author":"josemando","children":[29900623,29900757,29900824,29900905,29900906,29900907,29900962,29900964,29901114,29901229,29901259,29901392,29901441,29901470,29901687,29901726,29901727,29901739,29901766,29901780,29901903,29901960,29902009,29902050,29902467,29902472,29903279,29903315,29903357,29903387,29903426,29903576,29903610,29903791,29904423,29904482,29904625,29905923,29905998,29906356,29906656,29908667,29913193],"created_at":"2022-01-12T00:21:34Z","created_at_i":1641946894,"num_comments":183,"objectID":"29900496","points":301,"story_id":29900496,"story_text":"Every time there is a post here related to Firefox, I see a lot of people complaining about the state of browsers and the utterly dominance of Chrome and Chromium based browsers.
I\u2019m wondering if that are technical reasons why the newish browsers (such as Brave or Edge) are choosing Chromium instead of Firefox as their starting point.
If so, shouldn\u2019t this be a priority for Mozilla to change that?","title":"Ask HN: Why do new browsers use Chromium instead of Firefox as their base?","updated_at":"2025-01-20T17:07:24Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"MagMueller"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Hey HN, we\u2019re Gregor and Magnus, the founders of browser-use (https://browser-use.com/), an easy way to connect AI agents with the browser. Our agent library is open-source (https://github.com/browser-use/browser-use) and we have what is the biggest open-source community for browser agents. And now we have a cloud offering\u2014hence our Launch HN today!
Check out this video to see it in action: https://preview.screen.studio/share/r1h4DuAk. There are lots more demos at https://github.com/browser-use/browser-use on how we control the web with prompts.
We started coding a decade ago with Selenium bots and macros to automate tasks. Then we both moved into ML. Last November, we asked ourselves, \u201cHow hard could it be to build the interface between LLMs and the web?\u201d
We launched on Show HN (https://news.ycombinator.com/item?id=42052432) and have since been addressing various challenges of browser automation, such as: - Automation scripts break when the website changes - Automation scripts are annoying to build - Captchas and rate limits - parsing errors and API key management - and perhaps worst of all, login screens.
People use us to fill out their forms, extract data behind login walls, or automate their CRM. Others use the xPaths browser-use clicked on and build their scripts faster, or directly rerun the actions of browser-use deterministically. We\u2019re currently working on robust task reruns, agent memory for long tasks, parallelization for repetitive tasks, and many other sweet improvements.
One interesting aspect is that some companies now want to change their UI to be more agent-friendly. Some developers even replace ugly UIs with nice ones and use browser-use to copy data over.
Besides the open-source we have an API. We host the browser and LLMs for you and help you with handling proxy rotation, persistent sessions and allowing you to run multiple instances in parallel. We price at $30/month\u2014significantly lower than OpenAI\u2019s Operator.
On the open-source side, browser use remains free. You can use any LLM, from Gemini to Sonnet, Qwen, or even DeepSeek-R1. It\u2019s licensed under MIT, giving you full freedom to customize it.
We\u2019d love to hear from you\u2014what automation challenges are you facing? Any thoughts, questions, experiences are welcome!"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Launch HN: Browser Use (YC W25) \u2013 open-source web agents"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"https://github.com/browser-use/browser-use"}},"_tags":["story","author_MagMueller","story_43173378","launch_hn"],"author":"MagMueller","children":[43173943,43174113,43174203,43174273,43174410,43174518,43174557,43174646,43174648,43175070,43175657,43176291,43176692,43177852,43178320,43178708,43178842,43179025,43180320,43180349,43180630,43180642,43180816,43181274,43181425,43182614,43183138,43184102,43185275,43187833,43192608,43228887],"created_at":"2025-02-25T15:45:17Z","created_at_i":1740498317,"num_comments":100,"objectID":"43173378","points":259,"story_id":43173378,"story_text":"Hey HN, we\u2019re Gregor and Magnus, the founders of browser-use (https://browser-use.com/), an easy way to connect AI agents with the browser. Our agent library is open-source (https://github.com/browser-use/browser-use) and we have what is the biggest open-source community for browser agents. And now we have a cloud offering\u2014hence our Launch HN today!
Check out this video to see it in action: https://preview.screen.studio/share/r1h4DuAk. There are lots more demos at https://github.com/browser-use/browser-use on how we control the web with prompts.
We started coding a decade ago with Selenium bots and macros to automate tasks. Then we both moved into ML. Last November, we asked ourselves, \u201cHow hard could it be to build the interface between LLMs and the web?\u201d
We launched on Show HN (https://news.ycombinator.com/item?id=42052432) and have since been addressing various challenges of browser automation, such as: - Automation scripts break when the website changes - Automation scripts are annoying to build - Captchas and rate limits - parsing errors and API key management - and perhaps worst of all, login screens.
People use us to fill out their forms, extract data behind login walls, or automate their CRM. Others use the xPaths browser-use clicked on and build their scripts faster, or directly rerun the actions of browser-use deterministically. We\u2019re currently working on robust task reruns, agent memory for long tasks, parallelization for repetitive tasks, and many other sweet improvements.
One interesting aspect is that some companies now want to change their UI to be more agent-friendly. Some developers even replace ugly UIs with nice ones and use browser-use to copy data over.
Besides the open-source we have an API. We host the browser and LLMs for you and help you with handling proxy rotation, persistent sessions and allowing you to run multiple instances in parallel. We price at $30/month\u2014significantly lower than OpenAI\u2019s Operator.
