When using Opus or Codex vs. a chinese or Open source model, it feels like its reasoning capabilities are basically the same.
The difference is typically in coding. It looks like OpenAI and Anthropic invest a lot in pre-training (paying Mercor and the like).
Also a lot in creating synthetic data, I believe this has bigger AI research involvement and techniques.
Of course, there's the RLHF loop that developers using Anthropic/OpenAI products as well, which provides probably yields very good data.
This ends up creating the perspective that it is smart, after all, it has been trained with what you want to do, so it can do that for you.
But overall, is there really much AI research being done on those companies, or are the AI researchers mostly fine-tuning small aspects of the model, akin to what Google engineers used to do for Google search?
I ask this because this all looks like somebody with money could throw money at the problem and end up with a better model at the end, provided they do what I outlined above better -- with AI research being really not that important.
It still often feels like talking with ChatGPT 4 with just better data.
Even the big upgrade of Claude Code being able to work autonomously looks to be mainly due to it knowing how to grab context and do tool calls (not saying that this is easy), rather than the model's raw performance being better.
Or am I wrong, is there something extremely good on those models that AI researchers discovered that the others don't have? Or is it really mostly Data?"},"title":{"matchLevel":"none","matchedWords":[],"value":"Ask HN: Is the ongoing AI research driving LLM models to be better?"}},"_tags":["story","author_thiago_fm","story_47872916","ask_hn"],"author":"thiago_fm","children":[47873222],"created_at":"2026-04-23T07:03:22Z","created_at_i":1776927802,"num_comments":4,"objectID":"47872916","points":3,"story_id":47872916,"story_text":"I'm just a curious hobbyist that has ran LLM models locally and follow a lot of content about it. Hope we have a few AI researchers here on HN to clarify this.
When using Opus or Codex vs. a chinese or Open source model, it feels like its reasoning capabilities are basically the same.
The difference is typically in coding. It looks like OpenAI and Anthropic invest a lot in pre-training (paying Mercor and the like).
Also a lot in creating synthetic data, I believe this has bigger AI research involvement and techniques.
Of course, there's the RLHF loop that developers using Anthropic/OpenAI products as well, which provides probably yields very good data.
This ends up creating the perspective that it is smart, after all, it has been trained with what you want to do, so it can do that for you.
But overall, is there really much AI research being done on those companies, or are the AI researchers mostly fine-tuning small aspects of the model, akin to what Google engineers used to do for Google search?
I ask this because this all looks like somebody with money could throw money at the problem and end up with a better model at the end, provided they do what I outlined above better -- with AI research being really not that important.
It still often feels like talking with ChatGPT 4 with just better data.
Even the big upgrade of Claude Code being able to work autonomously looks to be mainly due to it knowing how to grab context and do tool calls (not saying that this is easy), rather than the model's raw performance being better.
Or am I wrong, is there something extremely good on those models that AI researchers discovered that the others don't have? Or is it really mostly Data?","title":"Ask HN: Is the ongoing AI research driving LLM models to be better?","updated_at":"2026-04-23T10:00:54Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"yogoism"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"While big tech often outsources data annotation to firms like Scale AI, TURING, and Mercor, \ncompanies such as Tesla and Google run in-house teams.
Which approach do you think is better for AI and robotics development, and how will this trend evolve?
Please share your data annotation insights and experiences."},"title":{"matchLevel":"none","matchedWords":[],"value":"Ask HN: In-house or outsourced data annotation? (2025)"}},"_tags":["story","author_yogoism","story_44209338","ask_hn"],"author":"yogoism","children":[44214562,44214563],"created_at":"2025-06-07T12:55:33Z","created_at_i":1749300933,"num_comments":4,"objectID":"44209338","points":3,"story_id":44209338,"story_text":"While big tech often outsources data annotation to firms like Scale AI, TURING, and Mercor, \ncompanies such as Tesla and Google run in-house teams.
Which approach do you think is better for AI and robotics development, and how will this trend evolve?
