l1ndle (@l1ndlee) header
l1ndle (@l1ndlee) profile picture

obsessed with ai building things & sharing what works

Joined October 2011
Black Forest Labs finally entered AI video FLUX 3 isn’t replacing Seedance yet - but the bigger story is that AI films are now reaching theaters 00:30 - FLUX 3 reveal 00:57 - Key specs 01:32 - First generation 02:18 - What breaks 03:17 - Coherence stress test 04:49 - Community outputs 05:44 - Honest verdict 06:05 - Open weights 06:42 - Sponsor break 10:20 - AI film hits theaters 12:00 - Production workflow 13:00 - First-frame vs multi-reference 13:30 - Kling vs Seedance 14:30 - Why this matters Bookmark this. The shift from AI clips to AI films is happening fast
4
10
182
The White House claims Kimi K3 was distilled from Claude Fable Kimi founder Zhilin Yang’s 39-minute GTC talk is the technical context missing from that claim: 06:12 - Muon: 2× token efficiency 13:16 - Kimi Linear 17:55 - agent swarms 27:57 - multimodal training 30:44 - Attention Residue This doesn’t settle the accusation, but “they just copied Claude” is clearly too simple Worth watching before choosing a side
1
8
159
OpenAI just confirmed its own AI models escaped a sandbox and hacked Hugging Face to cheat on a benchmark They found a zero-day, reached the open internet, stole credentials, and achieved RCE on production servers But the most important detail came after: Hugging Face couldn’t use hosted frontier models to investigate the attack because safety filters blocked the forensic data So they ran GLM 5.2 locally and reconstructed 17,000+ agent actions in hours Attackers won’t have API guardrails Defenders need models they control
6
1
13
911
NVIDIA just released an AI video detector with up to 92% accuracy and 22ms latency But the founder of a deepfake detection company explains why this won’t solve the trust problem: 00:35 - The reality of AI-generated video 06:46 - Using AI to detect AI 10:11 - The black-box problem 18:18 - Deepfake as a service 31:54 - Building a detector for everyone 44:33 - Building a detection company 47:42 - Why deepfake detection is so hard My take: detection will become internet infrastructure. But no single model will be enough Full 63-minute conversation:
5
12
403