On the open-source side, browser use remains free. You can use any LLM, from Gemini to Sonnet, Qwen, or even DeepSeek-R1. It\u2019s licensed under MIT, giving you full freedom to customize it.
We\u2019d love to hear from you\u2014what automation challenges are you facing? Any thoughts, questions, experiences are welcome!","title":"Launch HN: Browser Use (YC W25) \u2013 open-source web agents","updated_at":"2026-08-13T14:27:33Z","url":"https://github.com/browser-use/browser-use"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"jerheinze"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"An advanced browser fingerprint calculator aimed mainly at Tor Browser users"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://fpcentral.irisa.fr"}},"_tags":["story","author_jerheinze","story_13925238"],"author":"jerheinze","children":[13927941,13928066,13928474,13928999,13929170,13929916,13929919,13930358,13930662,13930773,13932841],"created_at":"2017-03-21T19:11:44Z","created_at_i":1490123504,"num_comments":77,"objectID":"13925238","points":158,"story_id":13925238,"title":"An advanced browser fingerprint calculator aimed mainly at Tor Browser users","updated_at":"2024-09-20T00:33:22Z","url":"https://fpcentral.irisa.fr"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"cheiVia0"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Tor Browser User Manual"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"https://blog.torproject.org/blog/announcing-tor-browser-user-manual"}},"_tags":["story","author_cheiVia0","story_12852912"],"author":"cheiVia0","children":[12853056,12853266,12853747,12854095,12854861],"created_at":"2016-11-02T07:43:04Z","created_at_i":1478072584,"num_comments":18,"objectID":"12852912","points":134,"story_id":12852912,"title":"Tor Browser User Manual","updated_at":"2024-09-19T23:54:54Z","url":"https://blog.torproject.org/blog/announcing-tor-browser-user-manual"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"nan60"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Why do all browsers' user agents start with \"Mozilla/\"? (2008)"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"https://stackoverflow.com/questions/1114254/why-do-all-browsers-user-agents-start-with-mozilla"}},"_tags":["story","author_nan60","story_44349845"],"author":"nan60","children":[44349882,44350092,44350173,44350289,44350293,44350379,44350507,44350837,44350865,44351728],"created_at":"2025-06-22T20:04:33Z","created_at_i":1750622673,"num_comments":62,"objectID":"44349845","points":116,"story_id":44349845,"title":"Why do all browsers' user agents start with \"Mozilla/\"? (2008)","updated_at":"2025-08-14T22:03:30Z","url":"https://stackoverflow.com/questions/1114254/why-do-all-browsers-user-agents-start-with-mozilla"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"bhaisaab"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the browser user-agent string"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://webaim.org/blog/user-agent-string-history/"}},"_tags":["story","author_bhaisaab","story_6674812"],"author":"bhaisaab","children":[6675100,6675135,6675366,6675412,6675484,6675597,6675609,6675659,6676087,6676284,6676529,6676714,6679533],"created_at":"2013-11-05T13:08:45Z","created_at_i":1383656925,"num_comments":32,"objectID":"6674812","points":100,"story_id":6674812,"story_text":"","title":"History of the browser user-agent string","updated_at":"2023-09-06T22:35:39Z","url":"http://webaim.org/blog/user-agent-string-history/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"choult"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the browser user-agent string"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://webaim.org/blog/user-agent-string-history/"}},"_tags":["story","author_choult","story_31246438"],"author":"choult","children":[31263253,31263381,31263390,31263584,31263761,31263822,31265441,31266867,31267045],"created_at":"2022-05-03T09:47:24Z","created_at_i":1651571244,"num_comments":20,"objectID":"31246438","points":87,"story_id":31246438,"title":"History of the browser user-agent string","updated_at":"2026-05-13T09:00:36Z","url":"https://webaim.org/blog/user-agent-string-history/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"edgarvm"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Windows DRM Files Used to Decloak Tor Browser Users"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"https://www.bleepingcomputer.com/news/security/windows-drm-files-used-to-decloak-tor-browser-users/"}},"_tags":["story","author_edgarvm","story_13559608"],"author":"edgarvm","children":[13559765,13559834,13559862,13560328,13560436,13560597,13560600,13560699,13561008,13561402],"created_at":"2017-02-03T13:24:39Z","created_at_i":1486128279,"num_comments":32,"objectID":"13559608","points":79,"story_id":13559608,"title":"Windows DRM Files Used to Decloak Tor Browser Users","updated_at":"2024-09-20T00:21:41Z","url":"https://www.bleepingcomputer.com/news/security/windows-drm-files-used-to-decloak-tor-browser-users/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"tosh"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the browser user-agent string (2008)"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://webaim.org/blog/user-agent-string-history/"}},"_tags":["story","author_tosh","story_21085388"],"author":"tosh","children":[21086065,21099095,21099180,21099276,21099380,21100853,21101963],"created_at":"2019-09-26T20:49:16Z","created_at_i":1569530956,"num_comments":23,"objectID":"21085388","points":62,"story_id":21085388,"title":"History of the browser user-agent string (2008)","updated_at":"2024-09-20T05:02:17Z","url":"https://webaim.org/blog/user-agent-string-history/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"QuantumGood"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Browsers used in 5500+ HN visits yesterday"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://i.imgur.com/nXTkk.png"}},"_tags":["story","author_QuantumGood","story_1755686"],"author":"QuantumGood","children":[1755762,1755816,1755824,1755884,1755913,1756055,1756077,1756100,1756220,1756324,1756368,1757301,1759055,1763324],"created_at":"2010-10-04T14:17:06Z","created_at_i":1286201826,"num_comments":32,"objectID":"1755686","points":58,"story_id":1755686,"story_text":"","title":"Browsers used in 5500+ HN visits