Two Minute Papers (@twominutepapers) header
Two Minute Papers (@twominutepapers) profile picture

🙌📜 What a time to be alive!

Joined January 2015
Photo by Two Minute Papers (@twominutepapers)Photo by Two Minute Papers (@twominutepapers)Photo by Two Minute Papers (@twominutepapers)Photo by Two Minute Papers (@twominutepapers)Photo by Two Minute Papers (@twominutepapers)Photo by Two Minute Papers (@twominutepapers)Photo by Two Minute Papers (@twominutepapers)Photo by Two Minute Papers (@twominutepapers)Photo by Two Minute Papers (@twominutepapers)Photo by Two Minute Papers (@twominutepapers)
Really excited to try Gemini 4 Argon, but it does not appear to be available in Antigravity with the Ultra subscription. The announcement says it should be there. Am I too early?
7
19
3,743
These comments really made my day. Thank you so much, Fellow Scholars! 🙏
2
8
3,378
Seeing Klangphonics today and man, they are so good!
1
9
2,733
A beautiful sight, isn't it? 🙂
2
1
26
3,706
Wow
Full benchmark results for GLM-5.3
3
2
45
7,157
🙂🙏
This quoted post is unavailable.
1
1
6
3,132
This is amazing. Jeff is the best!
In 2022, as a Google intern, I sent a random calendar invite to Jeff Dean. He accepted. His calendar was accessible even to interns, so I figured I had nothing to lose. The next day, I sent this photo to my team. Everyone was shocked. It’s hard to imagine Google without Jeff. He’s been one of my biggest inspirations. Thank you for everything, Jeff.
2
2
28
8,165
Another DeepSeek moment has arrived. The new Flash model is way better than the previous one. Many of these results more than doubled. One went 7x. In just one revision. What happened? How did they do this black magic? Full video: youtube.com/watch?v=bm1BjOjS…
6
1
14
3,024
Incredible AI parkour simulation from @NVIDIAAI. It learned to be a pro from looking at just 30 seconds of footage! ▶️Full video: youtu.be/8B05cy3UuSE
7
30
1
249
28,905
Compiling your own Linux kernel makes for the most blissful Saturday morning experience.
5
2
45
3,679
Claude AI knows more than it tells you. Here's how they found out about it. Full video: youtube.com/watch?v=l72ufA-4…
2
4
20
3,713
After I think about 10 years of waiting, a new @Jonathan_Blow game demo is available! Not having a Windows machine, just found out that @SinkingStarGame is on GeForce Now and I am unreasonably happy about this! I get to try it, yipee! 🥳
4
1
44
4,777
NVIDIA's Nemotron 3 Ultra AI is a gift to all of us. I ran my experiments and in my cases, found the following: - Blazing fast 🔥 - Agent stuff ✅ - Math and science ✅ - Hard coding tasks 😅 Full video: youtube.com/watch?v=zJvN8PDX…
1
6
1
94
18,396
If this is true this is stunning.
Kimi 2.7 ranked 2nd after Fable 5 and before GPT-5 xhigh We have re-run our ErdosBench smoke test on 14 problems with Kimi 2.7, Qwen 3.7 Max, Grok 4.3 and compared it with the top performers from previous runs. Kimi 2.7 is amazingly good. More below.
2
6
97
14,405
Let's goooooo
Meet DiffusionGemma! An experimental open model that explores a fast approach to text generation, released under an Apache 2.0 license. Moving beyond sequential, token-by-token processes to generate entire blocks of text simultaneously. Here’s what’s new with DiffusionGemma: 👇
2
3
34
2,821
Loving this
“AI agents will outperform humans at almost all jobs by 2026–2027.” - The forecast is everywhere. So we built the exam to test that claim, on real labor-market aligned work. On the hardest tier, top agents pass 2.6%. Meet Agents' Last Exam (ALE), a rolling benchmark measuring whether agents can actually do real jobs. 🧵👇
2
17
2,915
Is Claude Opus 4.8 still hiding its mistakes from us?
1
2
1
22
2,749
Internal thought process of an LLM I'm running. Chop chop, little AI! 🙂
2
32
2,560
I spent a few days testing NVIDIA's Nemotron 3 Ultra AI. 550 billion parameters running *blazing fast* so we had quite a few interesting experiments with it. Video hopefully on our channel soon-ish!
Today we're shipping Nemotron 3 Ultra. A 550B MoE frontier-intelligence open model built for long-running agents. It delivers 5x faster inference and lowers the cost of complex agentic tasks by up to 30% versus other open frontier models.
4
2
34
5,529