Please share your data annotation insights and experiences.","title":"Ask HN: In-house or outsourced data annotation? (2025)","updated_at":"2025-06-09T13:05:27Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"_ocin"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"I know the job marker is pretty rough these days, but still, I'm wondering.
Is it possible for a western company to hire me, as an African, perhaps remotely?
I have been lured by websites like Mercor, etc... offering nice remote work, but I figured out that they mostly wanted my data.
Is it possible for me to get hired at all or should I just try entrepreneurship and mainly focus on copying ideas that worked in the western world for the African market?"},"title":{"matchLevel":"none","matchedWords":[],"value":"Ask HN: Is it possible to get hired as an African software engineer"}},"_tags":["story","author__ocin","story_47998209","ask_hn"],"author":"_ocin","children":[47998374,48014332],"created_at":"2026-05-03T15:54:44Z","created_at_i":1777823684,"num_comments":2,"objectID":"47998209","points":3,"story_id":47998209,"story_text":"I know the job marker is pretty rough these days, but still, I'm wondering.
Is it possible for a western company to hire me, as an African, perhaps remotely?
I have been lured by websites like Mercor, etc... offering nice remote work, but I figured out that they mostly wanted my data.
Is it possible for me to get hired at all or should I just try entrepreneurship and mainly focus on copying ideas that worked in the western world for the African market?","title":"Ask HN: Is it possible to get hired as an African software engineer","updated_at":"2026-05-06T07:01:55Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"ozgurozkan"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"Automate red teaming of your voice AI agent.
In YC terms Request for startups:\nFull stack voice AI Red Teamer:\nAutomates the work below https://work.mercor.com/jobs/list_AAABmX4FiBxRn-I1VM5KeZ2h"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Agentic Adversarial testing for your voice AI"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://audn.ai"}},"_tags":["story","author_ozgurozkan","story_45921051","show_hn"],"author":"ozgurozkan","created_at":"2025-11-13T21:48:27Z","created_at_i":1763070507,"num_comments":0,"objectID":"45921051","points":1,"story_id":45921051,"story_text":"Automate red teaming of your voice AI agent.
In YC terms Request for startups:\nFull stack voice AI Red Teamer:\nAutomates the work below https://work.mercor.com/jobs/list_AAABmX4FiBxRn-I1VM5KeZ2h","title":"Show HN: Agentic Adversarial testing for your voice AI","updated_at":"2026-07-26T15:40:44Z","url":"https://audn.ai"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"JeanKage"},"title":{"matchLevel":"none","matchedWords":[],"value":"You could be next The lawyers and scientists training AI to steal their careers"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://www.theverge.com/cs/features/877388/white-collar-workers-training-ai-mercor"}},"_tags":["story","author_JeanKage","story_47321047"],"author":"JeanKage","created_at":"2026-03-10T09:46:21Z","created_at_i":1773135981,"num_comments":0,"objectID":"47321047","points":12,"story_id":47321047,"title":"You could be next The lawyers and scientists training AI to steal their careers","updated_at":"2026-03-14T19:19:59Z","url":"https://www.theverge.com/cs/features/877388/white-collar-workers-training-ai-mercor"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"rvz"},"title":{"matchLevel":"none","matchedWords":[],"value":"You Could Be Next"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://www.theverge.com/cs/features/877388/white-collar-workers-training-ai-mercor"}},"_tags":["story","author_rvz","story_47347216"],"author":"rvz","created_at":"2026-03-12T06:32:39Z","created_at_i":1773297159,"num_comments":0,"objectID":"47347216","points":5,"story_id":47347216,"title":"You Could Be