yesterday","updated_at":"2024-09-19T17:14:31Z","url":"http://i.imgur.com/nXTkk.png"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"jtchang"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the browser user-agent string (2008)"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://webaim.org/blog/user-agent-string-history/"}},"_tags":["story","author_jtchang","story_11693455"],"author":"jtchang","children":[11693731,11693789,11693820,11693872,11693881,11693934,11694288,11696225,11697783],"created_at":"2016-05-13T21:54:23Z","created_at_i":1463176463,"num_comments":12,"objectID":"11693455","points":38,"story_id":11693455,"title":"History of the browser user-agent string (2008)","updated_at":"2024-09-19T23:15:19Z","url":"http://webaim.org/blog/user-agent-string-history/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"rbdeveloper"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Brave Browser uses 100% CPU when back end servers are down"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/brave/brave-browser/issues/19125"}},"_tags":["story","author_rbdeveloper","story_29061436"],"author":"rbdeveloper","children":[29061437,29062998],"created_at":"2021-10-31T23:20:48Z","created_at_i":1635722448,"num_comments":5,"objectID":"29061436","points":37,"story_id":29061436,"title":"Brave Browser uses 100% CPU when back end servers are down","updated_at":"2024-09-20T09:42:47Z","url":"https://github.com/brave/brave-browser/issues/19125"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"spydez"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the browser user-agent string"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://www.webaim.org/blog/user-agent-string-history/"}},"_tags":["story","author_spydez","story_298844"],"author":"spydez","children":[299145],"created_at":"2008-09-08T22:07:23Z","created_at_i":1220911643,"num_comments":4,"objectID":"298844","points":36,"story_id":298844,"title":"History of the browser user-agent string","updated_at":"2023-09-06T19:22:56Z","url":"http://www.webaim.org/blog/user-agent-string-history/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"dmitrygr"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Google introduces a new system for tracking Chrome browser users"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.nytimes.com/2022/01/25/business/google-topics-chrome-tracking.html"}},"_tags":["story","author_dmitrygr","story_30088089"],"author":"dmitrygr","children":[30089009,30089146,30089876,30089962],"created_at":"2022-01-26T17:01:07Z","created_at_i":1643216467,"num_comments":11,"objectID":"30088089","points":35,"story_id":30088089,"title":"Google introduces a new system for tracking Chrome browser users","updated_at":"2024-09-20T10:23:46Z","url":"https://www.nytimes.com/2022/01/25/business/google-topics-chrome-tracking.html"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"zerojames"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"It seems routine to see a bunch of browser User-Agents from the same IP"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://utcc.utoronto.ca/~cks/space/blog/web/MultipleUserAgentsSameIP"}},"_tags":["story","author_zerojames","story_40749211"],"author":"zerojames","children":[40749877,40749934,40749949,40749965,40749975,40749978,40750070,40750127,40750139,40750170,40750413,40750424,40751473,40752181],"created_at":"2024-06-21T13:05:21Z","created_at_i":1718975121,"num_comments":29,"objectID":"40749211","points":31,"story_id":40749211,"title":"It seems routine to see a bunch of browser User-Agents from the same IP","updated_at":"2025-11-28T11:50:41Z","url":"https://utcc.utoronto.ca/~cks/space/blog/web/MultipleUserAgentsSameIP"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"todsacerdoti"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Should the browser use all available memory?"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://jmmv.dev/2021/08/using-all-memory-as-a-cache.html"}},"_tags":["story","author_todsacerdoti","story_28161538"],"author":"todsacerdoti","children":[28165457,28166429,28167636,28168069],"created_at":"2021-08-12T21:12:04Z","created_at_i":1628802724,"num_comments":10,"objectID":"28161538","points":30,"story_id":28161538,"title":"Should the browser use all available memory?","updated_at":"2024-09-20T09:10:29Z","url":"https://jmmv.dev/2021/08/using-all-memory-as-a-cache.html"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"dsr12"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Hack could let browsers use cloud to carry out big attacks on the cheap"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"http://arstechnica.com/security/2012/11/hack-could-let-browsers-use-cloud-to-carry-out-big-attacks-on-the-cheap/"}},"_tags":["story","author_dsr12","story_4847918"],"author":"dsr12","children":[4848239,4850353],"created_at":"2012-11-29T14:44:48Z","created_at_i":1354200288,"num_comments":3,"objectID":"4847918","points":23,"story_id":4847918,"story_text":"","title":"Hack could let browsers use cloud to carry out big attacks on the cheap","updated_at":"2024-09-19T19:06:23Z","url":"http://arstechnica.com/security/2012/11/hack-could-let-browsers-use-cloud-to-carry-out-big-attacks-on-the-cheap/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"rntn"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Google's Privacy Sandbox accused of misleading Chrome browser users"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.theregister.com/2024/06/13/noyb_gdpr_privacy_sandbox/"}},"_tags":["story","author_rntn","story_40671576"],"author":"rntn","children":[40671704],"created_at":"2024-06-13T16:20:39Z","created_at_i":1718295639,"num_comments":2,"objectID":"40671576","points":23,"story_id":40671576,"title":"Google's Privacy Sandbox accused of misleading Chrome browser users","updated_at":"2024-09-20T17:17:26Z","url":"https://www.theregister.com/2024/06/13/noyb_gdpr_privacy_sandbox/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"smushy"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the browser user-agent string (2008)"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://webaim.org/blog/user-agent-string-history/"}},"_tags":["story","author_smushy","story_46784676"],"author":"smushy","children":[46784681,46786918,46847143],"created_at":"2026-01-27T19:03:45Z","created_at_i":1769540625,"num_comments":2,"objectID":"46784676","points":20,"story_id":46784676,"title":"History