Gemini 3 destroyed all other LLMs
| \"''''"\\'"\'\"\'"\' | 11/24/25 | | christmas online | 11/24/25 | | \"''''"\\'"\'\"\'"\' | 11/24/25 | | thankman | 11/24/25 | | Hallmark Channel white chad driving pickup truck | 11/24/25 | | christmas online | 11/24/25 | | Hallmark Channel white chad driving pickup truck | 11/24/25 | | ,.,....,..,.,.,,,,..,..,.,..,.,.,.,... | 11/25/25 | | ,.,....,..,.,.,,,,..,..,.,..,.,.,.,... | 11/24/25 | | thankman | 11/24/25 | | crucified baby yeshua | 11/24/25 | | christmas online | 11/24/25 | | crucified baby yeshua | 11/24/25 | | christmas online | 11/24/25 | | crucified baby yeshua | 11/24/25 | | christmas online | 11/24/25 | | t.rick o'treat panama | 11/25/25 | | .,.,...,..,.,.,:,,:,...,:::,...,:,.,.:..:. | 11/24/25 | | metaepistemology is trans | 11/24/25 | | thankman | 11/24/25 | | UN peacekeeper | 11/24/25 | | thankman | 11/24/25 | | .,.,...,..,.,.,:,,:,...,:::,...,:,.,.:..:. | 11/25/25 | | .,.,...,..,.,.,:,,:,...,:::,...,:,.,.:..:. | 11/25/25 | | thankman | 11/25/25 | | metaepistemology is trans | 11/24/25 | | Die Hard 2: Die Harder | 11/24/25 | | christmas online | 11/24/25 | | metaepistemology is trans | 11/24/25 | | https://imgur.com/a/o2g8xYK | 11/24/25 | | crucified baby yeshua | 11/25/25 | | Smoker | 11/25/25 | | Mainlining the $ecret Truth of the Univer$e | 11/25/25 | | metaepistemology is trans | 11/25/25 | | Mainlining the $ecret Truth of the Univer$e | 11/25/25 | | ,.,....,..,.,.,,,,..,..,.,..,.,.,.,... | 11/25/25 | | Smoker | 11/25/25 | | \"''''"\\'"\'\"\'"\' | 11/25/25 | | ,.,....,..,.,.,,,,..,..,.,..,.,.,.,... | 11/26/25 | | Mainlining the $ecret Truth of the Univer$e | 11/27/25 | | metaepistemology is trans | 11/27/25 | | Patel Philippe | 11/27/25 | | metaepistemology is trans | 11/27/25 | | ,.,....,..,.,.,,,,..,..,.,..,.,.,.,... | 11/27/25 | | average/ordinary/typical citizen/person | 11/27/25 | | thankman | 11/27/25 |
Poast new message in this thread
 |
Date: November 25th, 2025 10:03 AM
Author: ,.,....,..,.,.,,,,..,..,.,..,.,.,.,...
(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2)#49458928) |
Date: November 24th, 2025 4:13 PM
Author: ,.,....,..,.,.,,,,..,..,.,..,.,.,.,...
looks like Opus 4.5 might be a better coding model now. hard to know for sure, but it's close. the rate of advancement here is nuts. pretty decent improvement over Sonnet 4.5, which isn't even 2 months old.
https://www.anthropic.com/news/claude-opus-4-5
(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2)#49457074)
|
 |
Date: November 24th, 2025 6:35 PM
Author: .,.,...,..,.,.,:,,:,...,:::,...,:,.,.:..:.
I am skeptical about the value of comparing models on benchmarks where they cluster together at high scores. There is not much room to detect model differences when they are at 90 %
(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2)#49457525) |
Date: November 24th, 2025 6:39 PM Author: thankman
Nano Banana Pro
Create an image at 31.7785° N, 35.2296° E, April 3, 33 CE, 15:00 hours
Thinking...
Examining the Details
I'm currently focused on the specific coordinates and date provided. My goal is to produce an image reflective of that location and timeframe. I am now working on identifying relevant historical data points.
Pinpointing the Scene
I'm now zeroing in on the specific elements needed for the historical recreation. The latitude and longitude lead me to Jerusalem, and the date focuses me on a well-documented event. I'm building a picture of the scene, considering details like Roman soldiers, a crowd, and the crosses on a hill, all against a fittingly somber backdrop.
Analyzing Visual Depiction
I'm currently reviewing the match between the visual elements and the user's input, specifically focusing on the accurate representation of the specified time, date, and location. I am ensuring that all the parameters are correctly translated into the visual depiction. The focus is on the correct elements of the crucifixion scene itself.
Assessing Scene Correspondence
I've determined that the generated image aligns completely with the user's input. The specified historical setting, time, and date are all evident in the scene. The elements described in the request are correctly reflected in the visual output. The historical context, Roman soldiers, and onlookers are consistent with the prompt.
https://imgur.com/a/ZYhQE7s
(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2)#49457550)
|
 |
Date: November 25th, 2025 1:15 AM
Author: .,.,...,..,.,.,:,,:,...,:::,...,:,.,.:..:.
It’s so much better than other generators.
(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2)#49458420) |
 |
Date: November 25th, 2025 1:20 AM
Author: .,.,...,..,.,.,:,,:,...,:::,...,:,.,.:..:.
https://x.com/karpathy/status/1992655330002817095?s=46
They must have trained it with Gemini because it can solve complex technical problems in the image generation process.
(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2)#49458422) |
 |
Date: November 25th, 2025 10:02 AM
Author: ,.,....,..,.,.,,,,..,..,.,..,.,.,.,...
cr. this is why I enjoy vibe coding even if doesn’t save time. I would rather just keep playing the code lottery and ask it to regenerate than have to think
(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2)#49458925) |
 |
Date: November 26th, 2025 12:24 AM
Author: ,.,....,..,.,.,,,,..,..,.,..,.,.,.,...
TPUs can do anything anyone would want to do with neural networks.
(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2)#49461564) |
Date: November 27th, 2025 3:08 AM Author: Patel Philippe
This has to be a joke. Other than synthesizing massive information contexts for simplistic Google search style prompting, or analyzing a video on YouTube in detail, it is still complete trash compared to GPT and now Claude
Still no projects or OEM prompt optimization tool. Looks like they removed agentic prompting
Today it could not review a 4-page PDF in order after 4 tries (kept saying something on page 1 was on page 3).
Tempted to try Ultra but honestly why bother
(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2)#49465146) |
 |
Date: November 27th, 2025 11:52 AM
Author: ,.,....,..,.,.,,,,..,..,.,..,.,.,.,...
There seems to be a limited amount of benchmark gaming going on. Private benchmarks or new benchmarks released after the models come out usually produce a rank ordering similar to other benchmarks. The labs have an incentive to be straight about benchmarks because people quickly figure out what models actually work and they won’t trust companies that consistently game them.
(http://www.autoadmit.com/thread.php?thread_id=5801840&forum_id=2)#49465528) |
|
|