Next","updated_at":"2026-03-13T21:34:26Z","url":"https://www.theverge.com/cs/features/877388/white-collar-workers-training-ai-mercor"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Michelangelo11"},"title":{"matchLevel":"none","matchedWords":[],"value":"You Could Be Next"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://www.theverge.com/cs/features/877388/white-collar-workers-training-ai-mercor"}},"_tags":["story","author_Michelangelo11","story_47411247"],"author":"Michelangelo11","created_at":"2026-03-17T11:30:22Z","created_at_i":1773747022,"num_comments":0,"objectID":"47411247","points":4,"story_id":47411247,"title":"You Could Be Next","updated_at":"2026-03-17T18:16:27Z","url":"https://www.theverge.com/cs/features/877388/white-collar-workers-training-ai-mercor"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"doppp"},"title":{"matchLevel":"none","matchedWords":[],"value":"22-Year-Olds Are Now the Youngest Self-Made Billionaires"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://www.forbes.com/sites/richardnieva/2025/10/30/mercor-youngest-self-made-billionaires/"}},"_tags":["story","author_doppp","story_45798695"],"author":"doppp","created_at":"2025-11-03T13:22:53Z","created_at_i":1762176173,"num_comments":0,"objectID":"45798695","points":3,"story_id":45798695,"title":"22-Year-Olds Are Now the Youngest Self-Made Billionaires","updated_at":"2026-03-05T23:01:03Z","url":"https://www.forbes.com/sites/richardnieva/2025/10/30/mercor-youngest-self-made-billionaires/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"paulpauper"},"title":{"matchLevel":"none","matchedWords":[],"value":"Feeding the Machine"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://www.theverge.com/cs/features/831818/ai-mercor-handshake-scale-surge-staffing-companies"}},"_tags":["story","author_paulpauper","story_46293355"],"author":"paulpauper","created_at":"2025-12-16T19:40:27Z","created_at_i":1765914027,"num_comments":0,"objectID":"46293355","points":2,"story_id":46293355,"title":"Feeding the Machine","updated_at":"2026-03-05T23:13:44Z","url":"https://www.theverge.com/cs/features/831818/ai-mercor-handshake-scale-surge-staffing-companies"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"gk1"},"title":{"matchLevel":"none","matchedWords":[],"value":"Scale AI sues Mercer for trade secret theft"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://www.bloomberg.com/news/articles/2025-09-03/scale-ai-sues-rival-startup-mercor-alleging-trade-secret-theft"}},"_tags":["story","author_gk1","story_45119084"],"author":"gk1","created_at":"2025-09-03T18:40:10Z","created_at_i":1756924810,"num_comments":0,"objectID":"45119084","points":2,"story_id":45119084,"title":"Scale AI sues Mercer for trade secret theft","updated_at":"2026-03-05T22:38:04Z","url":"https://www.bloomberg.com/news/articles/2025-09-03/scale-ai-sues-rival-startup-mercor-alleging-trade-secret-theft"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"rhttrhtrq"},"title":{"matchLevel":"none","matchedWords":[],"value":"AI Job"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://work.mercor.com/jobs/list_AAABl9D1yGSu5jR9BrtP-5nX?referralCode=35d084dc-8a89-4237-8562-9797f845044a&utm_source=referral&utm_medium=share&utm_campaign=job_referral"}},"_tags":["story","author_rhttrhtrq","story_44466254"],"author":"rhttrhtrq","children":[44466255],"created_at":"2025-07-04T17:16:00Z","created_at_i":1751649360,"num_comments":0,"objectID":"44466254","points":2,"story_id":44466254,"title":"AI Job","updated_at":"2025-07-04T19:47:30Z","url":"https://work.mercor.com/jobs/list_AAABl9D1yGSu5jR9BrtP-5nX?referralCode=35d084dc-8a89-4237-8562-9797f845044a&utm_source=referral&utm_medium=share&utm_campaign=job_referral"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"gmays"},"title":{"matchLevel":"none","matchedWords":[],"value":"AI Can't Read an Investor Deck"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://www.mercor.com/blog/Finance-tasks-ai-failures-modes/"}},"_tags":["story","author_gmays","story_47735664"],"author":"gmays","children":[47735701,47735884],"created_at":"2026-04-12T02:27:17Z","created_at_i":1775960837,"num_comments":2,"objectID":"47735664","points":1,"story_id":47735664,"title":"AI