of the browser user-agent string (2008)","updated_at":"2026-03-05T23:27:01Z","url":"https://webaim.org/blog/user-agent-string-history/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"jiggle123"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Fixing Failures in Browser-Use Models: Why More Data Isn't Enough"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"https://www.fig.inc/blog/fixing-failures-in-browser-use/"}},"_tags":["story","author_jiggle123","story_48686732"],"author":"jiggle123","children":[48686807],"created_at":"2026-06-26T13:59:09Z","created_at_i":1782482349,"num_comments":1,"objectID":"48686732","points":17,"story_id":48686732,"title":"Fixing Failures in Browser-Use Models: Why More Data Isn't Enough","updated_at":"2026-06-26T14:33:55Z","url":"https://www.fig.inc/blog/fixing-failures-in-browser-use/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"cookiecaper"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Why do web browsers use such bad languages?"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"http://www.deseret-tech.com/journal/why-do-web-browsers-use-such-bad-languages/"}},"_tags":["story","author_cookiecaper","story_7233081"],"author":"cookiecaper","children":[7233654,7233745,7233852,7234145,7234217,7234444,7234870,7235693,7235855],"created_at":"2014-02-13T18:24:53Z","created_at_i":1392315893,"num_comments":9,"objectID":"7233081","points":13,"story_id":7233081,"story_text":"","title":"Why do web browsers use such bad languages?","updated_at":"2024-09-19T20:33:30Z","url":"http://www.deseret-tech.com/journal/why-do-web-browsers-use-such-bad-languages/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"jiggle123"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"GUI-Perturbed: Breaking Browser-Use Models Using Domain Randomization"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"https://www.fig.inc/blog/gui-pertubed-breaking-browser-use-models/"}},"_tags":["story","author_jiggle123","story_48659948"],"author":"jiggle123","created_at":"2026-06-24T13:58:30Z","created_at_i":1782309510,"num_comments":0,"objectID":"48659948","points":12,"story_id":48659948,"title":"GUI-Perturbed: Breaking Browser-Use Models Using Domain Randomization","updated_at":"2026-06-24T17:11:19Z","url":"https://www.fig.inc/blog/gui-pertubed-breaking-browser-use-models/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"stanleydrew"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the browser user-agent string (2008)"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://webaim.org/blog/user-agent-string-history/"}},"_tags":["story","author_stanleydrew","story_2224026"],"author":"stanleydrew","created_at":"2011-02-15T23:51:18Z","created_at_i":1297813878,"num_comments":0,"objectID":"2224026","points":11,"story_id":2224026,"story_text":"","title":"History of the browser user-agent string (2008)","updated_at":"2023-09-06T19:21:27Z","url":"http://webaim.org/blog/user-agent-string-history/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Umesh110"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Open-source browser-use/CUA equivalent for androids: mobile-use"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/runablehq/mobile-use"}},"_tags":["story","author_Umesh110","story_43517243"],"author":"Umesh110","children":[43517244],"created_at":"2025-03-29T17:52:05Z","created_at_i":1743270725,"num_comments":0,"objectID":"43517243","points":8,"story_id":43517243,"title":"Open-source browser-use/CUA equivalent for androids: mobile-use","updated_at":"2025-04-01T12:03:05Z","url":"https://github.com/runablehq/mobile-use"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"maalber"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"We made human-use. Similar to how browser-use connects agents to the web, human-use connects agents to people all over the world in real-time using the Rapidata API. This allows the LLM to crowdsource human feedback and insights when it deems necessary.
Free to use for anyone, you can enable your agent to use humans in real-time to:
- Do preference research\n- Check for hallucinations\n- Capture sentiment\n- Get feedback\n- etc. Whatever you would want from humans.
We expose certain parts of the Rapidata API to the agent through the MCP server framework. Additionally we provide a custom client. More details are available on the repo.
Happy for feedback and to answer any questions."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Show HN: You know browser-use, here's human-use for LLMs to get info from humans"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/RapidataAI/human-use"}},"_tags":["story","author_maalber","story_43983619","show_hn"],"author":"maalber","children":[43983637,43985390],"created_at":"2025-05-14T12:17:44Z","created_at_i":1747225064,"num_comments":4,"objectID":"43983619","points":7,"story_id":43983619,"story_text":"We made human-use. Similar to how browser-use connects agents to the web, human-use connects agents to people all over the world in real-time using the Rapidata API. This allows the LLM to crowdsource human feedback and insights when it deems necessary.
Free to use for anyone, you can enable your agent to use humans in real-time to:
- Do preference research\n- Check for hallucinations\n- Capture sentiment\n- Get feedback\n- etc. Whatever you would want from humans.
We expose certain parts of the Rapidata API to the agent through the MCP server framework. Additionally we provide a custom client. More details are available on the repo.
Happy for feedback and to answer any questions.","title":"Show HN: You know browser-use, here's human-use for LLMs to get info from humans","updated_at":"2025-05-20T05:45:05Z","url":"https://github.com/RapidataAI/human-use"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"wayy"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Hey all, Justin here. I previously built Phind, the AI search engine for developers.
One of the biggest problems we had there was figuring out what went wrong with bad searches. We had tons of searches per day, but less than 1% of users gave any explicit feedback. So we were either manually digging through searches or making general system improvements and hoping they helped.
