{"id":"pNItkmxeRaQ","title":"GPT 5.2 Best Model Yet? AGI Check, AI Agents Outperform 9/10 Security Professionals in Testing","description":"Join the NEAR Legion at https://nearlegion.com/\n\nFollow us on X:   / nearlegion  \n\n#NEAR #nearprotocol #NEARAI #ai #ainews","channel_name":"NEAR Protocol","published_at":"2025-12-12T00:00:00.000Z","thumbnail_url":"https://img.youtube.com/vi/pNItkmxeRaQ/hqdefault.jpg","transcript_text":"Good morning Legion. Happy Friday. J1 kicks off today with over 70 teams ready to race. As a reminder, this is only the beginning of championship season, so there will be many more races to come. Our first story today should be obvious GPT 5.2 is rolled out with OpenAI dropping the latest evals. and looking at this graphic, GPT 5.2 is state of the art all around very powerful, outperforming both Anthropic and Google in most major categories. In spite of all the skepticism, Sam Altman continues to deliver. Our second story is actually a very nice, albeit technical, macro commentary on what is actually changing in the Western AI race. Give the full thread by Akash Gupta a read, but for now, focus in on this specific paragraph. If you're building products in 2025 and assuming AI reasoning is expensive, you're building for a world that no longer exists. The benchmark that was supposed to prove AI couldn't generalize just became another line item on a pricing page. 390 times efficiency improvement in one year. This is exactly the context that tells us that despite many skeptics to the contrary, the race to AGI remains alive and well with massive performance improvements year over year. Automation, replacement of white collar jobs and hints of NEARVANA remain on schedule. Our last story today has to do with a paper published by a group of Stanford researchers who compared head to head, security professionals and AI agents in a real world scenario. This type of study might also be particularly relevant for teams looking to get a product audited in the future. You can just use these agents. The AI agents outperform nine out of ten professional penetration testers in the study, paving the way towards proving that security testing will likely not be spared from the AI future. With that in mind, the paper did clarify that there remain a number of capability gaps among these agents. The agents exhibit higher false positive rate and struggle with GUI based tasks. At least there's still some time before even these get fixed. We hope you have a wonderful weekend. We'll do this all again real soon.","transcript_synced_at":"2026-11-03T04:45:53.391Z","segments":[{"start_seconds":0,"text":"Good morning Legion. Happy Friday.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=0s"},{"start_seconds":2,"text":"J1 kicks off today with over 70 teams ready to race.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=2s"},{"start_seconds":6,"text":"As a reminder, this is only the beginning of championship season, so there will be many more races to come.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=6s"},{"start_seconds":11,"text":"Our first story today should be obvious GPT 5.2 is rolled out with OpenAI dropping the latest evals.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=11s"},{"start_seconds":18,"text":"and looking at this graphic, GPT 5.2 is state of the art all around very powerful, outperforming both Anthropic and Google in most major categories.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=18s"},{"start_seconds":27,"text":"In spite of all the skepticism, Sam Altman continues to deliver.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=27s"},{"start_seconds":31,"text":"Our second story is actually a very nice, albeit technical, macro commentary on what is actually changing in the Western AI race.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=31s"},{"start_seconds":38,"text":"Give the full thread by Akash Gupta a read, but for now, focus in on this specific paragraph.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=38s"},{"start_seconds":43,"text":"If you're building products in 2025 and assuming AI reasoning is expensive, you're building for a world that no longer exists.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=43s"},{"start_seconds":50,"text":"The benchmark that was supposed to prove AI couldn't generalize just became another line item on a pricing page.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=50s"},{"start_seconds":56,"text":"390 times efficiency improvement in one year.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=56s"},{"start_seconds":60,"text":"This is exactly the context that tells us that despite many skeptics to the contrary, the race to AGI remains alive and well with massive performance improvements year over year.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=60s"},{"start_seconds":70,"text":"Automation, replacement of white collar jobs and hints of NEARVANA remain on schedule.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=70s"},{"start_seconds":75,"text":"Our last story today has to do with a paper published by a group of Stanford researchers who compared head to head, security professionals and AI agents in a real world scenario.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=75s"},{"start_seconds":84,"text":"This type of study might also be particularly relevant for teams looking to get a product audited in the future. You can just use these agents.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=84s"},{"start_seconds":91,"text":"The AI agents outperform nine out of ten professional penetration testers in the study, paving the way towards proving that security testing will likely not be spared from the AI future.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=91s"},{"start_seconds":101,"text":"With that in mind, the paper did clarify that there remain a number of capability gaps among these agents.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=101s"},{"start_seconds":107,"text":"The agents exhibit higher false positive rate and struggle with GUI based tasks.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=107s"},{"start_seconds":112,"text":"At least there's still some time before even these get fixed. We hope you have a wonderful weekend. We'll do this all again real soon.","url":"https://www.youtube.com/watch?v=pNItkmxeRaQ&t=112s"}],"youtube_url":"https://youtube.com/watch?v=pNItkmxeRaQ","url":"https://near-pedia.xyz/videos/pNItkmxeRaQ"}