Can't Read an Investor Deck","updated_at":"2026-04-13T13:36:33Z","url":"https://www.mercor.com/blog/Finance-tasks-ai-failures-modes/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Reviewerfy"},"title":{"matchLevel":"none","matchedWords":[],"value":"Research and Data Quality Reviewer"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://work.mercor.com/jobs/list_AAABmNNqhzophcucz_5Ny48q"}},"_tags":["story","author_Reviewerfy","story_45024660"],"author":"Reviewerfy","children":[45024661],"created_at":"2025-08-26T10:31:27Z","created_at_i":1756204287,"num_comments":1,"objectID":"45024660","points":1,"story_id":45024660,"title":"Research and Data Quality Reviewer","updated_at":"2026-03-05T22:31:01Z","url":"https://work.mercor.com/jobs/list_AAABmNNqhzophcucz_5Ny48q"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"amai"},"title":{"matchLevel":"none","matchedWords":[],"value":"Feeding the Machine"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://www.theverge.com/cs/features/831818/ai-mercor-handshake-scale-surge-staffing-companies"}},"_tags":["story","author_amai","story_48068677"],"author":"amai","created_at":"2026-05-08T20:55:41Z","created_at_i":1778273741,"num_comments":0,"objectID":"48068677","points":1,"story_id":48068677,"title":"Feeding the Machine","updated_at":"2026-05-08T21:00:20Z","url":"https://www.theverge.com/cs/features/831818/ai-mercor-handshake-scale-surge-staffing-companies"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"handfuloflight"},"title":{"matchLevel":"none","matchedWords":[],"value":"Apex-Agents"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://www.mercor.com/blog/introducing-apex-agents/"}},"_tags":["story","author_handfuloflight","story_47111725"],"author":"handfuloflight","created_at":"2026-02-22T15:23:43Z","created_at_i":1771773823,"num_comments":0,"objectID":"47111725","points":1,"story_id":47111725,"title":"Apex-Agents","updated_at":"2026-03-05T23:34:54Z","url":"https://www.mercor.com/blog/introducing-apex-agents/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Terretta"},"title":{"matchLevel":"none","matchedWords":[],"value":"Apex: AI Productivity Index for investment banking, law, consulting, medical"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"https://mercor.com/blog/introducing-apex-ai-productivity-index/"}},"_tags":["story","author_Terretta","story_45461962"],"author":"Terretta","created_at":"2025-10-03T12:04:57Z","created_at_i":1759493097,"num_comments":0,"objectID":"45461962","points":1,"story_id":45461962,"title":"Apex: AI Productivity Index for investment banking, law, consulting, medical","updated_at":"2026-03-05T22:49:25Z","url":"https://mercor.com/blog/introducing-apex-ai-productivity-index/"},{"_highlightResult":{"author":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["mercor"],"value":"mercora"},"title":{"matchLevel":"none","matchedWords":[],"value":"Data leak on Europol terrorism investigations"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://www.reuters.com/article/us-europol-cyber-idUSKBN13P0M6"}},"_tags":["story","author_mercora","story_13074380"],"author":"mercora","children":[13074804],"created_at":"2016-11-30T20:25:46Z","created_at_i":1480537546,"num_comments":3,"objectID":"13074380","points":44,"story_id":13074380,"title":"Data leak on Europol terrorism investigations","updated_at":"2024-09-20T00:04:55Z","url":"http://www.reuters.com/article/us-europol-cyber-idUSKBN13P0M6"}],"hitsPerPage":50,"nbHits":3869,"nbPages":20,"page":0,"params":"query=Mercor&tags=story&hitsPerPage=50&advancedSyntax=true&analyticsTags=backend","processingTimeMS":9,"processingTimingsMS":{"_request":{"roundTrip":17},"afterFetch":{"merge":{"mergeLoop":{"prepareNextHit":5,"total":5},"total":5},"total":5},"fetch":{"query":1,"scanning":1,"total":3},"total":9},"query":"Mercor","serverTimeMS":11}