This problem gets harder with agents. Traces are longer and more complex. It takes more effort to review them, so I'm building a tool that lets you analyze LLM outputs directly to help developers of LLM apps and agents understand where things are breaking and why.
I've put together a demo using browser-use agent traces (gpt-5): https://trails-red.vercel.app/viewer
It's early, but I have lots of ideas - live querying of past failures for currently-running agents, preference models to expand sparse signal data.
Would love feedback on the demo. Also if you're building agents and have 10k+ traces per day that you're not looking at but would like to, I'd love to talk."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Show HN: 83 browser-use trajectories, visualized"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://trails-red.vercel.app/viewer"}},"_tags":["story","author_wayy","story_46736978","show_hn"],"author":"wayy","children":[46739500],"created_at":"2026-01-23T19:50:03Z","created_at_i":1769197803,"num_comments":1,"objectID":"46736978","points":7,"story_id":46736978,"story_text":"Hey all, Justin here. I previously built Phind, the AI search engine for developers.
One of the biggest problems we had there was figuring out what went wrong with bad searches. We had tons of searches per day, but less than 1% of users gave any explicit feedback. So we were either manually digging through searches or making general system improvements and hoping they helped.
This problem gets harder with agents. Traces are longer and more complex. It takes more effort to review them, so I'm building a tool that lets you analyze LLM outputs directly to help developers of LLM apps and agents understand where things are breaking and why.
I've put together a demo using browser-use agent traces (gpt-5): https://trails-red.vercel.app/viewer
It's early, but I have lots of ideas - live querying of past failures for currently-running agents, preference models to expand sparse signal data.
Would love feedback on the demo. Also if you're building agents and have 10k+ traces per day that you're not looking at but would like to, I'd love to talk.","title":"Show HN: 83 browser-use trajectories, visualized","updated_at":"2026-03-05T23:24:17Z","url":"https://trails-red.vercel.app/viewer"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"darien"},"story_text":{"matchLevel":"none","matchedWords":[],"value":"This means about 10% of all firefox browsers on any given day block online ads. (See http://blog.mozilla.com/addons/2009/08/11/how-many-firefox-users-use-add-ons/) Is it right for mozilla to advocate adblock? Is mozilla liable due to promotion? Will the online industry ever address the rising use of adblock a la MPAA/RIAA?"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"9.5mm Firefox browsers use Adblock Plus Daily"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://addons.mozilla.org/en-US/statistics/addon/1865"}},"_tags":["story","author_darien","story_803165"],"author":"darien","children":[803167,803605,804016],"created_at":"2009-09-03T18:41:29Z","created_at_i":1252003289,"num_comments":12,"objectID":"803165","points":5,"story_id":803165,"story_text":"This means about 10% of all firefox browsers on any given day block online ads. (See http://blog.mozilla.com/addons/2009/08/11/how-many-firefox-users-use-add-ons/) Is it right for mozilla to advocate adblock? Is mozilla liable due to promotion? Will the online industry ever address the rising use of adblock a la MPAA/RIAA?","title":"9.5mm Firefox browsers use Adblock Plus Daily","updated_at":"2024-09-19T16:38:17Z","url":"https://addons.mozilla.org/en-US/statistics/addon/1865"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"vortex_ape"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Show HN: Browser-Use MCP for Claude that works without an API key"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://twitter.com/vortex_ape/status/1900953901588729864"}},"_tags":["story","author_vortex_ape","story_43379473","show_hn"],"author":"vortex_ape","children":[43379507],"created_at":"2025-03-16T14:48:18Z","created_at_i":1742136498,"num_comments":5,"objectID":"43379473","points":5,"story_id":43379473,"title":"Show HN: Browser-Use MCP for Claude that works without an API key","updated_at":"2025-03-17T01:26:27Z","url":"https://twitter.com/vortex_ape/status/1900953901588729864"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"signa11"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the browser user-agent string "},"url":{"matchLevel":"none","matchedWords":[],"value":"http://webaim.org/blog/user-agent-string-history/"}},"_tags":["story","author_signa11","story_521138"],"author":"signa11","created_at":"2009-03-18T03:51:57Z","created_at_i":1237348317,"num_comments":0,"objectID":"521138","points":5,"story_id":521138,"story_text":"","title":"History of the browser user-agent string ","updated_at":"2024-09-19T16:30:03Z","url":"http://webaim.org/blog/user-agent-string-history/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"stevep2007"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Top secret firm awarded contract by Russians to expose Tor browser users"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://www.networkworld.com/article/2598401/opensource-subnet/russias-bid-to-expose-users-highlights-law-enforcements-tricky-relationship-with.html#linkedin"}},"_tags":["story","author_stevep2007","story_8224244"],"author":"stevep2007","children":[8224246,8224378],"created_at":"2014-08-25T21:04:39Z","created_at_i":1409000679,"num_comments":2,"objectID":"8224244","points":4,"story_id":8224244,"story_text":"","title":"Top secret firm awarded contract by Russians to expose Tor browser users","updated_at":"2024-09-19T21:04:46Z","url":"http://www.networkworld.com/article/2598401/opensource-subnet/russias-bid-to-expose-users-highlights-law-enforcements-tricky-relationship-with.html#linkedin"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"thepasswordapp"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Show HN: AI agent that rotates your passwords (browser-use and zero-knowledge)"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://thepassword.app/"}},"_tags":["story","author_thepasswordapp","story_46089114","show_hn"],"author":"thepasswordapp","children":[46089413,46091597],"created_at":"2025-11-29T17:15:17Z","created_at_i":1764436517,"num_comments":2,"objectID":"46089114","points":4,"story_id":46089114,"title":"Show HN: AI agent that rotates your passwords (browser-use and zero-knowledge)","updated_at":"2026-03-05T23:10:18Z","url":"https://thepassword.app/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"hsikka"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Breaking Browser-Use Models Using Domain Randomization"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"https://www.fig.inc/blog/gui-pertubed-breaking-browser-use-models/"}},"_tags":["story","author_hsikka","story_48630784"],"author":"hsikka","created_at":"2026-06-22T14:37:48Z","created_at_i":1782139068,"num_comments":0,"objectID":"48630784","points":4,"story_id":48630784,"title":"Breaking Browser-Use Models Using Domain Randomization","updated_at":"2026-06-22T17:15:26Z","url":"https://www.fig.inc/blog/gui-pertubed-breaking-browser-use-models/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"whytai"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Perplexity computer is based on the OSS browser use library"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://twitter.com/mamagnus00/status/2042339700082610345"}},"_tags":["story","author_whytai","story_47710327"],"author":"whytai","created_at":"2026-04-09T21:22:05Z","created_at_i":1775769725,"num_comments":0,"objectID":"47710327","points":4,"story_id":47710327,"title":"Perplexity computer is based on the OSS browser use library","updated_at":"2026-04-09T23:43:50Z","url":"https://twitter.com/mamagnus00/status/2042339700082610345"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"isomierism"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"The Bear Case on Browser Use"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.lomondlabs.com/bear-case-on-browsers"}},"_tags":["story","author_isomierism","story_45235052"],"author":"isomierism","created_at":"2025-09-13T20:27:57Z","created_at_i":1757795277,"num_comments":0,"objectID":"45235052","points":4,"story_id":45235052,"title":"The Bear Case on Browser Use","updated_at":"2026-03-05T22:40:42Z","url":"https://www.lomondlabs.com/bear-case-on-browsers"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"vortex_ape"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Hey everyone,
I built an MCP that enables the Claude desktop app to control your browser with Browser Use (YC W25).
MCPs (based on the Model Context Protocol by Anthropic) enable Claude to use tools that can do a variety of functions like searching the web, executing code, manipulating files, accessing APIs; basically interacting with things outside the conversation context.
The main reason I built this MCP was because I saw that existing Browser Use MCPs required me to get an API key from OpenAI/Anthropic playgrounds in order to use them with the Claude desktop app, which I didn't want to do because I already pay for Claude.
It was fun to read the Browser Use code (https://github.com/browser-use/browser-use) to see how everything works! There's basically an agent (powered by an LLM of your choice) which has access to a bunch of tools. I removed that agent and made the Claude desktop app the agent instead using the Model Context Protocol by Anthropic, and finally gave it access to the Browser Use tools to tie everything together.
Hope you find it useful!"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Show HN: Browser-Use MCP for Claude that works without an API key"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://twitter.com/vortex_ape/status/1900953901588729864"}},"_tags":["story","author_vortex_ape","story_43617033","show_hn"],"author":"vortex_ape","created_at":"2025-04-07T23:38:27Z","created_at_i":1744069107,"num_comments":0,"objectID":"43617033","points":4,"story_id":43617033,"story_text":"Hey everyone,
I built an MCP that enables the Claude desktop app to control your browser with Browser Use (YC W25).
MCPs (based on the Model Context Protocol by Anthropic) enable Claude to use tools that can do a variety of functions like searching the web, executing code, manipulating files, accessing APIs; basically interacting with things outside the conversation context.
The main reason I built this MCP was because I saw that existing Browser Use MCPs required me to get an API key from OpenAI/Anthropic playgrounds in order to use them with the Claude desktop app, which I didn't want to do because I already pay for Claude.
It was fun to read the Browser Use code (https://github.com/browser-use/browser-use) to see how everything works! There's basically an agent (powered by an LLM of your choice) which has access to a bunch of tools. I removed that agent and made the Claude desktop app the agent instead using the Model Context Protocol by Anthropic, and finally gave it access to the Browser Use tools to tie everything together.
Hope you find it useful!","title":"Show HN: Browser-Use MCP for Claude that works without an API key","updated_at":"2025-04-10T15:08:54Z","url":"https://twitter.com/vortex_ape/status/1900953901588729864"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"ledwards"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Amazon Nova Act Browser Use Model"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://labs.amazon.science/"}},"_tags":["story","author_ledwards","story_43538465"],"author":"ledwards","created_at":"2025-03-31T18:58:14Z","created_at_i":1743447494,"num_comments":0,"objectID":"43538465","points":4,"story_id":43538465,"title":"Amazon Nova Act Browser Use Model","updated_at":"2025-03-31T19:33:01Z","url":"https://labs.amazon.science/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"franze"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Browser Use \u2013 Simple Web AI Agents"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"https://github.com/browser-use/browser-use"}},"_tags":["story","author_franze","story_42677425"],"author":"franze","created_at":"2025-01-12T22:04:51Z","created_at_i":1736719491,"num_comments":0,"objectID":"42677425","points":4,"story_id":42677425,"title":"Browser Use \u2013 Simple Web AI Agents","updated_at":"2025-01-13T15:03:12Z","url":"https://github.com/browser-use/browser-use"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Alfagun74"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the Browser User-Agent String (2008)"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://webaim.org/blog/user-agent-string-history/"}},"_tags":["story","author_Alfagun74","story_41808856"],"author":"Alfagun74","children":[41808899],"created_at":"2024-10-11T12:23:15Z","created_at_i":1728649395,"num_comments":0,"objectID":"41808856","points":4,"story_id":41808856,"title":"History of the Browser User-Agent String (2008)","updated_at":"2024-10-11T12:55:19Z","url":"https://webaim.org/blog/user-agent-string-history/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"bryanrasmussen"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Few Brave Browser Users Take Advantage of Crypto Features, Says CEO"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"https://cointelegraph.com/news/few-brave-browser-users-take-advantage-of-crypto-features-says-ceo"}},"_tags":["story","author_bryanrasmussen","story_23081910"],"author":"bryanrasmussen","created_at":"2020-05-05T16:40:08Z","created_at_i":1588696808,"num_comments":0,"objectID":"23081910","points":4,"story_id":23081910,"title":"Few Brave Browser Users Take Advantage of Crypto Features, Says CEO","updated_at":"2024-09-20T06:08:04Z","url":"https://cointelegraph.com/news/few-brave-browser-users-take-advantage-of-crypto-features-says-ceo"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mnmlsm"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Microsoft lost 300M browser users in 2016, tracking site shows"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://www.businessinsider.com/300-million-users-ditched-microsoft-browsers-2016-11?utm_source=feedburner&utm_medium=referral"}},"_tags":["story","author_mnmlsm","story_12871348"],"author":"mnmlsm","created_at":"2016-11-04T09:49:51Z","created_at_i":1478252991,"num_comments":0,"objectID":"12871348","points":4,"story_id":12871348,"title":"Microsoft lost 300M browser users in 2016, tracking site shows","updated_at":"2024-09-19T23:56:43Z","url":"http://www.businessinsider.com/300-million-users-ditched-microsoft-browsers-2016-11?utm_source=feedburner&utm_medium=referral"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"woldemariam"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Why do all browsers' user agents start with \u201cMozilla/\u201d?"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"https://stackoverflow.com/questions/1114254/why-do-all-browsers-user-agents-start-with-mozilla"}},"_tags":["story","author_woldemariam","story_27425988"],"author":"woldemariam","children":[27426059],"created_at":"2021-06-07T18:12:29Z","created_at_i":1623089549,"num_comments":1,"objectID":"27425988","points":3,"story_id":27425988,"title":"Why do all browsers' user agents start with \u201cMozilla/\u201d?","updated_at":"2024-09-20T08:42:20Z","url":"https://stackoverflow.com/questions/1114254/why-do-all-browsers-user-agents-start-with-mozilla"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"thunderbong"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Writing a browser-use agent from scratch \u2013 Part 1/3 \u2013 The Capture"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"https://pdufour.substack.com/p/writing-a-browser-use-agent-from"}},"_tags":["story","author_thunderbong","story_48552718"],"author":"thunderbong","created_at":"2026-06-16T09:29:54Z","created_at_i":1781602194,"num_comments":0,"objectID":"48552718","points":3,"story_id":48552718,"title":"Writing a browser-use agent from scratch \u2013 Part 1/3 \u2013 The Capture","updated_at":"2026-06-16T16:29:35Z","url":"https://pdufour.substack.com/p/writing-a-browser-use-agent-from"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"LorenDB"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the browser user-agent string"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://webaim.org/blog/user-agent-string-history/"}},"_tags":["story","author_LorenDB","story_46543591"],"author":"LorenDB","created_at":"2026-01-08T17:16:54Z","created_at_i":1767892614,"num_comments":0,"objectID":"46543591","points":3,"story_id":46543591,"title":"History of the browser user-agent string","updated_at":"2026-03-05T23:19:21Z","url":"https://webaim.org/blog/user-agent-string-history/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"MagMueller"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"I asked OpenAI's Atlas browser agent:
"""go to browser-use.com and use the computer.get_dom tool. Share\nthe extracted DOM exactly with me."""
The response:\n|SCROLL|<body node_id=9d5f6b01> (vertical view=749px, 0px above, 11932px below)
<a node_id=f9367e7b>\n\n Browser Use\n\n <button node_id=eaeb1667 aria-label="Open menu">\nThen I checked how it clicks: It clicks by node_id (e.g. f9367e7b) and as alternative coordinates.
In browser-use we
1. interact with the DOM by backend_node_id and coordinate fallback
2. use the exact same token for scroll containers with "|" and caps lock (|SCROLL|)
3. use scroll containers with context how much above / below
4. use the same llm representation with \n<tag filtered_attributes>
5. use element texts in new lines with indentation
Things I noticed they could improve:
1. Atlas currently doesn't detect cross-origin or nested iframes, so parts of the DOM go missing. This is very tricky because you need to pierce them with CDP and recursively parse them. (e.g. https://csreis.github.io/tests/cross-site-iframe.html)
2. They waste 10 tokens every item: [tab]<div node_id=83876787. They could cut that to <a id3. (3 Tokens)
3. They keep full links -> They could shorten them easily to save tokens.
4. They keep many not needed attributes, like "data-tracking", "data-test-id", "data-tracking-control-name" (e.g. on LinkedIn.com)
5. For all elements they use [tabs] before which is not needed.
6. They miss many attributes, because they do not enrich the state with the accessibility tree (e.g. for min/max values or hints like required)"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"I reverse engineered OpenAI's Atlas, it uses my open-source library browser-use"}},"_tags":["story","author_MagMueller","story_46019851","ask_hn"],"author":"MagMueller","created_at":"2025-11-23T01:16:34Z","created_at_i":1763860594,"num_comments":0,"objectID":"46019851","points":3,"story_id":46019851,"story_text":"I asked OpenAI's Atlas browser agent:
"""go to browser-use.com and use the computer.get_dom tool. Share\nthe extracted DOM exactly with me."""
The response:\n|SCROLL|<body node_id=9d5f6b01> (vertical view=749px, 0px above, 11932px below)
<a node_id=f9367e7b>\n\n Browser Use\n\n <button node_id=eaeb1667 aria-label="Open menu">\nThen I checked how it clicks: It clicks by node_id (e.g. f9367e7b) and as alternative coordinates.
In browser-use we
1. interact with the DOM by backend_node_id and coordinate fallback
2. use the exact same token for scroll containers with "|" and caps lock (|SCROLL|)
3. use scroll containers with context how much above / below
4. use the same llm representation with \n<tag filtered_attributes>
5. use element texts in new lines with indentation
Things I noticed they could improve:
1. Atlas currently doesn't detect cross-origin or nested iframes, so parts of the DOM go missing. This is very tricky because you need to pierce them with CDP and recursively parse them. (e.g. https://csreis.github.io/tests/cross-site-iframe.html)
2. They waste 10 tokens every item: [tab]<div node_id=83876787. They could cut that to <a id3. (3 Tokens)
3. They keep full links -> They could shorten them easily to save tokens.
4. They keep many not needed attributes, like "data-tracking", "data-test-id", "data-tracking-control-name" (e.g. on LinkedIn.com)
5. For all elements they use [tabs] before which is not needed.
6. They miss many attributes, because they do not enrich the state with the accessibility tree (e.g. for min/max values or hints like required)","title":"I reverse engineered OpenAI's Atlas, it uses my open-source library browser-use","updated_at":"2026-03-05T23:05:31Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"hkpatel3"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Web Agents like Browser-use and Browserbase need a way to authenticate using their own credentials not yours. I made the first OpenID Connect Provider for Web Agents. Please check out the repo and the website auth-agent.com and let me know your thoughts."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Show HN: Web Agents like Browser-use and Browserbase need a way to authenticate"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/auth-agent/auth-agent"}},"_tags":["story","author_hkpatel3","story_45980074","show_hn"],"author":"hkpatel3","created_at":"2025-11-19T14:39:31Z","created_at_i":1763563171,"num_comments":0,"objectID":"45980074","points":3,"story_id":45980074,"story_text":"Web Agents like Browser-use and Browserbase need a way to authenticate using their own credentials not yours. I made the first OpenID Connect Provider for Web Agents. Please check out the repo and the website auth-agent.com and let me know your thoughts.","title":"Show HN: Web Agents like Browser-use and Browserbase need a way to authenticate","updated_at":"2026-03-05T23:02:57Z","url":"https://github.com/auth-agent/auth-agent"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"susam"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the browser user-agent string"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://webaim.org/blog/user-agent-string-history/"}},"_tags":["story","author_susam","story_44354063"],"author":"susam","created_at":"2025-06-23T10:09:20Z","created_at_i":1750673360,"num_comments":0,"objectID":"44354063","points":3,"story_id":44354063,"title":"History of the browser user-agent string","updated_at":"2025-06-23T10:53:01Z","url":"https://webaim.org/blog/user-agent-string-history/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Garbage"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Interesting history of the browser user-agent string"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://webaim.org/blog/user-agent-string-history/?"}},"_tags":["story","author_Garbage","story_3889973"],"author":"Garbage","created_at":"2012-04-25T17:09:43Z","created_at_i":1335373783,"num_comments":0,"objectID":"3889973","points":3,"story_id":3889973,"story_text":"","title":"Interesting history of the browser user-agent string","updated_at":"2024-09-19T18:32:06Z","url":"http://webaim.org/blog/user-agent-string-history/?"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Garbage"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"History of the browser user-agent string"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://webaim.org/blog/user-agent-string-history/?"}},"_tags":["story","author_Garbage","story_14521917"],"author":"Garbage","created_at":"2017-06-09T15:08:37Z","created_at_i":1497020917,"num_comments":0,"objectID":"14521917","points":3,"story_id":14521917,"title":"History of the browser user-agent string","updated_at":"2024-09-20T00:54:41Z","url":"http://webaim.org/blog/user-agent-string-history/?"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"tiny-automates"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["browser","use"],"value":"Security audit of Browser Use: prompt injection, credential exfil, domain bypass"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://arxiv.org/abs/2505.13076"}},"_tags":["story","author_tiny-automates","story_46955396"],"author":"tiny-automates","children":[46955397],"created_at":"2026-02-10T04:28:37Z","created_at_i":1770697717,"num_comments":1,"objectID":"46955396","points":2,"story_id":46955396,"title":"Security audit of Browser Use: prompt injection, credential exfil, domain bypass","updated_at":"2026-03-05T23:33:41Z","url":"https://arxiv.org/abs/2505.13076"}],"hitsPerPage":50,"nbHits":252,"nbPages":6,"page":0,"params":"query=browser-use&tags=story&hitsPerPage=50&advancedSyntax=true&analyticsTags=backend","processingTimeMS":26,"processingTimingsMS":{"_request":{"roundTrip":19},"afterFetch":{"format":{"highlighting":1,"total":1},"merge":{"mergeLoop":{"prepareNextHit":1,"total":1},"total":1},"total":1},"fetch":{"query":7,"scanning":15,"total":23},"total":26},"query":"browser-use","serverTimeMS":28}